When a court determines that an agency action violates the Administrative Procedure Act, the conventional remedy is to invalidate the action and remand to the agency. Only rarely do the courts entertain the possibility of holding agency errors harmless. The courts’ strict approach to error holds some appeal: Better a hard rule that encourages procedural fastidiousness than a remedial standard that might tempt agencies to cut corners. But the benefits of this rule-bound approach are more elusive, and the costs much larger, than is commonly assumed. Across a wide range of cases, the reflexive invalidation of agency action appears wildly excessive. Although the adoption of a context-sensitive remedial standard would increase decision costs and generate inconsistency, the exercise of remedial restraint in appropriate cases may prove superior to a clumsy approach that treats every trans­gression as worthy of equal sanction.


    1. The Rule of Prejudicial Error
    2. Administration and Distrust
    3. The Consequences of Remedial Purity
    1. Legislative and Nonlegislative Rules
    2. Logical Outgrowth
    3. Good Cause
    1. Failures to Resolve Statutory Ambiguities
    2. Harmless Error
    3. Remand Without Vacatur



In United States v. Texas, the Supreme Court upheld by an equally divided vote a nationwide injunction prohibiting the Obama Administra­tion from implementing its “deferred action program” with respect to certain groups of unauthorized aliens. 1 136 S. Ct. 2271, 2272 (2016) (mem.) (per curiam). According to the Fifth Circuit opinion under review, the program, which was announced in an enforce­ment memo directed at officers of the Department of Homeland Security (DHS), 2 Texas v. United States, 809 F.3d 134, 146–47 (5th Cir. 2015), aff’d by an equally divided court, United States v. Texas, 136 S. Ct. 2271 (per curiam). amounted to a new legislative rule. 3 Id. at 184. Because the Administrative Procedure Act (APA) requires all such rules to pass through the notice-and-comment process prior to their adoption, the memo, which passed through no such process, was null and void. 4 Id. at 176–78.

Much ink has been spilled over whether the deferred action program is in fact a legislative rule. But there’s something odd about the Fifth Circuit’s tacit assumption—widely shared and rarely questioned—that invalidation of the program was the appropriate remedy if DHS improp­erly skipped notice and comment. 5 See id. at 178 (emphasizing that the proper remedy for failing to adhere to notice and comment is to “set aside” and “enjoin” the agency action). Of all the possible defects in the deferred action program, lack of public input was not one of them. The Administration had been considering this deferred action program for months. Not only its broad strokes but even its details had been widely shared. 6 See, e.g., Christi Parsons et al., White House Pursuing Plan to Expand Immigrant Rights, L.A. Times (July 25, 2014, 7:30 pm), []. Public debate over the program was intense, and those with concerns about the rule were not shy about voicing them. 7 See, e.g., Texas Governor-Elect Greg Abbott Details Illegal Immigration Crisis, Fox News (Nov. 10, 2014), []. And DHS responded to those concerns at length when the program was formally announced. 8 The Dep’t of Homeland Sec.’s Auth. to Prioritize Removal of Certain Aliens Unlawfully Present in the U.S. & to Defer Removal of Others, 38 Op. O.L.C. 1 (2014) [hereinafter OLC Memo].

Keeping all that in mind, one can reasonably wonder what would be gained from forcing DHS to move its deferred action program through the conventional rulemaking process. 9 See infra section II.A. The better approach might have been to find any error harmless, especially given the APA’s instruction that “due account shall be taken of the rule of prejudicial error” in conducting judicial review. 10 Administrative Procedure Act, 5 U.S.C. § 706 (2012). That possibility, however, was so far beyond the pale that the Administration didn’t even make the argument.

The arguable lack of fit between error and remedy in United States v. Texas exemplifies administrative law’s systematic inattention to remedial questions. With rare exceptions, agency actions that contravene the APA are invalidated and returned to the agency. 11 See Ronald M. Levin, Judicial Remedies, in A Guide to Judicial and Political Review of Federal Agencies 251, 251 (2d ed. 2015) (“As a general rule, when an agency action fails the standards of judicial review, the court is expected to set the action aside and remand it to the agency for further consideration.”). Across a range of cases, the remedy appears disproportionate to the underlying infraction. 12 See infra Parts II–III. The courts’ strict approach could perhaps be justified on prophylactic grounds: Better a hard rule that encourages procedural fastidiousness than a reme­dial standard that might tempt agencies to cut corners. 13 See infra Part IV. It’s not obvious, however, that the benefits of a rule-bound approach outweigh its costs; indeed, there’s reason to fear that they don’t. But we’re not even debating the question.

The lack of a robust debate is especially peculiar given the attention lavished on remedy in related contexts. Exceptions have been made to the exclusionary rule, for example, when exclusion would lead to the loss of probative evidence and would not much deter law enforcement officers from violating the Fourth Amendment. 14 See Utah v. Strieff, 136 S. Ct. 2056, 2059 (2016) (“[T]he Court has . . . held that, even when there is a Fourth Amendment violation, this exclusionary rule does not apply when the costs of exclusion outweigh its deterrent benefits.”). Yet the heated debate over the proper scope of the exclusionary rule finds no counterpart in administrative law, even though the structure of the problem is the same: To what extent is a costly judicial remedy (exclusion in one context, invali­dation in the other) necessary to encourage government actors to respect legal rights?

This Article aims to open that conversation. The bulk of the Article will canvass categories of cases in which there is often a mismatch between the underlying violation and the harshness of the conventional remedy. 15 See infra Parts II–III. That mismatch is common, as United States v. Texas suggests, in cases in which agencies skip notice and comment because they believe a rule is exempt. 16 See Brief for the Petitioners at 65–73, United States v. Texas, 136 S. Ct. 2271 (2016) (No. 15-674), 2016 WL 836758 (arguing notice-and-comment rulemaking was unnecessary). Rarely do courts ask whether the agency found other means for soliciting public input; instead, prejudice is presumed. Courts have a similar blind spot when agencies offer inadequate explanations for agency action or erroneously determine that the law forecloses a particular regulatory option. They typically don’t inquire into whether the agency’s mistake made a difference to its decision or otherwise caused injury to the challenger.

This Article, however, only scratches the surface of overlooked remedial questions. When agencies adopt rules that exceed their legal authority, should courts invalidate entire rules or just sever the offending provisions? 17 For recent work on this front, see generally Charles W. Tyler & E. Donald Elliott, Administrative Severability Clauses, 124 Yale L.J. 2286 (2015). Given the traditional rule that injunctions should be narrowly tailored, should the federal courts limit injunctive relief to the particular plaintiffs? 18 Cf. Summers v. Earth Island Inst., 555 U.S. 488, 491–92 (2009) (declining to address the government’s argument to this effect). Should the lower courts eschew nationwide injunc­tive relief? 19 Cf. Nat’l Mining Ass’n v. U.S. Army Corps of Eng’rs, 145 F.3d 1399, 1408–10 (D.C. Cir. 1998) (considering the appropriateness of nationwide injunctions in APA cases). See generally Samuel L. Bray, Multiple Chancellors: Reforming the National Injunction (UCLA Sch. of Law, Pub. Law Research Paper No. 16-54, 2017), [] (arguing against the practice of imposing nationwide injunctions). When the courts rule against the government, should courts more frequently enter stays to give the political branches an opportunity to respond? 20 Cf. N. Pipeline Constr. Co. v. Marathon Pipe Line Co., 458 U.S. 50, 88–89 (1982) (entering a stay of its judgment to allow Congress to respond).

Questions like these pervade administrative law, but they don’t get the attention they deserve. To be sure, they are not entirely disregarded: The D.C. Circuit, for example, has developed a practice of remanding an agency action without vacating it when the defects in the agency’s rule are modest and invalidation would be disruptive. 21 See, e.g., Allied-Signal, Inc. v. U.S. Nuclear Regulatory Comm’n, 988 F.2d 146, 150–51 (D.C. Cir. 1993). But the remand-without-vacatur innovation is controversial. Several judges believe it to be unlawful, 22 See Milk Train, Inc. v. Veneman, 310 F.3d 747, 758 (D.C. Cir. 2002) (Sentelle, J., dissenting) (“[W]hen we hold that the conclusion heretofore improperly reached should remain in effect, we are substituting our decision of an appropriate resolution for that of the agency to whom the proposition was legislatively entrusted.”); Checkosky v. SEC, 23 F.3d 452, 491 (D.C. Cir. 1994) (Randolph, J., dissenting) (“Once a reviewing court determines that the agency has not adequately explained its decision, the Administrative Procedure Act requires the court—in the absence of any contrary statute—to vacate the agency’s action.”). and agencies face heavy criticism for declining to respond to remand orders. 23 See Kristina Daugirdas, Note, Evaluating Remand Without Vacatur: A New Judicial Remedy for Defective Agency Rulemakings, 80 N.Y.U. L. Rev. 278, 301–05 (2005) (provid­ing examples of cases in which agencies did not prioritize responding to remand-without-vacatur decisions). In this Article, I hope to allay both concerns. 24 See infra section III.C. The instruction to take account of “prejudicial error” provides an unimpeach­able statutory foundation for the practice. 25 See Administrative Procedure Act, 5 U.S.C. § 706 (2012). And in many cases, it would be senseless to compel agencies to devote scarce resources to fixing insub­stantial problems. Upholding a rule without any expectation of an agency response may reflect a basic tenet of playground justice: no harm, no foul.

