Since the dawn of Chevron, people have been arguing that Chevron's second step is or should be hard-look, arbitrary and capricious review. Some of the arguments for this move are rather prosaic. First, Chevron itself comes close to saying that Step Two is arbitrary and capricious review ("legislative regulations are given controlling weight unless they are arbitrary, capricious, or manifestly contrary to the statute."). Second, equating Step Two with arbitrary and capricious review is said to square Chevron with the Administrative Procedure Act and thereby legitmate Chevron; Justice Thomas' important concurring opinion last week in Cuozzo is an example of this view.
Third, as Ronald Levin first argued, Chevron and the Court's subsequent opinions on Chevron says that Chevron has two steps, not one, and absent equating Step Two with arbitrary and capricious review, it's very difficult to see how Step Two isn't a reprise of Step One. Courts that decline to inquire into the policy reasoning an agency gives for choosing one of several possible meanings of a statute have nothing much to do at Step Two besides repeat their Step One conclusion that the statute doesn't unambiguously foreclose the agency's reading. For a randomly chosen but poignant example of this redundancy, see the D.C. Circuit's recent opinion in Van Hollen v. FEC, a case in which that court reversed a district court's Step One decision for the challenger, remanded the case to the district court to consider Step Two, and on further appeal from the remand, reasoned that its prior Step One holding "largely foreordains our . . . Step Two answer," because the court had already determined that the agency's reading "was within the range of linguistically possible constructions" -- a determination which the court treated as sufficient to decide Step Two, even though such a determination should always be required for an agency to survive Step One. (There are courts who only ask at Step One whether a statute can bear multiple meanings, without deciding whether it can bear the agency's; this saves Step Two from redundancy at the cost of making Step One a pointless exercise.)
Though it's virtually undeniable that only arbitrary and capricious review can make Chevron non-redundantly live up to its purported two-step framework (but see Richard Re's novel suggestion that Step Two, instead of being a space for arbitrary and capricious review, could non-redundantly determine whether an agency's reading is compelled), some scholars have more ambitiously argued that regardless of the number of steps Chevron has, Chevron deference is only justifiable if arbitrary and capricious review is a part of it. Chevron, the argument goes, famously claims that in cases of statutory indeterminacy, courts should defer to agencies' policy choices rather than making their own, because of agencies' superior political accountability and expertise. If that's the rationale for Chevron deference, though, why should a court defer to an agency that doesn't purport to be making a policy choice, but relies only on interpretive claims about statutory meaning, or an agency that purports to be making a policy choice as between permissible interpretations of a statute but does so in an arbitrary way? Arguments of this kind have been made, at various times, by Michael Herz, Peter Strauss, Kenneth Bamberger and Strauss, Jeffrey Pojanowski and Randy Kozel, Richard Murphy, and many others, and I agree with them wholeheartedly.
2. How Chevron theory's success, as theory, caused its operational failure.
This theory of Chevron was substantially victimized by its own success. Courts and commentators bought in so heavily to Chevron's prescriptive claims on why courts should defer to agencies that they came to mistake them for descriptive truisms about what agencies were actually doing, or regularly observed mandates to agencies to behave in the manner that Chevron suggested they did. Very early in Chevron's history, courts largely stopped reading the regulatory documents to which they were deferring , assumed that, given Chevron's account of agency statutory interpretation, whatever was contained in them must be a deliberative policy choice of some sort, and either gave agencies deference on this premise without even doing arbitrary and capricious review, or, when they did arbitrary and capricious review at Step Two, largely turned to agency briefs for policy reasoning to review.
Though I could adduce any number of cases in which courts have unknowingly mistaken pure agency statutory interpretation for agency policymaking and deferred on that ground, two extrajudicial examples of the mystification Chevron would cause for even the leading lights of administrative law particularly strike me. The first is a personal anecdote. Once, Justice Scalia visited a seminar I was taking at Georgetown and took questions from the class. When I asked him what he would do in a Chevron case if an agency gave purely interpretive reasons in a regulation for its choice to read a statute one way or another, he tentatively ventured that Chevron deference might not be warranted in such a case, but said that he had never seen such a case in his years on the Court or D.C. Circuit. This surprised me because, as I'd recently learned in writing a note on Chevron deference to the Board of Immigration Appeals, many agencies, especially adjudicative bodies like the Board or NLRB, engage in purely interpretive reasoning more often than not when choosing what meaning to give to an ambiguous statute. I can only conclude that Scalia had come to mistake Chevron's picture of agency practice for reality.
