Harvard Law Review Harvard Law Review Harvard Law Review

Administrative Law


Responding to Jeffrey A. Pojanowski, Neoclassical Administrative Law

Professor Jeff Pojanowski offers an elegant legal framework1×1. For the notion of a legal framework, see Adrian Vermeule, Chevron as a Legal Framework, Jotwell (Oct. 24, 2017), https://adlaw.jotwell.com/chevron-as-a-legal-framework [https://perma.cc/LP3T-F22E]. with roots in the history of American public law, embodied most famously in Crowell v. Benson.2×2. 285 U.S. 22 (1932). This framework, “neoclassical administrative law,”3×3. Jeffrey A. Pojanowski, Neoclassical Administrative Law, 133 Harv. L. Rev. 852, 857 (2020). has two main features: de novo judicial decisionmaking on questions of law, and substantial judicial deference on discretionary questions of policy choice (in modern terms, deferential arbitrary and capricious review).4×4. Id. Pojanowski suggests that neoclassical administrative law represents a via media between recent sweeping, putatively originalist criticisms of the administrative state on the one hand, and the abnegation of law to administrative supremacy on the other.5×5. See id. at 895.

From the standpoint of its originalist critics, the administrative state’s most flagrant departures from the constitutional plan are massive grants of statutory authority to agencies (amounting to an invalid delegation of legislative authority, on the originalist view) and judicial deference to agencies on questions of law, including agency interpretations both of statutes and of their own regulations.6×6. See Jeremy K. Kessler, The Struggle for Administrative Legitimacy, 129 Harv. L. Rev. 718, 721 (2016) (reviewing Daniel R. Ernst, Tocqueville’s Nightmare: The Administrative State Emerges in America (2014)) (describing originalist critique). Pojanowski’s framework attempts to eliminate the judicial deference problem by drawing a sharp distinction between legal interpretation and policymaking, a distinction that also implies a reduction in the intensity of arbitrariness review.7×7. Pojanowski, supra note 3, at 884, 892. Admittedly, the neoclassical framework does nothing to address the delegation problem, but neither is it intended to.8×8. Id. at 900. Pojanowski’s approach is best understood as a kind of originalism tempered by an appreciation of the difference between ideal and nonideal theory, between the first-best and the attainable second-best. Pojanowski refers to this as “faint-hearted originalism,”9×9. Id. at 895 (citing Antonin Scalia, Originalism: The Lesser Evil, 57 U. Cin. L. Rev. 849, 861–62, 864 (1989)). but perhaps “second-best originalism” might be the better term.10×10. See Adrian Vermeule, The System of the Constitution 29–35 (2012) (explaining the theory of system effects and the second best). It is an approach for a world in which sweeping grants of discretion to agencies cannot be eliminated, and can only be controlled. Judges will at least vigilantly police the boundaries of law and agency “jurisdiction,” while leaving policymaking to administrative discretion.

Pojanowski’s achievement is admirable, yet I do not think the framework can ultimately succeed. Let us bracket all questions about whether neoclassical administrative law represents the best reading of the underlying legal texts and constitutional principles as a matter of first impression, and instead look to administrative law doctrine and feasible reforms of that doctrine. Because Pojanowski aims for a via media, he quite explicitly aims to advance a view that has a reasonable degree of fit with our current law as it has developed; his framework does not claim to be a radical proposal for legal revolution.

As it turns out, however, neoclassical administrative law is almost as radical as the non-fainthearted originalism to which it claims to provide a moderate alternative. Not only did the Supreme Court reject one of the central tenets of neoclassical administrative law this past Term, as we will see in a moment, but also in a larger sense the framework has already been tried and rejected by the development of our law.11×11. This is the claim of Adrian Vermeule, Law’s Abnegation: From Law’s Empire to the Administrative State (2016). Pojanowski seems to imply that the book is a proposal for how administrative law ought to be; in fact it is an interpretive argument about the current state and future direction of administrative law doctrine. In my own view, in other words, it is Pojanowski’s putative via media that is itself a departure from the mature equilibrium of the administrative state. Pojanowski’s putatively neoclassical administrative law is in fact a lightly reworked version of a classical position, one that proved deeply unstable and unworkable under the institutional conditions of the modern administrative state. In this sense, it is not obvious what exactly is “neo” about his position. (Hence my title.)