The Article closes by critically assessing the enduring appeal of a rule-like approach to administrative remedies. 26 See infra Part IV. In part, the approach responds to lingering discomfort with delegations of discretionary authority to federal agencies and a concomitant assumption that only the vigilance of the courts prevents tho­se agencies from running amok. A strict remedial approach is also attractive because of the serious epistemic challenges associated with the exercise of remedial discretion. Courts little under­stand how their decisions influence agency behavior. How can they possibly tell whether a given exercise of remedial restraint might encour­age conduct that the courts believe it their duty to discourage?

Neither objection is compelling, however. Holding more agency errors harmless may not much affect agency incentives; to the extent it does, any uptick in agency misbehavior may not be sufficiently worrisome to warrant the reflexive invalidation of agency action. In any event, uncertainty about how agencies might respond is not sufficient reason, by itself, to justify a rule-like approach to remedy. The adoption of any remedial approach—including a default rule of invalidation—necessarily rests on uncertain and untested beliefs about how agencies respond to judicial review. Yet courts make context-sensitive remedial judgments in a host of other areas in which they have only a rough sense of their effects. 27 See infra Part IV. Why not in administrative law? Although a standard will increase decision costs and generate inconsistency, the exercise of remedial restraint in appropriate cases may well prove superior to a clumsy approach that treats every transgression as worthy of equal sanction.

I. Why Remedial Purity?

Modern administrative law reflects the dominance of those whom Judge Henry Friendly once called remedial “purists,” who

insist that any guessing by a court about what the agency might do when apprised of . . . an error is an unlawful intrusion into the sanctity of the administrative process, and once such an error is detected, the case must go back so that the agency, as the sole repository of authority, can decide it right. 28 Henry J. Friendly, Chenery Revisited: Reflections on Reversal and Remand of Administrative Orders, 1969 Duke L.J. 199, 223.

This Part offers some thoughts about why the purists have won the day, before turning, in Part II, to those categories of cases in which the lack of attention to remedy is most troubling.

A. The Rule of Prejudicial Error

The APA instructs federal courts to “hold unlawful and set aside” arbitrary or unlawful agency action. 29 5 U.S.C. § 706(2). When the APA was enacted in 1946, that instruction reflected a consensus that judicial review of agency action should be modeled on appellate review of trial court judgments. 30 See Thomas W. Merrill, Article III, Agency Adjudication, and the Origins of the Appellate Review Model of Administrative Law, 111 Colum. L. Rev. 939, 942–43 (2011) (“The appellate review model was fully entrenched before the onset of the New Deal and was later incorporated into the Administrative Procedure Act in 1946.”). The APA’s embrace of the appellate model reflected the contempora­neous fact that agencies overwhelmingly operated through adjudication, not rulemaking. Just as a district court judgment infected with error should be invalidated and returned for reconsideration, so too with agency action. At the same time, some errors were harmless and could properly be treated as such: “[D]ue account shall be taken of the rule of prejudicial error.” 31 5 U.S.C. § 706.

The open-textured commands of the APA papered over consider­able disagreement in Congress over the stringency of judicial review. 32 See Nicholas Bagley, The Puzzling Presumption of Reviewability, 127 Harv. L. Rev. 1285, 1307–09 (2014). Measured against today’s standards, however, the APA antici­pated that reviewing courts would take a light touch. Arbitrariness review was thought to require little more from agencies than the courts would insist upon from an act of Congress. 33 See Pac. States Box & Basket Co. v. White, 296 U.S. 176, 182 (1935) (applying arbitrariness review to a state regulation and noting that the standard presumes the exis­tence of facts that would sustain the action in question). In contemplating only limited judicial superintendence of the administrative state, the APA reflected the era’s optimism that expert agencies could largely be trusted to advance the public interest. 34 See generally James Landis, The Administrative Process 123–34 (1938).

By the same token, the instruction to take account of prejudicial error was meant to prevent judges from needlessly overturning agency deci­sions. According to the final lines of the Attorney General’s influential 1947 manual on the APA, the instruction “sums up in succinct fashion the ‘harmless error’ rule applied by the courts in the review of lower court decisions as well as of administrative bodies.” 35 U.S. Dep’t of Justice, Attorney General’s Manual on the Administrative Procedure Act 110 (1947) [hereinafter AG’s Manual]. In his earlier comments on the draft bill that became the APA, the Attorney General similarly emphasized that the “prejudicial error” language was meant to track the common law: “[N]ot every failure to observe the requirements of this statute or of the law is ipso facto fatal to the validity of an order. The statute adopts the rule now well established as a matter of common law in all jurisdictions that error is not fatal unless prejudicial.” S. Rep. No. 79-752, at 44–45 (1945) (reprinting the Attorney General’s comments). The committee reports in both the House and the Senate are more grudging in tone: “The requirement that account shall be taken ‘of the rule of prejudicial error’ means that a procedural omission which has been cured prior to the finality of the action involved by affording the party the procedure to which he was originally entitled is not a reversible error.” H.R. Rep. No. 79-1980, at 46 (1946); S. Rep. No. 79-752, at 28. What did that rule entail? “[E]rrors which have no substantial bearing on the ultimate rights of the parties will be disregarded.” 36 AG’s Manual, supra note 35, at 110. The Attorney General’s articulation of the rule echoed, probably consciously, Rule 61 of the Federal Rules of Civil Procedure, which had been adopted a decade earlier and applied in all federal civil litigation. 37 See Fed. R. Civ. P. 61 (“At every stage of the proceeding, the court must disregard all errors and defects that do not affect any party’s substantial rights.”). For support, the Attorney General cited Market Street Railway Co. v. Railroad Commission, in which the Supreme Court, in 1945, had upheld an agency ratemaking decision that was improperly premised on evidence outside the record. 38 324 U.S. 548, 561–62 (1945); AG’s Manual, supra note 35, at 110. “It does not appear,” the Court wrote, “that the Company was in any way prejudiced thereby, and it makes no showing that, if a rehearing were held to introduce its own reports, it would gain much by cross-examination, rebuttal, or impeachment.” 39 Market Street Ry. Co., 324 U.S. at 562.

As a matter of what the APA says, then, harmless error review should be as central to administrative law as it is to conventional litigation. In the decades since the APA’s adoption, however, the rule of prejudicial error has been all but forgotten. The Supreme Court never invoked the rule to decide an agency case until 2007, 40 See Nat’l Ass’n of Home Builders v. Defs. of Wildlife, 551 U.S. 644, 659 (2007). and it wasn’t until 2009, in Shinseki v. Sanders, that the Court offered any discussion of the provision at all. 41 See 556 U.S. 396, 406–11 (2009). There, the Court simply quoted the final lines of the Attorney General’s Manual for the proposition that the APA’s harmless error rule was no different from the rule applied in everyday civil litigation. 42 Id. at 406 (quoting AG’s Manual, supra note 35, at 110). In the scholarly literature, the significance of the prejudicial error provision is either downplayed or ignored. 43 The major exception is Professor Ronald Levin. See Ronald M. Levin, Scope-of-Review Doctrine Restated: An Administrative Law Section Report, 38 Admin. L. Rev. 239, 284 (1986) (describing the APA’s prejudicial error principle); Ronald M. Levin, “Vacation” at Sea: Judicial Remedies and Equitable Discretion in Administrative Law, 53 Duke L.J. 291, 312 (2003) [hereinafter Levin, Vacation] (“This clause, a harmless error principle, necessarily implies that the ‘shall . . . set aside’ language found earlier in the provision must mean ‘shall generally,’ not ‘shall always.’” (quoting Administrative Procedure Act, 5 U.S.C. § 706 (2000))). The lower courts have not been as deaf to the rule of prejudicial error as the Supreme Court and commentators—harmlessness has worked its way into the corners of the law. 44 See infra section III.B–.C. But it’s only modest hyperbole to say that the rule of prejudicial error has been lost to administrative law. Why?