Second, in 2005, Jerry Mashaw wrote perhaps the first major article on agency statutory interpretation and found, in a small empirical study of EPA and HHS regulations, that agencies engaged in textualist statutory interpretation all the time. Two years later, Richard Pierce, a giant of administrative law, wrote a short response that questioned the point of the whole project, disregarded, without mention, Mashaw's empirical evidence of agency practice, and claimed that when agencies "give meaning to ambiguous provisions in the statutes they administer . . . they are not involved in the process of statutory interpretation. Instead, they are engaged in a policymaking process . . . ." The only evidence cited for this empirical claim was Chevron itself. And while it's possible to read Pierce as merely making a prescriptive claim about what agencies are supposed to do under Chevron, even that overreads Chevron, which only states what courts are supposed to do when agencies respond to statutory ambiguity in certain ways, not how agencies are supposed to respond to statutory ambiguity.
3. A new age for hard-look review at Step Two?
Though three decades under Chevron's rule have passed without the Court, or D.C. Circuit for that matter, truly heeding the call to turn Step Two into a meaningful policy review of agency discretion, or even clarifying what Step Two is if not arbitrary and capricious review, in recent years there have been signs that the Court was beginning to inch towards adopting the prevailing academic view. First, in Judulang v. Holder, Justice Kagan surprisingly wrote for a unanimous Court that it didn't matter whether the Court applied Chevron to a Board of Immigration Appeals decision or arbitrary and capricious review, because Step Two review "ask[s] whether an agency interpretation is arbitrary or capricious in substance."
Second, and more tellingly, last year in Michigan v. EPA Justice Scalia would invoke State Farm arbitrary and capricious review to reject an EPA interpretation of a provision of the Clean Air Act giving the EPA power to regulate if "appropriate and necessary" that declined to consider cost. Meanwhile, Justice Kagan in dissent conceded that the EPA's interpretation of "appropriate and necessary" would have been unreasonable had it refused to consider cost, but argued that the EPA had considered cost. In an exciting working paper presented earlier this month at George Mason's Chevron conference, Catherine Sharkey reads Michigan to suggest that the present moment of Chevron retreat (though I'd call it revanchism) seen in cases like Michigan and Burwell may be finally ushering in meaningful arbitrary and capricious review at Step Two. Like Herz, Strauss, Bamberger, Pojanowski, Kozel, Murphy, and others before her, she also argues that courts should be looking for "policy-relevant evidence" at Step Two, "not [agencies'] view of the best legal interpretation." What may be new in Sharkey's paper, besides the suggestion that her view is at long last becoming doctrine, is the claim that hard look review at Step Two will not only lead courts to defer to the right agency interpretations, but cause agencies to make better decisions. I would only add, probably unoriginally, that it would also encourage agencies to disclose the real policy reasons they have for doing what they're doing, instead of hiding behind motivated reasoning regarding statutory meaning and legislative intent.
4. Why Encino is a Step Two case that matters.
Last Monday, the Court decided a pair of cases, Encino Motorcars and Cuozzo, that appeared, on first blush, to confirm Sharkey's reading of Michigan and the current moment of Chevron retreat. But before anyone on the hard-look side of the Step Two debates could claim Encino and Cuozzo as a victory, Adrian Vermeule, the leading advocate for a one-step Chevron, cast Encino as a banal non-event, while Daniel Hemel argued that Encino wasn't even a Step Two case, but was really all about Step "0.5." I think Vermeule and Hemel are wrong; Encino (and Cuozzo) isn't banal, and it is a Step Two case, even if its author didn't write it that way. But Encino, rather than ushering in the era of hard-look Step Two review so many have hoped for for so long, only holds that agency reason-giving of some kind is required at Step Two, and actually reifies the last thirty years of doctrinal indifference to, and uncuriosity about, what kinds of reasons agencies give.