The point comes in narrower and broader versions, focusing respectively on current doctrine and on the larger path of the law since the 1930s. As for current doctrine, the main appearance of neoclassical administrative law in the 2018 Term was a footnote to Justice Thomas’s dissenting opinion in Department of Commerce v. New York,12×12. See Dep’t of Commerce v. New York, 139 S. Ct. 2551, 2578 n.3 (2019) (Thomas, J., dissenting) (stating that under the APA, “[d]eferential review of the agency’s discretionary choices and reasoning under the arbitrary-and-capricious standard stands in marked contrast to a court’s plenary review of the agency’s interpretation and application of the law”). This is almost a précis of Pojanowksi, and the majority opinion in effect rejected both halves of Justice Thomas’s formulation. a decision that in effect took the opposite of Pojanowski’s approach by affording the Department broad statutory authority while closely scrutinizing its policy choices for pretext. So too, the Court dealt a serious blow to neoclassical administrative law by reaffirming, in Kisor v. Wilkie,13×13. 139 S. Ct. 2400 (2019). deference to agency interpretations of their own regulations.14×14. See Auer v. Robbins, 519 U.S. 542 (1997); Bowles v. Seminole Rock & Sand Co., 325 U.S. 410 (1945). Kisor is squarely inconsistent with the neoclassical approach; it amounts to a bad setback for those who would eliminate all agency deference from administrative law, whether deference to agency interpretations of their own rules or deference to agency statutory interpretations.

Pojanowski’s treatment of Kisor is unconvincing. He emphasizes that authority-based deference under Auer v. Robbins,15×15. 519 U.S. 542. as clarified in Kisor, may largely overlap, in practice, with deference based on the plausibility of the agency’s reasoning under Skidmore v. Swift & Co.16×16. Skidmore v. Swift & Co., 323 U.S. 134, 140 (1944) (noting that under certain circumstances, agency interpretations may have the “power to persuade, if lacking power to control”); see Pojanowski, supra note 3, at 888. Whatever the precise scope of these doctrines, however, the larger significance of Kisor is that it leaves embedded firmly within the law a principle of authority-based deference on legal questions. That principle is squarely inconsistent with neoclassical administrative law, and lies around like a loaded gun, just waiting to be picked up by future lawyers and judges when deference strikes them as useful.17×17. Cf. Korematsu v. United States, 323 U.S. 214, 246 (1944) (Jackson, J., dissenting) (explaining that some legal principles can “lie[] about like a loaded weapon ready for the hand of any authority that can bring forward a plausible claim”).

The result in Kisor also bodes extremely ill for the project of overruling Chevron18×18. Chevron U.S.A. Inc. v. Nat. Res. Def. Council, Inc., 467 U.S. 837 (1984). deference, much discussed and much hoped-for in certain conservative-libertarian legal circles.19×19. In that sense, the outcome in Kisor was of a piece with the outcome in Gundy v. United States, 139 S. Ct. 2116 (2019), in which five Justices voted, albeit for different reasons, to reject a nondelegation challenge that had been widely anticipated to reinvigorate that shadowy doctrine. Id. at 2121; id. at 2131 (Alito, J., concurring); see also Paul v. United States, 140 S. Ct. 342, 342 (2019) (Kavanaugh, J., statement respecting the denial of certiorari) (stating a willingness to consider future nondelegation challenges). For my reasons for skepticism that another such challenge will succeed anytime soon, see Adrian Vermeule, Never Jam Today, Yale J. on Reg.: Notice & Comment (June 20, 2019), https://yalejreg.com/nc/never-jam-today-by-adrian-vermeule [https://perma.cc/452W-EWBM]. A more likely future will see the majority invoking the nondelegation doctrine solely as a narrowing canon at the level of statutory interpretation. The basis for the Chief Justice’s controlling concurrence in Kisor was longstanding precedent, and although he carefully reserved the Chevron issue as distinct,20×20. Kisor v. Wilkie, 139 S. Ct. 2400, 2425 (2019) (Roberts, C.J., concurring in part). it is highly likely that if Auer cannot be overruled, Chevron won’t be either. Before Kisor, conservative-libertarian legalists saw the overruling of Auer as the first and easier step on the path to overruling Chevron. Having clearly failed at Step One, as it were, it is unreasonable to expect success at Step Two. Chevron may well be increasingly cabined in various ways, but there too, narrowing dispositions will leave the basic principle of deference embedded in the law, ready to be employed as and when the Justices please.21×21. Pojanowski points to several other cases to support a picture of the Court claiming to firm up restraints on the administrative state, such as Lucia v. SEC, 138 S. Ct. 2044 (2018), and Free Enterprise Fund v. Public Company Accounting Oversight Board, 561 U.S. 477 (2010). See Pojanowski, supra note 3, at 900. But these involve the internal structure of the executive branch, rather than the relationship between the executive branch and the courts directly; hence they are orthogonal to the major features of neoclassical administrative law, and to our disagreements. (Indeed, Free Enterprise Fund is a welcome development that I take to be consistent with my account in Law’s Abnegation; it puts in place a new modest limit on the administrative state, but in favor of presidential power rather than legalized judicial control.) The fact remains that the Court’s important recent pronouncements squarely relevant to the main features of the neoclassical framework — Kisor on judicial review of law, and Department of Commerce on arbitrary and capricious review — both came out the wrong way for Pojanowski.