B. Administration and Distrust

By the 1960s, postwar optimism about the administrative state had dimmed. 45 See generally Richard B. Stewart, The Reformation of American Administrative Law, 88 Harv. L. Rev. 1669 (1975) (noting the influence of the claim that agencies exercise their discretion in favor of organized and regulated interests). Uneasiness about the public mindedness of agencies brought to the fore concerns about agency capture—the notion, bolstered by public choice theory, that concentrated interests would exploit their organizational advantages over a diffuse public to twist agency decisions to their private ends. 46 See generally George J. Stigler, The Theory of Economic Regulation, 2 Bell J. Econ. & Mgmt. Sci. 3, 3 (1971) (arguing regulations will inevitably favor regulated interests). Courts and commentators also grew disenchanted with the idea that expertise was sufficient to enable agencies to identify some determinate public good. 47 See Stewart, supra note 45, at 1683 (observing agency decisionmaking requires interest balancing, which requires more than technical expertise). Even as agencies made the sorts of contestable value judgments that had traditionally been reserved to Congress, they lacked Congress’s democratic pedigree or its institutional capacity to channel public values.

To reconcile agencies’ democratic deficit with their immense power, the courts began to insist on agencies’ adherence to procedures that would assure some measure of public input and deliberation. Bare rationality would no longer suffice; agencies would have to defend the reasona­bleness of their decisions based on a record compiled at the time they were made. 48 See Citizens to Pres. Overton Park, Inc. v. Volpe, 401 U.S. 402, 416 (1971). And the courts transformed notice and com­ment from the almost-ministerial act contemplated by the spare text of the APA into an elaborate justificatory process that would, it was hoped, foster mean­ingful public input, force agencies to grapple with objections, and enable judicial oversight. 49 Compare Administrative Procedure Act, 5 U.S.C. § 553 (2012) (setting out the APA’s rulemaking requirements), with United States v. N.S. Food Prods. Corp., 568 F.2d 240, 251–53 (2d Cir. 1977) (requiring the agency to disclose all the scientific data upon which it relied and to respond to all “vital questions”).

Around the same time, the rise of rulemaking put pressure on the traditional notion that an agency action could be challenged only in an enforcement proceeding. 50 On the rise of rulemaking, see Thomas W. Merrill & Kathryn Tongue Watts, Agency Rules with the Force of Law: The Original Convention, 116 Harv. L. Rev. 467, 545–70 (2002). Regulated entities pointed out that, as a practical matter, they would have no choice but to come into compliance with agency rules long before the rules were ever enforced. 51 See Abbott Labs. v. Gardner, 387 U.S. 136, 152–53 (1967). Fearing that agencies might exploit that fact to evade legal restraints or to shirk their procedural obligations, the Supreme Court in Abbott Laboratories v. Gardner blessed preenforcement review of agency rulemaking. 52 Id. at 148.

Together with the advent of hard-look review, preenforcement review made judicial oversight into something resembling the final step in the adoption of any major agency policy. No longer episodic and unusual, judicial review became the norm—and it has stayed the norm. Indeed, it is fair to say that the necessity of robust judicial review has hardened into an infrequently challenged convention of administrative law.

Against this backdrop, it’s natural to see judicial review primarily as a mechanism to stop agencies from evading the procedural safeguards and legal constraints that make agency action functionally and constitu­tionally tolerable. 53 For recent articulations of deep anxiety about the administrative state, see gener­ally Dep’t of Transp. v. Ass’n of Am. R.Rs., 135 S. Ct. 1225, 1240–55 (2015) (Thomas, J., concurring in the judgment); Philip Hamburger, Is Administrative Law Unlawful? (2014). For the classic exposition, see generally Theodore J. Lowi, The End of Liberalism (2d ed. 1979). A tacit presumption of agency distrust has all but displaced the presumption of regularity. Reinforcing the point are the canonical arbitrariness cases—Chenery, 54 SEC v. Chenery Corp., 318 U.S. 80 (1943). Overton Park, 55 Citizens to Pres. Overton Park, Inc. v. Volpe, 401 U.S. 402 (1971). State Farm, 56 Motor Vehicle Mfrs. Ass’n v. State Farm Mut. Ins., 463 U.S. 29 (1983). Massachusetts v. EPA 57 549 U.S. 497 (2007). —which are widely admired for their flinty refusal to accept agency half-measures. Going easy on errant agencies, the cases seem to teach, would license sloppiness and sidestepping. The systemic consequences of remedial restraint would be especially disquieting: Instead of hewing to the procedural rules and statutory constraints that cleanse agency action of the taint of executive diktat, agencies would flout those rules in the service of their narrow agendas.

Excusing agency errors as harmless would thus run counter to the broader project of using judicial review to discipline executive discretion. Indeed, harmless error could send courts down a slippery slope. For any given agency action, the costs of vacating and remanding are relatively concrete. The systemic benefits of vacatur, however, are diffuse and speculative. If licensed to hold errors harmless, courts might too readily sustain agency actions that ought to be invalidated. Best hold the line with a rule than risk the inexorable expansion of harmless error. 58 Cf. Plaut v. Spendthrift Farm, Inc., 514 U.S. 211, 239 (1995) (“[T]he doctrine of separation of powers is a structural safeguard rather than a remedy to be applied only when specific harm, or risk of specific harm, can be identified. In its major features . . . it is a prophylactic device . . . .”).

And so, as the common law of agency review has developed, the rule of prejudicial error has gone missing. 59 Cf. United States v. Johnson, 632 F.3d 912, 931 (5th Cir. 2011) (noting that concerns an agency might evade procedural rules “support the limited role of the harmless error doctrine in administrative law”). It stays missing, in part, because of the Supreme Court’s discretionary docket, which enables it to opine on agency error while leaving remedial questions to the lower courts. The Court, for example, has never passed on the validity of the D.C. Circuit’s practice of remand without vacatur, despite opportunities to do so. 60 See Stephanie J. Tatham, Admin. Conference of the U.S., The Unusual Remedy of Remand Without Vacatur 8–9 (2014), [] (identifying two such opportunities).

Remedial questions also receive little attention in briefing. 61 See Daugirdas, supra note 23, at 310. As Professor Kent Barnett has recently argued, a similar inattention to remedy plagues much separation-of-powers litigation. See Kent Barnett, To the Victor Goes the Toil—Remedies for Regulated Parties in Separation-of-Powers Litigation, 92 N.C. L. Rev. 481, 486–87 (2014). In part, that’s because remedial arguments are unlikely to find a receptive audience. But it’s also risky for a federal agency to argue for remedial re­straint. Emphasizing the triviality of an error is another way of admitting how easy it would have been to address the error in the first place, suggesting the arbitrariness of the agency’s failure to do it right the first time around. 62 See Daugirdas, supra note 23, at 310. And arguing that the agency would have reached the same conclusion even if it had adhered religiously to the APA can alienate judges who think that the agency must therefore not take administrative procedures seriously. Often, it’s better to press a confident argument on the merits and hope for the best.

C. The Consequences of Remedial Purity

The common law development of administrative law could have gone in a different direction, one that was more tolerant of agency mistakes and less insistent that judicial review holds the key to agency accountability and legitimacy. Nothing in the APA precludes that ap­proach; if anything, it is more faithful to the APA’s original meaning than the severe interpretation that governs today.

As it stands, however, the courts are committed to remedial purity. Whatever the merits of that approach, its costs are large. When a court vacates an agency action, the agency must decide whether to correct whatever deficiency the court has identified. Rectifying the mistake may be no mean feat, especially if doing so requires the agency to trudge through the procedural thicket surrounding notice and comment. 63 See ICF Consulting, The Reg Map: Informal Rulemaking (2003), [] (representing visually the extensive process required for informal rulemaking). In the meantime, the agency action will be put on hold—delayed, often for years, as the agency decides how to respond. In the end, the agency might choose to abandon the action altogether: Its priorities may have changed, its staff may have been reassigned, or the external groups supporting action may have dispersed. 64 See Jerry L. Mashaw & David L. Harfst, Regulation and Legal Culture: The Case of Motor Vehicle Safety, 4 Yale J. on Reg. 257, 295 (1987) (“The idea that an agency can or will quickly turn to remedying the factual or analytic defects in its remanded rule is surely naive, however minor those problems might appear in the abstract.”). Judicial review can thus derail or delay significant government programs, sometimes at substantial cost to the public welfare. 65 For classic pieces documenting how judicial review can harm the public, see generally Jerry L. Mashaw & David L. Harfst, The Struggle for Auto Safety (1990); R. Shep Melnick, Regulation and the Courts: The Case of the Clean Air Act (1983); Richard J. Pierce, Jr., The Unintended Effects of Judicial Review of Agency Rules: How Federal Courts Have Contributed to the Electricity Crisis of the 1990s, 43 Admin. L. Rev. 7 (1991).