Encino is a case about whether a car dealership's service adviser, who essentially sells customers car service but doesn't perform the service himself, is within the FLSA exemption to overtime compensation for a "salesman, partsman, or mechanic primarily engaged in selling or servicing automobiles." The Department of Labor determined, in a notice-and-comment rulemaking, that a service adviser was not included within that language.
With all respect to Justice Thomas' dissent, it's plain to see that the statute is ambiguous on whether a service adviser is included. An excellent argument can be made that a service adviser neither is engaged in selling automobiles, nor servicing them, though an equally reasonable argument can be made that a service adviser is engaged in servicing. Further, the government plausibly argues that the statute really means mean "salesman primarily engaged in selling, or partsman or mechanic primarily engaged in servicing, automobiles." A little-known canon of interpretation, reddendo singula singulis, suggests that the statute is best read in this way, though there are strong arguments that reddendo doesn't apply. In short, the statute is ambiguous, and whatever the Department of Labor had concluded about whether service advisers were in or out would have been a permissible reading of the statute.
However, the Department of Labor said virtually nothing about why it read service advisers out, only noting that service advisers were not explicitly in, stating that its interpretation was "reasonable," and cryptically remarking that it "recognize[d] there are circumstances under which the requirements for the exemption would not be met." (The relevant page of the rule is here.) The Department's failure to give reasons for adopting the eminently permissible interpretation it chose created the perfect test case of whether Step Two requires anything more from an agency than mere permissibility.
In an opinion by Justice Kennedy, the Court declined to defer to the Department's rule, and remanded the case to the Ninth Circuit to interpret the exemption without deference to the Department. The extent to which the Court's denial of deference turned on the agency's interpretation being a surprising shift from longstanding views is unclear; Justice Kennedy at one point suggests that less reason-giving might have been needed if not for the change in view, while Justice Ginsburg and Justice Sotomayor concur in full but claim that the change was not a reason, or at least not much of a reason, for the Court's demanding heightened reason-giving in this case.
I am not very interested in that question. What is important is that, for apparently the first time, the Court denied deference to an agency's interpretation of confessedly ambiguous statutory language, contained in a formal document that satisfied Mead's safe harbors, because the agency failed to give reasons for picking the permissible interpretation it chose. Vermeule says this is banal; arbitrary regulations are invalid, so of course, he reasons, they get no deference. Indeed, Encino should be banal. But until last week, the Court had never held that arbitrary regulations get no deference (though it had said it in cases like Judulang and Chevron itself), nor had it held that a regulation that arrives at a permissible reading of a statute is arbitrary and invalid because it gave no reason for picking the permissible reading it picked. To illustrate how important and novel Encino is, consider that in Burwell, the government seriously urged, and the Fourth Circuit gave, Chevron deference to a regulation that addressed whether federal exchanges were exchanges "established by the State" as follows (and that the Court only denied deference to that regulation on novel grounds which had nothing to do with the quality of the agency's reasoning):
Hemel, for his part, claims that in spite of Encino's four citations to State Farm, repeated criticisms of the agency for failing to produce a "reasoned explanation" of its actions, and the Court's description of the agency's rule as "an arbitrary and capricious regulation," Encino isn't a Step Two or State Farm case. Part of his argument is purely circular, or question-begging; he claims that Encino isn't a State Farm case because State Farm has nothing to do with agency statutory interpretation. Once one acknowledges that courts can, in fact, apply arbitrary and capricious review to agencies' policy choices to pick one permissible reading of a statute over others, this argument collapses.
The statutory language of section 36B and other provisions of the Affordable Care Act support the interpretation that credits are available to taxpayers who obtain coverage through a State Exchange, regional Exchange, subsidiary Exchange, and the Federally-facilitated Exchange. Moreover, the relevant legislative history does not demonstrate that Congress intended to limit the premium tax credit to State Exchanges. Accordingly, the final regulations maintain the rule in the proposed regulations because it is consistent with the language, purpose, and structure of section 36B and the Affordable Care Act as a whole.