The problem with Pojanowski’s framework is more than (although it is not less than) a question of understanding recent doctrine. It is also a much larger question of feasible paths for the law, under the real constraints of time, information, and capacities that afflict judges. Why would neoclassical administrative law prove any more stable, over time, than did its classical counterpart? The main thing to observe about the Crowell framework, the major doctrinal inspiration for neoclassical administrative law, is that it began to come undone within about a decade of its creation in 1932. By 1943, in NLRB v. Hearst,22×22. 322 U.S. 111 (1944). the Court was speaking of deference to agency interpretations with a “reasonable basis in law,”23×23. Id. at 131. as well as according deference to agency determinations of fact and policy under the substantial evidence test.24×24. See id. at 131–32. After a period of intermittent and frankly inconsistent opinions, the Court coalesced around deference to agency interpretations of law in Chevron in 1984.25×25. See Chevron U.S.A. Inc. v. Nat. Res. Def. Council, Inc., 467 U.S. 837, 844 (1984). That deference, which is antithetical to the letter and the spirit of Crowell’s classical administrative law, has remained the law to date, albeit with a number of important modifications and vicissitudes.

Pragmatically speaking, again bracketing all the arguments about whether deference to agency interpretations of law is consistent with the text, structure, and original understanding of the APA and with the Constitution, this sort of breakdown of (neo)classical administrative law seems entirely predictable. The reasons for this spillover of deference from questions of policy and fact to questions of law are not mysterious.

First is the notorious slipperiness, especially for real-world judges, of the distinctions among questions of “law,” “fact,” and “discretion” or “policy,”26×26. The Administrative Procedure Act distinguishes “issues” of “fact, law, or discretion.” 5 U.S.C. § 557(c)(3)(A) (2018). at least in hard cases. In Hearst, the problem involved a so-called “mixed” question of law and fact: whether to classify newsboys as “employees” or as independent contractors within the terms of the National Labor Relations Act.27×27. 29 U.S.C. §§ 151–169 (2012); see Hearst, 322 U.S. at 120. Hearst perfectly exemplifies the Court’s inability, over time, to sustain a clear distinction between “pure” and “mixed” questions of law in controversial cases at the moving frontier of administrative law. Is the question presented there one of “law,” “fact,” or “policy”? In truth it is all three, inextricably and simultaneously, and even if, by elaborate analytic argument, one could disentangle all the components, federal judges lack the time and inclination for elaborate analytic argument. The same is true for the many statutes that require or authorize agencies to take “reasonable,” “appropriate,” or “necessary” action; it is chronically true in such cases that lines between law, fact, and policy discretion are uncertain and unstable.