The harshness of the vacate-and-remand remedy may also affect the standards that courts employ to gauge the legality of agency action. As Professor Daryl Levinson has argued, a disproportionate remedy can lead courts to reshape the substance of constitutional rights. 66 See Daryl J. Levinson, Rights Essentialism and Remedial Equilibration, 99 Colum. L. Rev. 857, 884–85 (1999). So too with administrative law, in which the prospect of invalidation may push courts to narrow the scope of what counts as arbitrary. It’s an open secret in agency cases, for example, that the right question is not whether the agency has acted arbitrarily or capriciously. Errors and arbitrariness are as inevitable in agency action as they are in any complex human activity. The right question is how arbitrarily the agency has acted. A court can—indeed, must—set the stringency of review to suit whatever rough sense of policy guides it in reviewing agency action. Courts may therefore water down the substantive standard to avoid a remedy that seems dispropor­tionate to the perceived error.

Some evidence suggests that this occurs. Professors Jacob Gersen and Adrian Vermeule have recently found, for example, that arbitrariness review is much less rigorous in practice than cases like Overton Park and State Farm would suggest—not only at the Supreme Court, but in the lower courts as well. 67 Jacob E. Gersen & Adrian Vermeule, Thin Rationality Review, 114 Mich. L. Rev. 1355, 1361–67 (2016). To match their descriptive account, Gersen and Vermeule offer a theory to support what they call “thin rationality review.” 68 Id. at 1357. As they see it, agencies are obliged only to decide on the basis of reasons, not necessarily to examine the full range of feasible options and offer a convincing explanation for why the chosen policy is superior to alternatives. 69 Id. Given resource constraints and uncertainty, agencies might reasonably leave some options unexamined. 70 Id. And, given the cost and difficulty of conveying the tacit knowledge that underwrites their expertise, agencies may be unwilling or unable to fully explain the basis of their decisions to generalist judges. 71 Id.

As theory, Gersen and Vermeule’s account is compelling. It is also provocative: Most judges would bristle at the suggestion that agencies need not ventilate all plausible options or fully explain themselves. If that’s right, however, what accounts for the pattern that Gersen and Vermeule observe of “soft look” review? 72 See id. at 1361–67. It could be that the courts have relaxed the stringency of hard-look review out of some unexpressed and inchoate appreciation of what agencies can be expected to communicate. But latent remedial concerns may also contribute to the disconnect that Gersen and Vermeule observe between the law in the casebooks and the law on the ground. 73 See id. Judges weaned on Overton Park and State Farm need some motivation to overcome their doctrinal commitment to reason-giving and thoroughness. The felt obligation to avoid a harsh remedy might supply that motivation.

Suppressing the remedial debate thus might not only distort the law on the books. It might also hamper forthright discussion of agency error. As it stands, courts face an unappealing choice: Either the agency has acted arbitrarily (and its action must fall) or it hasn’t (and its action will stand). That binary fails to capture that arbitrariness is a matter of degree: Even the best-considered agency actions contain gaps, inconsistencies, and stupidities. Under current doctrine, it’s easier to deny the existence of those problems than to explain why they do not require invalidation.

II. Notice-and-Comment Failures

Under § 553 of the APA, an agency that wishes to adopt a rule must publish a “[g]eneral notice of proposed rule making” in the Federal Register, allow the public to comment, and “incorporate in the rule[] adopted a concise general statement of their basis and purpose.” 74 5 U.S.C. § 553 (2012). The courts have built on that spare language to impose wide-ranging proce­dural obligations on agencies in connection with their adoption of new rules. 75 See Jack M. Beermann & Gary Lawson, Reprocessing Vermont Yankee, 75 Geo. Wash. L. Rev. 856, 857 (2007). That practice can be criticized on any number of grounds, includ­ing that it can’t really be derived from the APA. 76 See id. at 866–68. Nonetheless, courts tend to deal harshly with agencies that skip notice and comment. 77 See, e.g., Allina Health Servs. v. Sebelius, 746 F.3d 1102, 1109 (D.C. Cir. 2014) (“We have not been hospitable to government claims of harmless error in cases in which the government violated § 553 of the APA by failing to provide notice.”). In the circuits, the typical standard is that “the failure to provide notice and comment is harmless only where the agency’s mistake ‘clearly had no bearing on the procedure used or the substance of decision reached.’” 78 Riverbend Farms, Inc. v. Madigan, 958 F.2d 1479, 1487 (9th Cir. 1992) (quoting Sagebrush Rebellion, Inc. v. Hodel, 790 F.2d 760, 764–65 (9th Cir. 1986)); see also Animal Legal Def. Fund v. U.S. Dep’t of Agric., 789 F.3d 1206, 1225 n.13 (11th Cir. 2015); United States v. Utesch, 596 F.3d 302, 311–13 (6th Cir. 2010); U.S. Steel Corp. v. EPA, 595 F.2d 207, 215 (5th Cir. 1979). In application, that means that notice-and-comment errors are almost never held harmless.

Closer attention to remedy would require the courts to ask whether the deficiencies of the agency process have materially frustrated the values that notice and comment is meant to serve. Broadly speaking, notice-and-comment rulemaking serves both informational and participatory functions: It assures that agencies incorporate all relevant information into their decisionmaking and guarantees that members of the public have a voice in the decisions that affect their lives. 79 See Cass R. Sunstein, Democratizing Regulation, Digitally, Democracy (Fall 2014), [] (identifying the two primary functions of notice-and-comment rulemaking). Better information and more robust participation are thought to improve the quality of agency rules, enhance rules’ legitimacy, and make agencies more account­able to the public will. 80 See id.

But these functions can be served by means other than formal notice and comment. Start with the informational function. Did the agency tell the party complaining of the lack of notice and comment what it intended to do? Did it open lines of communication with that party about its proposed action? What comments would the complaining party have submitted through the formal process that it didn’t or couldn’t have offered in another setting? Did the agency already address the concerns that the party would have raised? Were they the kinds of comments that might have made a difference?

These questions are rarely asked and even more rarely answered, even though agencies solicit a lot of input outside the formal notice-and-comment process. Especially for significant actions, agencies use listening sessions, workshops, conference calls, informal discussions, interest-group meetings, and strategic leaks to learn about whether their pro­posals are workable, legal, and politically acceptable. 81 See Steven P. Croley, Regulation and Public Interests: The Possibility of Good Regulatory Government 258 (2008) (highlighting how agencies engaged in three high-profile rulemakings in the 1990s “provided more notice, data, and opportunities for participation . . . [than] the APA (or any other legal authority) demanded”); Nicholas Bagley & Helen Levy, Essential Health Benefits and the Affordable Care Act: Law and Process, 39 J. Health Pol. Pol’y & L. 441, 460 (2014) (“At least for salient policy questions of substantial importance . . . agencies have a number of incentives having little or nothing to do with formal legal requirements to secure public input and ensure political oversight.”); Elizabeth Magill, Agency Self-Regulation, 77 Geo. Wash. L. Rev. 859, 860 (2009) (describing how agencies “voluntarily constrain their discretion” and “limit their procedural freedom by committing to afford additional procedures, such as hearings, notices, and appeals, that are not required by any source of authority”); Adrian Vermeule, Deference and Due Process, 129 Harv. L. Rev. 1890, 1924 (2016) (“It is not the case that agencies want to offer as little procedure as possible.”). Indeed, the very regulated entities that file suit to challenge the absence of notice and comment are often intimately involved in crafting agency policy through these alterna­tive channels.

gencies solicit input for a simple reason: They generally want feed­back. 82 Cf. Vermeule, supra note 81, at 1925 (“[O]ne of the major reasons agencies go beyond the legal minimum is that, precisely when and to the extent that agencies are mission-oriented, they will have an interest in accuracy, and will sometimes provide extra procedure in order to ensure accuracy . . . .”). That’s one reason why agencies adhere to the notice-and-comment process even when they don’t have to. 83 See, e.g., 36 Fed. Reg. 2532, 2532 (Jan. 28, 1971) (subjecting all Department of Health, Education, and Welfare rules to notice and comment, even though Medicare and Medicaid rules are exempt under the APA because they relate to “benefits”). And it helps explain, too, why agencies that skip notice and comment prior to issuing a final rule will usually ask for public comments upon publication and, when they do, respond publicly. 84 See U.S. Gov’t Accountability Office, GAO-13-21, Federal Rulemaking: Agencies Could Take Additional Steps to Respond to Public Comments 24–25 (2012) (providing data showing that even those agencies that are not required to use notice and comment will frequently request and respond to comments).