More substantively, Hemel argues that Encino isn't a Step Two case, but is rather a Step Zero case, because Justice Kennedy emphatically declines to do Chevron at all, to the point where he declines to consider whether the statute is ambiguous at Step One. This is probably formally correct; Justice Kennedy doesn't do Step One, says at multiple points that Chevron deference doesn't apply to the agency's regulation, and at times appears to suggest that the rule he's announcing is a wrinkle of Mead, the Court's classic (or infamous) case on Step Zero. What Hemel misses, though, is that a Step Zero holding hinged solely on how an agency reasoned in a particular instance, rather than on whether a statute delegates gap-filling power to an agency if it uses notice-and-comment rulemaking or formal adjudication and reasons in a non-arbitrary fashion, is functionally equivalent to a Step Two holding. Unlike a "true" Step Zero holding, like Burwell, which would deny the agency deference on a particular issue or under a particular statute no matter the quality of its reasoning or the formal vehicle in which that reasoning appeared, Encino denies the Department of Labor Chevron deference only so long as it reasons arbitrarily; a non-arbitrary regulation on the same subject would receive Chevron deference. Had the Court styled its holding as a Step Two holding, the consequences for the Department of Labor would be exactly the same; the particular rule under review would not receive deference, but a different rule on the same question might.
Thus, by doing arbitrary and capricious review (which the Court is clearly doing, to an extent) at Step Zero, the Court has essentially acquiesced to the calls to do arbitrary and capricious review at Step Two. The only difference is the lack of a Step One ambiguity determination, which deprives the agency of information as to whether its prior reading was at least permissible and will survive review if re-arrived at in a non-arbitrary way. I would predict that in future cases, Encino will be applied as if it were a Step Two holding in order to avoid this problem.
5. Why Encino is not the advent of real hard-look review at Step Two, and why the Court doesn't want it to be.
But while Encino is not banal, nor a strange new permutation of Mead, but rather is an important new development on Step Two or the functional equivalent thereof, it is not the revolution in Step Two arbitrary and capricious review that one might have hoped for. The key sentence in the opinion isn't one of its many citations to State Farm, or its announcement that arbitrary and capricious regulations receive no Chevron deference, but this:
Although an agency may justify its policy choice by explaining why that policy “is more consistent with statutory language” than alternative policies, Long Island Care at Home, 551 U. S., at 175 (internal quotation marks omitted), the Department did not analyze or explain why the statute should be interpreted to exempt dealership employees who sell vehicles but not dealership employees who sell services (that is, service advisors).Here, the Court makes clear that the Department of Labor could have received Chevron deference had it only essayed a thumbnail sketch of the erudite arguments from reddendo singula singulis found in the SG's brief. Had the agency only said that the statute is ambiguous, but that the more ordinary reading of "engaged in servicing automobiles" excludes persons who don't actually service automobiles, or that the statute is ambiguous, but that a hoary canon of construction suggests that "salesman" should be read to be the subject of "selling," and only "selling," while "partsman or mechanic" should be read to be the subjects of "servicing," the Court would have deferred, so long as it agreed with the agency on ambiguity.
This is not real hard look review of the sort that people like Sharkey are calling for at Step Two. Rather, it is a minimal filter for agency interpretations that fail to give non-arbitrary reasons of any kind. 32 years after Chevron was decided, the Court still doesn't understand - or understands but refuses to admit - that there are no good reasons for mandatory deference (as opposed to contingent, Skidmore deference) to agency regulations that adopt a particular interpretation of a statute because the agency says the statute is textually best read that way. 32 years after Chevron was decided, the Court still doesn't understand - or understands but refuses to admit - that Chevron only gives agencies deference on the premise and condition that they do not interpret statutes that way, but instead make policy choices that call on expertise, public opinion, and presidential preferences.