The acknowledgment that such distinctions are not stable or tenable has been reflected in several strands of precedent, which together highlight the difficulty of drawing Pojanowski’s sharp distinction between review of legal questions, on the one hand, and review of facts and discretionary policymaking, on the other. One example, following directly on Hearst, has been the ongoing instability in practice of the distinction between so-called “pure questions of law” and “mixed questions of law and fact.”28×28. Compare INS v. Cardoza-Fonseca, 480 U.S. 421, 446–48 (1987) (distinguishing “a pure question of statutory construction for the courts to decide,” id. at 446, from a “question of interpretation . . . in which the agency is required to apply [a legal standard] to a particular set of facts,” id. at 448), with id. at 454–55 (Scalia, J., concurring in the judgment) (observing that the distinction is inconsistent with Chevron itself); cf. Nat’l Cable & Telecomms. Ass’n v. Brand X Internet Servs., 545 U.S. 967, 986–97 (2005) (finding Chevron deference applies to agency’s pure construction of a statutory definition). The instability of the distinction, which is far easier to state than to apply, undermines the sharp distinction between legal and nonlegal questions on which the (neo)classical framework is constructed.

Similarly, in City of Arlington v. FCC,29×29. 569 U.S. 290 (2013). Justice Scalia wrote for a majority (including Justice Thomas) to reject the idea that it is even coherent to draw a distinction between questions of agency “jurisdiction” and other questions of law bearing on agency authority.30×30. Id. at 297–98. This too is theoretically crucial, insofar as the category of agency “jurisdiction” was a centerpiece of the Crowell framework.31×31. See Crowell v. Benson, 285 U.S. 22, 54–55 (1932). Pojanowski, aware that this is a problem for his view, denies that he means to revive the “jurisdiction” exception, but then immediately says that courts should decide de novo the “scope” of the agency’s authority.32×32. Pojanowski, supra note 3, at 807. But this is a semantic distinction without a difference — and indeed it is a distinction specifically rejected by the Court, which defined questions of agency “jurisdiction” as questions about the scope of agency authority.33×33. City of Arlington, 569 U.S. at 293 (“We consider whether an agency’s interpretation of a statutory ambiguity that concerns the scope of its regulatory authority (that is, its jurisdiction) is entitled to deference under [Chevron].”).

Finally, consider the question of which factors are relevant to agency policy choice — under Citizens to Preserve Overton Park, Inc. v. Volpe,34×34. 401 U.S. 402 (1971). the first question in arbitrariness review.35×35. Id. at 416. The Court has generally said, quite sensibly, that if Congress has clearly ruled factors in or out, agencies must respect that decision, but that if Congress has been silent or ambiguous, agencies have discretion to decide which factors are relevant.36×36. See, e.g, Chevron U.S.A. Inc. v. Nat. Res. Def. Council, Inc., 467 U.S. 837, 842–43 (1984). For an excellent overview, see generally Richard J. Pierce, Jr., What Factors Can an Agency Consider in Making a Decision?, 2009 Mich. St. L. Rev. 67. The Court, in other words, has treated the first step in arbitrariness analysis as itself a Chevron question,37×37. See Pension Benefit Guar. Corp. v. LTV Corp., 496 U.S. 633, 647–48 (1990) (using the Chevron framework). partially collapsing the inquiries that Pojanowski would cleanly separate. This creates a serious dilemma for neoclassical administrative law. If statutes are what make relevant factors relevant, and if courts are to determine all statutory questions de novo, then courts must decide for themselves, all things considered, what factors agencies may, may not, or must consider when making policy choices. But that would hardly seem to produce the sort of deferential review of policymaking Pojanowski recommends. The neoclassical framework is either internally inconsistent, or else it must abandon even the Overton Park framework and all the law that has flowed from it — a much more radical enterprise than the advertisement for neoclassical administrative law lets on.