Stringent judicial review may therefore not be as essential to assuring public input as courts and commentators assume. That’s especially so because the executive branch has an institutional commit­ment to notice and comment. Under executive orders, agencies are obliged to solicit public feedback prior to issuing a notice of proposed rulemaking. 85 See Exec. Order No. 13,563, 76 Fed. Reg. 3821, 3822 (Jan. 21, 2011); Exec. Order No. 12,866, § 6(a)(1), 3 C.F.R. 638, 644 (1994), reprinted as amended in 5 U.S.C. § 601 app. at 802, 804 (2012). The Office of Information and Regulatory Affairs (OIRA) takes a similar view:

[The Office] . . . promot[es] public comment on all significant regulatory actions, including guidance documents and interpretive rules—not because such comment is required as a matter of law, and not because it is necessary or desirable in every instance, but because when the stakes are high and the issues novel, obtaining public comment is good practice as a way of avoiding mistakes. 86 Cass R. Sunstein, Commentary, The Office of Information and Regulatory Affairs: Myths and Realities, 126 Harv. L. Rev. 1838, 1854 (2013).

To an agency that has its own reasons to secure public input and that must answer to OIRA anyhow, the prospect that a court might someday excuse any breach of its notice-and-comment obligation might not affect its behavior. The marginal incentive effects of remedial purity may well approach zero.

Consider next the participatory function of notice-and-comment rulemaking. At a high level of abstraction, arguments about the need for assertive judicial review to address agencies’ “democracy deficit” have real force. But those arguments tend to come apart in concrete cases. Many, if not most, challenges to agency rules are brought by regulated firms and advocacy groups that have been in close contact with administrative agencies throughout a lengthy process of rule develop­ment. 87 Cf. Cary Coglianese, Litigating Within Relationships: Disputes and Disturbance in the Regulatory Process, 30 Law & Soc’y Rev. 735, 743 (1996) (“Both in rulemaking and in litigation, industry groups are the most common players.”). Participatory concerns recede when the agency has found other means of assuring participation.

Matters are different when a member of the public is blindsided by the adoption of a rule that should have, but did not, go through notice and comment. When such a person brings suit, there’s an argument for vacating the rule on the ground that the agency deprived her of her right to participate—a right with intrinsic value apart from the possibility that it might have affected the agency’s substantive decision. Much as depri­ving someone of her day in court can work an injury, it’s plausible to think that an individual suffers harm if she’s deprived of her chance to contribute to an agency rule.

Still, it pays to be skeptical of claims of prejudice arising purely from the deprivation of a participation right. As Matthew Lawrence has emphasized in a related context, “the inherent value of process—the value that comes from giving a claimant her ‘day in court,’ win or lose—can vary from claimant to claimant and . . . we can take advantage of that variation in distributing scarce procedural protections.” 88 Matthew J.B. Lawrence, Procedural Triage, 84 Fordham L. Rev. 79, 82 (2015). A small-business owner who is unwittingly ensnared by a new rule that never went through notice and comment may understandably be aggrieved by the loss of a chance to participate, even if the agency has heard and considered the objections she would have raised. But a public corporation’s loss of the right to participate, absent a showing of some reason to think that partici­pation might have changed the outcome, will typically not inflict the kind of dignitary harm that would support a finding of prejudice with respect to a real person. 89 A lawsuit brought by a closely held corporation could present a borderline case. See Burwell v. Hobby Lobby Stores, Inc., 134 S. Ct. 2751, 2759 (2014).

In any event, there’s no reason to credit a challenger’s participatory objections if the challenger actually participated in the agency’s deliber­ations. Yes, rebuking an agency for erroneously skipping notice and comment might deter other agencies from skipping notice and comment in the future, whether or not a particular challenger has suffered a participatory injury. To that extent, the challenger may serve as a kind of private attorney general to vindicate the public’s par­ticipatory rights. But there are at least two difficulties with this line of argument. First, it assumes that the marginal incentive effects are large enough, and the public’s right to formal notice and comment precious enough, that the benefits of remedial purity inevitably outweigh the costs to effective governance that such an approach entails. On that absolutist view, no notice-and-comment error, however trivial, could ever be harmless. Second, the public might not thank the courts for invalidating proce­durally defective rules at the behest of regulated firms that purport to represent the public’s interests. Instead, the public might believe that such rules advance their interests and ought to stand, even if agencies may be slightly less assiduous about soliciting public input in the future. What’s good for General Motors, in other words, might not be good for the public.

Some judicial humility may thus be in order. The notice-and-comment process is valuable, but it is not the only value—and judicial enforcement may make less of a difference than courts assume. Several discrete cate­gories of cases appear to call for the exercise of more remedial discretion than courts typically display.

A. Legislative and Nonlegislative Rules

The APA requires only “legislative rules” to pass through notice and comment. 90 See Perez v. Mortg. Bankers Ass’n, 135 S. Ct. 1199, 1206–07 (2015). In contrast, agencies can issue nonlegislative rules—specific­ally, policy statements and interpretive rules—without adhering to any particular procedural formalities. 91 See Administrative Procedure Act, 5 U.S.C. § 553(b)(3)(A) (2012) (exempting “general statements of policy” and “interpretative rules” from notice and comment). To prevent evasion of the notice-and-comment obligation, courts will invalidate policy statements and inter­pretive rules that are really disguised legislative rules. Policy statements are considered legislative when (1) they impose immediate, specific obli­gations on the regulated community and (2) the agency lacks genuine discretion to deviate from the policy in appropriate cases. 92 See Am. Bus. Ass’n v. United States, 627 F.2d 525, 529–30 (D.C. Cir. 1980). Interpretive rules are deemed legislative when they depart too far from the statute or regulation that they purport to interpret. 93 See Paralyzed Veterans of Am. v. D.C. Arena L.P., 117 F.3d 579, 588 (D.C. Cir. 1997) (explaining the distinction between interpretive and substantive rules “turns on how tightly the agency’s interpretation is drawn linguistically from the actual language of the statute or rule”), abrogated on other grounds by Perez, 135 S. Ct. 1199.

Not infrequently, however, it can be hard to distinguish nonleg­islative from legislative rules. Since a regulated entity that ignores a policy statement or interpretive rule risks an enforcement proceeding, all nonlegislative rules affect private conduct and constrain agency discretion, at least to some extent. When do those effects become so serious that the policy statement or interpretive rule must pass through notice and comment? Agencies have to guess—and without much helpful guidance from the courts. 94 See Am. Mining Cong. v. Mine Safety & Health Admin., 995 F.2d 1106, 1108 (D.C. Cir. 1993) (“The distinction between those agency pronouncements subject to APA notice-and-comment requirements and those that are exempt has been aptly described as ‘enshrouded in considerable smog.’” (quoting Gen. Motors Corp. v. Ruckelshaus, 742 F.2d 1561, 1565 (D.C. Cir. 1984))). For arguments that the courts should get out of the business of policing the distinction, see Jacob E. Gersen, Legislative Rules Revisited, 74 U. Chi. L. Rev. 1705, 1718–21 (2007); John F. Manning, Nonlegislative Rules, 72 Geo. Wash. L. Rev. 893, 894–97 (2004). Guess wrong and the courts will invalidate the nonleg­islative rule and, when necessary, enjoin the agency from relying on it. 95 See Chamber of Commerce v. OSHA, 636 F.2d 464, 470–71 (D.C. Cir. 1980) (enjoining an agency from relying on a nonlegislative rule).