Encino's indifference to the kinds of reasons agencies give was presaged at oral argument in a remarkable colloquy between Paul Clement and Justice Kagan, and confirmed last Monday in Cuozzo, which is a familiar exercise in judicial denialism and legal fiction-making about the realities of agency statutory interpretation. At oral argument in Encino, Clement attempted to argue that the Department of Labor shouldn't receive Chevron deference because it had solely addressed what it thought the statute meant, rather than what, given its ambiguity, the statute should be read to mean. He was immediately shut down:
JUSTICE KAGAN: ... I mean, agencies do this all the time. They say, this is the way we read the statute. This is the way we want to read the statute. And this Court has never been in the business of saying, oh, when you think that the statute says something, you don't get deference. Whereas when you think the statute is ambiguous but you give other reasons, you do get deference. I mean, that would be a a completely unadministrable line to use.This pithily sums up the two reasons the Court has probably never been inclined to separate the interpretive chaff from the policymaking wheat at Step Two: administrability, and the suspicion that when an agency says "this is the way we read the statute," it's because that's the way it wants to read the statute. Administrability is a legitimate concern. As I'll show below in connection with Cuozzo, it can be difficult to tell if an agency is doing statutory interpretation or policy reasoning, as all statutory interpretation is, to a certain extent, policy reasoning. Often the reason a court or agency will believe that Congress intended something is that it seems a better policy than the alternatives, and agencies are arguably entitled to Chevron deference when assessing what interpretations are better policies than others, even when they're only doing so (or say they're only doing so) to determine what Congress meant.
MR. CLEMENT: Well, I don't -- I don't want to quibble too long with you on that. I don't think that would be unadministrable at all. This Court has recognized the exact same principle in the Auer context, which is to say that if all you do is parrot the statute, then that doesn't really gain you any extra deference, and I think all --
JUSTICE KAGAN: This is not parroting the statute. This is saying we read this statute in a way which we think is better than another way.
I think this is plausible, but incorrect; the kind of deference owed to an agency claim that a policy is better and therefore the one Congress likely preferred is an epistemological deference, based, contingently, on how much the agency actually knows about what's better in the area, whereas the kind of deference owed to a true agency choice of one policy over another is a categorical deference, premised on the claim that agencies are more accountable than courts and more constitutionally licensed to make policy. And while a distinction between purely interpretive and policy reasoning might be unadministrable, the proper distinction between agencies that profess to say what Congress meant and agencies that profess to be making policy choices is probably much easier to draw.
Justice Kagan's second argument is weaker. It is doubtless the case that when an agency claims to have arrived at the best estimate of statutory meaning, it often, even usually, has arrived at the reading it wanted to arrive at for unstated policy reasons (including the President's actual or presumed preferences). Why, though, should agencies not be forced to state those reasons, which may, for all the courts know, be completely arbitrary? And why should courts defer to just any old agency policy choice, so long as it's permitted by a statute? While courts should generally defer to agencies' permissible policy choices, they shouldn't, and can't under the APA, defer to arbitrary ones, and it would seem that much could be gained, in the way of transparency, accountability, and the quality of agency reasoning by making agencies state their real policy reasons instead of hiding behind lawyerly exegeses of their preferred readings.
6. Cuozzo's confirmation of Encino as (basically) business as usual
In Cuozzo, the Court confirmed that it will still defer to more-or-less purely interpretive agency reasoning so long as it's reasoned, and that it will continue to perpetuate the Chevron myth of agencies as non-interpretive policymakers by recasting agency reasoning and performing arbitrary and capricious review on policy reasoning that an agency never gave. Cuozzo concerned the claim construction standard the Patent Office should use in "inter partes review," a procedure by which a third party can ask the Patent Office to reexamine and invalidate a patent. The statute creating inter partes review was silent on the matter; the Patent Office chose to use the standard it uses when reviewing patent applications in the first instance, namely the broadest reasonable construction standard, which tends to invalidate more patents than the standard the courts use in infringement litigation. Challengers to the regulation argued that inter partes review was intended to be a stand-in for judicial proceedings, and that the standard used in those proceedings should be used in inter partes review.
In adopting the broadest reasonable interpretation standard in inter partes review, the Patent Office gave a series of basically interpretive reasons, only two of which the Court mentioned. First, the Patent Office noted that the Federal Circuit had required it to use the broadest reasonable construction standard in patentability determination proceedings for thirty years.
Second, the Patent Office argued that the interpretive reasons the Federal Circuit had given for requiring the broadest reasonable construction standard in patentability proceedings - the preponderance of the evidence standard for patentability, the ability to amend claims, and the lack of a presumption of validity - all were present in the new statute's treatment of inter partes review, whereas different burdens of proof apply in litigation, where a more patent-friendly claim construction standard is used. Therefore, "[t]he provisions of the [statute] indicate[d] that the typical standard . . .should apply" to inter partes review.