I turn now to a distinct problem, although one that is related to the instability of law-fact-policy distinctions, especially as applied to complex modern statutes: the jurisprudential problem of legal realism. It is no accident that realism arises and flourishes roughly in conjunction with the growth of the administrative state, and with the spread of legal forms and instruments that challenge common law categories. The context of statutory delegations to administrative agencies tends to inspire the thought that “interpretation” is often, at least or especially in hard cases, an exercise in discretionary policymaking. Chevron itself is an obvious example: Was the validity of the “bubble concept” really a legal question in any sense that a nineteenth-century judge would recognize? Pojanowski says that “[e]ven if the law underdetermines a small fraction of the litigated cases posing legal questions, it does not follow that we should structure the entire system of judicial review based on those exceptional cases.”38×38. Pojanowski, supra note 3, at 905. The problem is that the “exceptional cases” do not come neatly prelabeled as such. Especially as one moves higher on the appellate ladder, it will become increasingly plausible for judges to argue about whether statutes are or are not ambiguous, and one ends up with the spectacle — often on display at the Supreme Court — of two groups of Justices of basically equal size, each arguing vehemently that the statute “clearly” favors their own view.39×39. See Eric A. Posner & Adrian Vermeule, The Votes of Other Judges, 105 Geo. L.J. 159, 164 (2016). In such a world, lawyers naturally begin to conclude that both groups are wrong to insist woodenly that the statute has a single determinate meaning (on which they differ), that in fact the statute is ambiguous, and that the tools of (neo)classical interpretation are simply inadequate to settle the issue.40×40. See Jacob E. Gersen & Adrian Vermeule, Chevron as a Voting Rule, 116 Yale L.J. 676, 679–80 (2007).

The problem of irreducible ambiguity in hard cases is exacerbated by the phenomenon of “old statutes, new problems.”41×41. Jody Freeman & David B. Spence, Old Statutes, New Problems, 163 U. Pa. L. Rev. 1, 1 (2014). As the administrative state confronts new problems and challenges under old statutes — statutes, like the Clean Air Act or the immigration laws, that a politically polarized and fractured Congress rarely updates, but that are expected to govern ever-changing problems42×42. See id. at 8–11. — it becomes less and less plausible to insist that statutes provide a single right answer, no matter what problems arise that were completely unforeseen by the statutes’ drafters. Hence the marked tendency for old-framework statutes and quasi-constitutional statutes is to become essentially common law constitutions governing part of the administrative state, and develop over time by interaction between changing agency interpretations and more or less deferential judicial review.

In this perspective Crowell, penned by a Chief Justice born in 1862, represents a kind of holdover from the world of classical legal thought, one that almost immediately broke down in the face of developing conditions. Chevron itself is, from a jurisprudential standpoint, best understood as a product of a limited form of legal realism, one that understands that when agencies “interpret” statutes like the Clean Air Act over time in changing unforeseen circumstances, giving rise to hard cases, those agencies will inevitably be faced with policy choices, whose resolution is not obviously better entrusted to a generalist and unaccountable judiciary. Put differently, the very point of Chevron is to articulate a conception of interpretation that opens up a “policy space” for agency discretion, as opposed to the attempt of classical legal interpretation to reduce statutory meaning to a single point.43×43. E. Donald Elliott, Chevron Matters: How the Chevron Doctrine Redefined the Roles of Congress, Courts and Agencies in Environmental Law, 16 Vill. Envtl. L.J. 1, 11–12 (2005).

Pojanowski argues that “[t]he structure of Chevron itself rests on pre-legal realist assumptions that pragmatists and supremacists ostensibly reject,” because at Step One the judge decides whether the statute is clear, and “[t]o stipulate that a question can be clear or not presupposes a stable measure with which to judge clarity.”44×44. Pojanowski, supra note 3, at 906. But this is not at all the prerealist legal approach to interpretation, which asked (barring special cases like mandamus, arguably the origin of Chevron itself45×45. See United States v. Mead Corp., 533 U.S. 218, 241–43 (2001) (Scalia, J., dissenting). ) not whether statutes were “clear” but simply what the best interpretation was, all things considered. Hearst and Chevron mark a fundamental conceptual break with this regime by introducing the supposition that in some range of cases, an administrative agency may reasonably disagree with the court’s judgment about what interpretation is best, and that where such reasonable disagreement occurs, the agency will prevail. That break with the past cannot be minimized, and once it has occurred, it is probably impossible to return to the belief structures of the old world by brute force, any more than we could induce in ourselves an unironic belief in the four humors of Hippocratic medicine.