In many cases, however, that remedy appears disproportionate to an agency’s failure to anticipate how a future court would apply an indeter­minate standard. Return to United States v. Texas, for example. The Fifth Circuit held that Deferred Action for Parents of Americans (DAPA) was a legislative rule that never went through the formal notice-and-comment process. 96 Texas v. United States, 809 F.3d 134, 150 (5th Cir. 2015). But what exactly did DHS fail to do? True, the agency never published the proposed policy in the Federal Register, as the APA requires. 97 See 5 U.S.C. § 553(b) (2012). But DHS provided notice in a much more effective manner: It leaked the proposal to the national media. 98 The proposal was described as follows:
“One option would allow immigrants who are parents of U.S. citizens to apply for temporary legal status which would let them work legally in the U.S. Because children born in the country automatically receive U.S. citizenship, that option could affect about 5 million people, researchers estimate.”
A second option would be to allow temporary legal status for the parents of young people already granted deportation deferrals by the Obama administration. That would affect a smaller, but still sizable, number of people.
Parsons et al., supra note 6.
Those hungry for additional details could find them in the extensive program rules govern­ing the Administration’s prior exercise of enforcement discretion—the Deferred Action for Childhood Arrivals (DACA) program—that DHS proposed to expand. 99 See Memorandum from Janet Napolitano, Sec’y, U.S. Dep’t of Homeland Sec., on Exercising Prosecutorial Discretion with Respect to Individuals Who Came to the United States as Children to David V. Aguilar, Acting Comm’r, U.S. Customs & Border Prot., Alejandro Mayorkas, Dir., U.S. Citizenship & Immigration Servs. & John Morton, Dir., U.S. Immigration & Customs Enf’t (June 15, 2012), [].

Was the problem that the public had no chance to voice its objections? That also can’t be right. The wisdom and legality of the Administration’s proposal became a flashpoint in the 2014 midterm elections, so much so that the Administration delayed announcing DAPA until after the elections in a (failed) effort to protect vulnerable Senate Democrats. 100 See Michael D. Shear, Obama Delays Immigration Action, Yielding to Democratic Concerns, N.Y. Times (Sept. 6, 2014), (on file with the Columbia Law Review); see also Senate Election Results, N.Y. Times, (on file with the Columbia Law Review) (last updated Dec. 17, 2014). Public officials, including many from Texas, objected vocif­erously to expanding DACA to cover more unauthorized immi­grants. The governor-elect of Texas, Greg Abbott, for one example among many, said on Fox News that he was “concerned that there will be a new surge in border activity in part because of the potential action the [P]resident may take” and warned President Barack Obama not to over­step his constitutional authority. 101 See, e.g., Texas Governor-Elect Greg Abbott Details Illegal Immigration Crisis, supra note 7. And, on the eve of DAPA’s announce­ment, CNN reported that congressional Republicans “have spent months preparing for the announcement by warning of executive overreach and political well-poisoning.” 102 Kevin Liptak & Dana Bash, The Immigration Waiting Game Will End Soon, CNN (Nov. 13, 2014, 7:24 pm), []. No one was surprised.

Is the problem that DHS didn’t respond point by point to the criticisms it received, as is usual in the announcement of a final rule? There, too, the criticism rings hollow. When DHS finally adopted DAPA, the administration offered lengthy explanations of both the desirability and legality of its program. 103 See OLC Memo, supra note 8; Memorandum from Jeh Charles Johnson, Sec’y, U.S. Dep’t of Homeland Sec., to León Rodriguez, Dir., U.S. Citizenship & Immigration Servs., Thomas S. Winkowski, Acting Dir., U.S. Immigration & Customs Enf’t & R. Gil Kerlikowske, Comm’r, U.S. Customs & Border Prot. (Nov. 20, 2014), []. Those explanations addressed the most important criticisms that had been lodged against the program in the public debate, including in particular the claim that DHS lacked the authority to adopt the program at all. 104 Cf. Jerry L. Mashaw et al., Administrative Law: The American Public Law System 619 (7th ed. 2014) (“Rarely has an agency rule been overturned because the required explanation failed to appear in the Federal Register text.”). Indeed, the Office of Legal Counsel (OLC) publicly released a dense and closely reasoned legal opinion justifying the program. 105 See OLC Memo, supra note 8, at 2–11. The opinion shows that OLC took the legal concerns raised by the program’s critics seriously—so seriously, in fact, that it concluded that DHS couldn’t extend DAPA to the parents of children who were neither citizens nor legal permanent residents. 106 See id. at 31–33.

Against this backdrop, why isn’t it incumbent on Texas to explain what a more formalized notice-and-comment procedure might have ac­complished? After all, the Administration loudly signaled what it planned to do. It acknowledged the criticisms aired in a raucous debate. And it responded at length to the most serious of those criticisms. 107 See id. By any measure, DHS’s actions substantially fulfilled the requirements the APA laid out. What more does Texas want?

It’s possible that Texas has an explanation for why it was harmed when the agency dodged notice and comment. Maybe Texas couldn’t formulate adequate comments because the notice it received was too scant. 108 See Fla. Power & Light Co. v. United States, 846 F.2d 765, 771 (D.C. Cir. 1988) (holding an agency’s notice must “provide sufficient factual detail and rationale for the rule to permit interested parties to comment meaningfully”). Maybe DHS refused to meet with Texas officials or sent its letters back unopened. Maybe the agency never addressed an important concern about DAPA’s effect on Texas’s economy. Indeed, the very fact that DHS declined to go through the notice-and-comment process is arguably evidence that the agency wanted to avoid a gloves-off public debate. 109 See Adam B. Cox & Cristina M. Rodríguez, The President and Immigration Law Redux, 125 Yale L.J. 104, 217–18 (2015) (arguing that “[s]ignificant policymaking of this sort would have benefitted from public scrutiny and involvement” through the formal notice-and-comment process). But why not at least insist that Texas offer up that explanation? It’s otherwise hard to resist the conclusion that Texas has tripped DHS up on a technicality.

For another high-profile example, consider the D.C. Circuit’s deci­sion in General Electric Co. v. EPA. 110 290 F.3d 377 (D.C. Cir. 2002). Pursuant to the Toxic Substances Control Act, EPA had adopted rules governing the cleanup and disposal of certain kinds of toxic waste associated with polychlorinated biphenyls (PCBs). 111 Id. at 379. Two of those rules laid out default compliance methods but also allowed companies to apply to use alternative methods, which EPA would approve if the alternative method did “not pose an unreasonable risk of injury to health or the environment.” 112 See 40 C.F.R. §§ 761.61(c)(2), 761.62(c) (2015). EPA then prepared a document describing what alternative approaches might look like. 113 Gen. Elec., 290 F.3d at 379. In particular, EPA said that it would approve methodologies that employed a particular toxicity factor in measuring the risk of any cleanup and disposal efforts. 114 Id.

General Electric (GE) wanted a less protective toxicity factor. In a lawsuit, GE argued that the guidance document’s incorporation of the specific toxicity factor made it a legislative rule that should have passed through notice and comment. 115 Id. The D.C. Circuit agreed and promptly vacated the rule: “[H]aving held that . . . the Guidance Document is a ‘rule’ . . . , it is clear that GE must prevail on the merits.” 116 Id. at 385.

As a matter of doctrine and practice, the remedy was unsurprising. EPA even confessed that vacatur would be appropriate if the guidance document was found to be a legislative rule. 117 Id. Taking a step back, though, the remedy appears extravagant. GE was no ordinary bystander to EPA’s decision. The company had been intimately involved in PCB-remediation efforts since its epic pollution of the Hudson River came to light in the mid-1970s. 118 See Harold Faber, Issue & Debate; Plan to Purify Upper Hudson Waters, N.Y. Times (Oct. 1, 1983), (on file with the Columbia Law Review). Between 1990 and 2005, GE spent nearly one billion dollars on PCB remediation, not including an additional “$11 million to monitor and participate in the legislative and regulatory process related to the development of PCB and hazardous waste laws and regulations.” 119 Gen. Elec., Report on PCB Expenditures, 1990–2005, [http://perma/cc/LLX9-TZZ2] (last visited Nov. 7, 2016).

In particular, GE had been fighting with EPA for years over the proper toxicity factor to use in judging alternative methods of cleaning up PCBs. EPA issued an advance notice of proposed rulemaking on that precise question in 1991 and a notice of proposed rulemaking in 1994, 120 See 59 Fed. Reg. 62,788 (proposed Dec. 6, 1994) (codified at 40 C.F.R. pt. 761); 56 Fed. Reg. 26,738 (proposed June 10, 1991) (codified at 40 C.F.R. pt. 761). receiving comments on each, including many from GE. 121 See Office of Prevention, Pesticides & Toxic Substances, EPA, Response to Comments Document on the Proposed Rule—Disposal of Polychlorinated Biphenyls 232–33 (1998), []. In 1998, EPA issued a final rule adopting the same toxicity factor that it later incorpo­rated in the guidance document. 122 See 63 Fed. Reg. 35,384 (June 29, 1998) (codified at 40 C.F.R. pts. 750, 761). In a lengthy response to com­ments, the agency acknowledged that a recent study suggested that the toxicity factor might be too stringent when it came to cancer risks. EPA none­theless concluded that a more protective standard was appropriate given the uncertainty associated with noncancer risks. 123 See Office of Prevention, Pesticides & Toxic Substances, supra note 122, at 128–29. Two years later, GE challenged the 1998 rule in the Fifth Circuit. 124 Cent. & S.W. Servs., Inc. v. EPA, 220 F.3d 683, 683 (5th Cir. 2000). Because EPA had just begun conducting a comprehensive evaluation of noncancer risks associ­ated with PCBs, the agency raised no objection to vacating that portion of its rule. 125 Id. at 695. During the pendency of that evaluation, EPA incorporated the toxicity factor into a guidance document. 126 Brief for Respondent United States EPA at 11, Gen. Elec. Co. v. EPA, 290 F.3d 377 (D.C. Cir. 2002) (No. 00-1394), 2001 WL 36039627 [hereinafter EPA Respondent Brief].