Third, the broadest reasonable interpretation standard was consistent with the legislative history, "which indicates that Congress was aware of the . . . standard and expected the Office to apply the standard to the new . . . review proceedings." Further, had Congress intended a new standard, it could have enacted one, but it didn't.
Fourth, the statute allowed the Office to consolidate inter partes review with certain other proceedings in which the broadest reasonable interpretation standard had long applied, and ["i]t would be anomalous for the [Office] to have to apply two different standards in the merged proceeding."
Fifth, inter partes review, as a statutory matter, turns out to be about patentability, the traditional Patent Office question, not validity, the traditional judicial question, and "[t]hat distinction confirms Congress' intent for the USPTO to apply the typical framework it currently applies in existing patentability determinations." Finally, the Office concluded that a single claim construction standard must be used to "prevent inconsistencies and inefficiencies" in a merged proceeding.
The Court deferred to the Patent Office after performing a sufficiently convincing facsimile of arbitrary and capricious review that Justice Thomas announced his constitutional concerns about Chevron had been cured, at least for the moment, by the Court's obesiance to the APA. According to the Court, "neither the statutory language, its purpose, or its history suggest that Congress considered what standard the agency should apply when reviewing a patent claim in inter partes review" - even though the regulation it deferred to repeatedly said that the statute's language, purpose and history suggested that Congress considered the question and intended the Office to use the broadest reasonable construction standard. According to the Court, the Patent Office made a "decision to prefer a degree of inconsistency in the standards used between the courts and the agency, rather than among agency proceedings." That decision, the Court says, was a reasonable choice for several reasons:
First, the broadest reasonable construction standard helps to protect the public from bad patents - a reason the Patent Office never gave or gestured at.
Second, here citing the agency rule, the Court notes that "past practice supports the Patent Office's regulation," which is the Court's repackaging of the Patent Office's claim that three decades of Federal Circuit jurisprudence more or less mandated that regulation.
Third, while the rule's challengers argue that the Patent Office's standards will create inconsistency between the Patent Office and subsequent or preceding litigation, different burdens of proof apply in the two, "mean[ing] that the possibility of inconsistent results is inherent to Congress' regulatory design" - a repackaging of the Patent Office's claim that the differing burdens of proof indicated as a matter of statutory interpretation that the broadest reasonable construction standard must be used in inter partes review.
Fourth, the Court repurposed the point about merger of proceedings, which in the Patent Office's version was a claim that the anomalies created by the rule's critics' position suggested their interpretation was incorrect, into an argument that it was reasonable for the Patent Office to decline to adopt the challengers' "difficult" rule.
Finally, the Court deferred to the Patent Office's supposed "prefer[ence]" for consistency between agency proceedings, concluding that "whether there is a better alternative as a policy matter" was left by Congress "to the particular expertise of the Patent Office."
In fairness to the Court, parts of my reading of the agency rule are tendentious. While the Office makes repeated claims about "Congress' intent" and what "Congress expected," it is certainly possible to read the Office's point about anomalies flowing from the challengers' preferred rule as pure policy reasoning, rather than as claims that Congress couldn't have intended for their rule. It may even be possible to read the Office's point about burden of proof and the like as an argument that, for the same reasons the Federal Circuit had thought the broadest reasonable construction standard made sense in patentability proceedings, it also made sense in inter partes review. The difficulties of classifying these arguments as instances of interpretive reasoning or freeform policymaking point up the potential unadministrability of my preferred approach to Chevron.
However, what is clear is that the Court deferred to a regulation that by the Court's own lights is erroneous, as it claims the statute has a precise meaning and definite intention on claim construction standards that the Court says isn't there. In order to justify deferring to that regulation, the Court is compelled to conjure up an exercise in policy choice that simply doesn't exist. And though it's certainly possible that the Court's imaginative reconstruction of what the Patent Office was thinking is entirely accurate, it's also possible that the Court deferred to a "choice" that really was the Patent Office's sincere estimate of congressional intent. Whichever is the case, so long as the Court declines to require that, or inquire into whether, agency statutory interpretation live up to Chevron's picture of what agencies do, we will continue to live with fictionalized renderings of agency statutory interpretation.