Pojanowski never focuses on the pragmatic impossibility of truly independent judicial analysis of highly complex modern statutes, whose interpretation carries enormous policy consequences, by judges laboring under realistic constraints of time and expertise. The canonical illustration came during the Court’s internal deliberations on Chevron itself. According to Justice Blackmun’s papers, Justice Stevens, the decision’s author, said at conference on the case: “When I am so confused, I go with the agency.”46×46. William N. Eskridge Jr. & John Ferejohn, A Republic of Statutes 277 (2010). This is an entirely rational decisionmaking strategy by generalist judges who face intricate, specialized regulatory statutes, who know the limits of their own knowledge, and who know that the consequences of a judicial blunder may be extremely serious. Even more importantly, this sort of deference is difficult to control by formal legal doctrine. Any Supreme Court decision, statute, or even constitutional amendment would be largely helpless to stamp out this sort of epistemological deference, which simply operates behind the scenes, in the judge’s internal deliberative processes. The choice is not between deference and no deference; it is between open and hidden deference.

The reasons for the instability of the classical Crowell framework carry over to the very similar neoclassical framework. Indeed, it is not obvious what is particularly “neo” about Pojanowski’s approach, which presupposes essentially the same prerealist distinction between legal interpretation and legal policymaking on which Crowell rested. Pojanowski tries to distinguish the two by saying, counterintuitively, that Crowell’s sharp law-fact distinction was insufficiently formalist, because courts of the era engaged in purposive legal interpretation, whereas “[t]he neoclassicist’s legal formalism . . . marks a return to pre-legal realist thought.”47×47. Pojanowski, supra note 3, at 898. Here is the full passage:

Crowell’s distinction between review of law and policy was unstable only so long as it rested on the interpretive antiformalism that dominated at the time of the New Deal and the subsequent Legal Process era. The neoclassicist’s legal formalism, however, marks a return to the pre–legal realist thought that, while aware of the blurriness in the lines between making, executing, and interpreting law, nevertheless insist[ed] that the division of these activities was coherent in theory and estimable in practice. To be sure, the tenability of such a classical approach to the legal craft in a post-realist world is an important challenge neoclassical administrative lawyers must address. But if it stands, the theory has better resources to patrol the line between law and policy than the strong purposivists who founded — and lost — the Crowell regime.48×48. Id. (internal quotation marks omitted) (alteration in original). As I say in text, what is unexplained here is how the “insistence” on the “division” between law and policy somehow becomes more tenable after legal realism than before. One cannot force oneself to believe things by fiat. Even accepting the premise that interpretation in the Crowell era was highly purposivist and antiformalist — a view that basically conflates the 1930s with the Legal Process approach of the 1950s49×49. See id. (conflating “the time of the New Deal” with “the subsequent Legal Process era”). — there can be no such return of the sort Pojanowski desires. Once the apple of realism has been tasted, everything changes, and the way back to the garden of naive classicism is forever barred. It is not possible to reinstate belief in a classical law-policy distinction by fiat, however useful the resulting framework would be, even as a kind of noble lie. To attempt to return to (a more formalist version of) the Crowell framework would, at best, merely recreate the adjudicative conditions and intellectual difficulties that led to the collapse of that framework in the first place.50×50. A claim fleshed out at length in Vermeule, supra note 11. As Valéry Giscard d’Estaing said in a different context: “There is no question of returning to the pre-1968 situation, if only for the reason that the pre-1968 situation included the conditions that led to 1968.”51×51. Id. at 42 (quoting Jon Elster, Sour Grapes: Studies in the Subversion of Rationality 87 n.5 (1983)).

Overall, Pojanowski’s proposal, whatever its abstract merits, is implicitly far more radical than it claims to be. It is out of step with too much doctrine, practice, and history; and lays out no feasible path for the law. Administrative law cannot go home again, even assuming home lies in the direction to which Pojanowski points. There is much to admire and to learn from in Pojanowski’s article. But there is not much that is truly “neo” in it, and a return to classical administrative law, no matter how ardently desired, is not a realistically possible future.


*Ralph S. Tyler Jr. Professor of Constitutional Law, Harvard Law School.