EPA’s evaluation of the noncancer risks—an evaluation in which GE actively participated 127 Id. at 3. —was still ongoing in 2002, when the D.C. Circuit decided General Electric v. EPA. 128 290 F.3d 377. Remember, the court invalidated the guidance document because it hadn’t gone through notice and comment. Never mind that the toxicity factor had actually been through seven years of notice and comment. Never mind that GE had been grousing to EPA about the toxicity factor for more than a decade and that the agency had responded, publicly and at length, to its concerns. Never mind that EPA had a convincing explanation for why it thought its guidance document was nonlegislative. Never mind that GE offered no explanation of what more it would have said that might have led EPA to change its mind. Never mind that the guidance document “was prepared with industry input,” including GE’s. 129 EPA Respondent Brief, supra note 126, at 2.

With all this, why wasn’t the appropriate response to hold EPA’s error harmless—to take due account of prejudicial error? 130 See Administrative Procedure Act, 5 U.S.C. § 706 (2012). After all, there’s no reason to think the dispute between GE and EPA is unusual. When industry groups challenge agency failures to conduct notice-and-comment rulemaking, it often turns out that they worked closely with the agency on that very issue. Instead of assuming—often contrary to evidence—that those groups haven’t been heard, why not require them to explain why additional notice and comment would have done any good?

Yet the courts are hostile to that common-sense approach in cases involving nonlegislative rules. 131 See McLouth Steel Prods. Corp. v. Thomas, 838 F.2d 1317, 1324 (D.C. Cir. 1988) (“Even if the challenger presents no bases for invalidating the rule on substantive grounds, we cannot say with certainty whether petitioner’s comments would have had some effect if they had been considered when the issue was open.”). As the D.C. Circuit explained in Sugar Cane Growers Cooperative of Florida v. Veneman, “if the government could skip [notice-and-comment] procedures, engage in informal consultation, and then be protected from judicial review unless a petitioner could show a new argument—not presented informally—section 553 obviously would be eviscerated.” 132 289 F.3d 89, 96 (D.C. Cir. 2002). As a statement of remedial purity, it is difficult to improve on this. But the court’s conclusion is by no means “obvious.” To the contrary, it depends on the false assumption that judicial review is the only reason agencies bother to solicit public input.

B. Logical Outgrowth

When an agency proposes a rule, the APA requires it to supply “either the terms or substance of the proposed rule or a description of the subjects and issues involved.” 133 5 U.S.C. § 553(b)(3). Because the point of proposing a rule is to notify the public about what the agency means to do, it would be troubling if the agency issued a final rule that bore little resemblance to its proposal. The courts have thus insisted that the final rule be a “logical outgrowth” of the proposed rule. 134 Long Island Care at Home, Ltd. v. Coke, 551 U.S. 158, 174 (2007) (internal quotation marks omitted) (quoting Nat’l Black Media Coal. v. FCC, 791 F.2d 1016, 1022 (2d Cir. 1986)). The touchstone of that inquiry is fair notice: Did the agency’s proposal give the public enough information to enable it to offer cogent comments? 135 Id.

It’s often tricky to discern whether some aspect of a final rule is the logical outgrowth of a proposed rule. In particular, at what level of generality must the agency provide the requisite notice? By allowing either “a description of the subjects and issues involved” or a more granular proffer of “the terms or substance of the proposed rule,” 136 5 U.S.C. § 553(b)(3). the APA suggests that notice can be quite general. But the courts, keen to prevent agency evasion, have demanded something more. At the same time, however, they don’t want to insist on too tight a connection between the final and proposed rules. Agencies might then be skittish about adjusting their rules in response to sensible comments. 137 See United Steelworkers v. Marshall, 647 F.2d 1189, 1221 (D.C. Cir. 1980) (“A contrary rule would lead to the absurdity that in rule-making under the APA the agency can learn from the comments on its proposals only at the peril of starting a new procedural round of commentary.” (internal quotation marks omitted) (quoting Int’l Harvester Co. v. Ruckelshaus, 478 F.2d 615, 632 n.51 (D.C. Cir. 1973))). The courts must therefore balance the need for meaningful notice against the risk of enervating the comment process.

That leaves a lot of room for judgment—and a lot of room for agencies to guess wrong about how much they can change a final rule. As a result, the presumptive vacate-and-remand remedy will sometimes appear disproportionate to an agency’s understandable mistake. That’s especially so when the agency gave general notice of the disputed issue in the Federal Register and offered additional details through alternative means.

Yet nearly every logical-outgrowth violation leads to vacatur. 138 See, e.g., Chamber of Commerce v. SEC, 443 F.3d 890, 904 (D.C. Cir. 2006) (“The court has not required a particularly robust showing of prejudice in notice-and-comment cases . . . .”). Consider the Sixth Circuit’s recent nationwide injunction against a joint rule from the Army Corps of Engineers and EPA interpreting the Clean Water Act’s use of the phrase “waters of the United States.” 139 See In re EPA, 803 F.3d 804, 808–09 (6th Cir. 2015). In two cases in the 2000s, the Supreme Court rebuked agencies for taking too expansive a view of their jurisdiction. 140 See Rapanos v. United States, 547 U.S. 715, 716 (2006) (Scalia, J.) (plurality opinion); Solid Waste Agency of N. Cook Cty. v. U.S. Army Corps of Eng’rs, 531 U.S. 159, 170–71 (2001). Agencies’ authority to regulate “waters of the United States,” the Court reasoned, couldn’t be stretched to cover any “storm drains” and “roadside ditches” that might conceiv­ably connect to a national waterway. 141 Rapanos, 547 U.S. at 722; see also id. at 759 (Kennedy, J., concurring in the judgment) (requiring a “significant nexus” to navigable waters).

In response to the losses, the Army Corps and EPA conducted a joint rulemaking. The key issue was how to treat waters that were not them­selves navigable waterways, interstate waters, territorial seas (collectively, “jurisdictional waters”) or their tributaries but were nonetheless so functionally related to those waters that they could be treated as “waters of the United States.” 142 See 80 Fed. Reg. 37,054, 37,105 (June 29, 2015) (codified at 33 C.F.R. pt. 328 and in scattered parts of 40 C.F.R.). In the end, the agencies opted to define such “adjacent waters” with a bright-line rule: They included waters within 100 feet of the 100-year floodplain, subject to a maximum of 1,500 feet from other jurisdictional waters. 143 See id.

Eighteen states filed lawsuits challenging the bright-line distance limitations on the ground that they weren’t a logical outgrowth of EPA’s proposed rule. 144 In re EPA, 803 F.3d at 805. In a split decision on a motion for a stay of the agencies’ rule, the Sixth Circuit held that the states had a substantial chance of prevailing on the merits. 145 Id. at 807. Why? Because the proposed rule would have used a standard—not a bright-line rule—to define adjacent waters as those “located within the riparian area or floodplain” of juris­dictional waters and their tributaries. 146 See id.; see also 79 Fed. Reg. 22,188, 22,263 (proposed Apr. 21, 2014) (codified in scattered parts of 40 C.F.R.). As such, the Sixth Circuit concluded that the public lacked “reasonably specific notice that the distance-based limitations adopted in the Rule were among the range of alternatives being considered.” 147 In re EPA, 803 F.3d at 807. The court’s decision left no doubt that vacating the rule would eventually be the appropriate remedy for the infraction.

On the merits, the Sixth Circuit’s decision is open to question. In their proposed rule, the agencies explained that “best professional judgments” had in the past been the lodestar of the inquiry into which waters counted as “adjacent.” 148 79 Fed. Reg. at 22,208. Because this open-ended standard gen­erated uncertainty, the agencies suggested more concrete possibil­ities, including “specific geographic limits” and “distance limitations.” 149 Id. The agencies did not spell out the particular limits and distances that they had in mind, but the Army Corps and EPA didn’t spring the possibility of bright-line rules on the states with no warning. Far from consciously evading their APA obligations, the agencies believed that they had pro­vided enough notice.

Stipulate, however, that the agencies should have been more specific. The agencies’ error doesn’t appear to have prevented the submission of incisive comments on specific geographical limitations. In the million-plus comments that the agencies received, the topic came up time and again. 150 See 80 Fed. Reg. 37,054, 37,057 (June 29, 2015) (codified at 33 C.F.R. pt. 328 and in scattered parts of 40 C.F.R.) (observing that many comments on the proposed rule “urged EPA to improve upon the April 2014 proposal, by providing more bright line boundaries and simplifying definitions that identify waters that are protected under the [Clean Water Act]”). Indeed, it’s not too strong to say that how to define adjacent waters—those waters that would be treated as “waters of the United States” even if they are not themselves navigable or interstate waters—was at the heart of the rulemaking. Environmental groups cautioned the agencies against bright-line geographic limitations: NRDC, for example, argued that “[t]he agencies have no reasonable basis for requiring a certain degree of proximity in order for a water body to qualify as ‘adjacent,’ or for disregarding shallow subsurface connections.” 151 Nat. Res. Def. Council, Comment Letter on Proposed Rule to Define “Waters of the United States” Under the Clean Water Act (Nov. 14, 2014), as reprinted in Respondents’ Opposition to State Petitioners’ Motion for Stay of Clean Water Rule Pending Review: Addendum Volume I at 13, In re EPA, 803 F.3d 804 (6th Cir. 2015) (Nos. 15-3799, 15-3822, 15-3853, 15-3887) [hereinafter Addendum Volume I].

Who lined up in favor of bright-line limitations? The very states that later claimed to be surprised when EPA adopted them. 152 Respondents’ Opposition to State Petitioners’ Motion for Stay of Clean Water Rule Pending Review at 3–4, In re EPA, 803 F.3d 804 (Nos. 15-3799, 15-3822, 15-3853, 15-3887). Georgia, for instance, offered its “support” for “specific geographic limits” for demarking adjacent waters, “including, for example, distance limitations.” 153 Ga. Dep’t of Transp., Comment Letter on Proposed Rule to Define “Waters of the United States” Under the Clean Water Act (Nov. 13, 2014), as reprinted in Addendum Volume I, supra note 151, at 9. North Carolina criticized the agencies’ proposed standard-based approach to defining a floodplain and encouraged them to instead use flood frequency—like the 100-year floodplain that the agency eventually adopted. 154 N.C. Soil & Water Conservation Comm’n, Comment Letter on Proposed Rule to Define “Waters of the United States” Under the Clean Water Act (Oct. 22, 2014), as reprinted in Addendum Volume I, supra note 151, at 16. One is reminded of Judge Carl McGowan’s wry comment from Automotive Parts & Accessories Ass’n v. Boyd, in which the petitioners criticized an agency for adopting a position the petitioners had endorsed during notice and comment. Judge McGowan wrote, “We find it hard to take petitioners seriously on this score, despite their effort to analogize themselves to private attorneys general with an unlimited right to expose all dangers to the public interest.” 155 407 F.2d 330, 342 (D.C. Cir. 1968).

Throughout a lengthy, complex rulemaking process, the agencies repeatedly courted public input on whether they should adopt bright-line geographic limitations. During the rulemaking, EPA released a 408-page report reviewing the science on the connectivity of streams and wetlands to other waters. 156 U.S. EPA, EPA/600/R-14/475F, Connectivity of Streams & Wetlands to Downstream Waters: A Review & Synthesis of the Scientific Evidence (2015). It then enlisted an independent Scientific Advisory Board—which held four open meetings and solicited comments on EPA’s work—to review both the proposed rule 157 Sci. Advisory Bd., U.S. EPA, Comment Letter on the Proposed Rule Titled “Definition of Waters of the United States Under the Clean Water Act” (Sept. 30, 2014) [hereinafter Science Advisory Board, Comment Letter],$FILE/EPA-SAB-14-007+unsigned.pdf []. and the conclusions in that report. 158 Sci. Advisory Bd., U.S. EPA, Comment Letter on the Draft EPA Report Connectivity of Streams and Wetlands to Downstream Waters: A Review and Synthesis of the Scientific Evidence (Oct. 17, 2014),$File/EPA-SAB-15-001+unsigned.pdf []. Meetings with relevant stakeholders were constant. Throughout the process, the agencies held “more than 400 meetings with states, small businesses, farmers, academics, miners, energy companies, counties, municipalities, environmental organizations, federal agencies, and others.” 159 Respondents’ Opposition to State Petitioners’ Motion for Stay of Clean Water Rule Pending Review at 3, In re EPA, 803 F.3d 804 (6th Cir. 2015) (Nos. 15-3799, 15-3822, 15-3853, 15-3887).

In its final review, the Scientific Advisory Board advised EPA and the Army Corps not to define adjacent waters “solely on the basis of geographical proximity or distance to jurisdictional waters.” 160 Science Advisory Board, Comment Letter, supra note 157, at 2–3. The agen­cies disagreed:

Several commenters expressed the view that the proposed definition of “adjacent,” and in particular the definition of “neighboring,” focused too heavily on “geographic adjacency” and should be revised to focus on “functional adjacency.” . . . The agencies, in response to other comments, sought to promulgate a definition of “adjacent” that draws reasonable boundaries in order to protect the waters that clearly have a significant nexus while minimizing uncertainty about the scope of “waters of the United States.” . . . [T]he agencies set the distance limits for adjacency based on both functional relationships and proximity, because those factors together identify the waters that clearly have a strong influence on the chemical, physical, or biological integrity of traditional navigable waters, interstate waters, or the territorial seas. 161 U.S. EPA & Dep’t of the Army, Clean Water Rule Response to Comments-–-Topic 3: Adjacent Waters 20, [] (last visited Oct. 26, 2016).

In other words, the agencies sided with the states over environmental groups and even over their own advisory board.

Taken together, the wealth of commentary on bright-line distance limitations, the diversity of open discussion channels, and the agencies’ sensitivity to the states’ desire for administrable rules make it hard to believe the states suffered prejudice on account of any logical-outgrowth error. As is often the case with high-profile rulemakings, the notice-and-comment process was only one part of a long, intensive effort to secure public input. 162 See Croley, supra note 81, at 159–61 (providing examples). Yet the Sixth Circuit did not put the states to the pain of explaining what they would have said had the agencies been more specific at the outset. Still less did the court ask whether the states expressed those concerns in other venues or whether those concerns stood a material chance of changing the agency decision. Had it done so, enjoining the water rule might not have seemed like the obvious choice. It might have seemed silly.

Other logical-outgrowth cases have a similar structure. Take Allina Health Services v. Sebelius, which involved a technical dispute about the calculation of Medicare payments to hospitals that treat a disproportion­ate share of low-income patients. 163 746 F.3d 1102, 1105 (D.C. Cir. 2014). The question at hand was how to classify Medicare beneficiaries who had enrolled in a private managed-care plan. 164 Id. Hundreds of millions of dollars turned on whether those beneficiaries were “entitled to benefits under” traditional Medicare. 165 Id. In context, the phrase is ambiguous. 166 See Ne. Hosp. Corp. v. Sebelius, 657 F.3d 1, 13 (D.C. Cir. 2011). On the one hand, individuals can enroll in Medicare managed-care plans only if they are eligible to receive benefits under traditional Medicare. 167 See Allina Health Servs., 746 F.3d at 1106. On the other hand, an enrollee in a private plan no longer receives benefits under traditional Medicare. 168 See Ne. Hosp. Corp., 657 F.3d at 4.

Prior to 2003, Medicare had an informal practice of classifying beneficiaries enrolled in private plans as not entitled to traditional Medicare benefits. 169 Allina Health Servs., 746 F.3d at 1106. But the practice was inconsistently applied and subject to confusion. 170 Id. To clarify its approach, the Centers for Medicare and Medicaid Services (CMS) flagged the statutory ambiguity and proposed a rule that would track its general practice. 171 See 68 Fed. Reg. 27,154, 27,208 (proposed May 19, 2003) (codified at 42 C.F.R. pts. 412–413). After receiving comments, the agency issued a final rule that diverged from its proposal. Medicare beneficiaries who were enrolled in private plans, CMS concluded, should be treated as entitled to benefits under traditional Medicare. 172 Allina Health Servs., 746 F.3d at 1106.

A number of hospitals filed suit, arguing that the final rule was not the logical outgrowth of the proposed rule. 173 Id. at 1107. Their litigating position was awkward. The agency’s decision was a binary one—either those enrolled in private plans would count as eligible for traditional Medicare or they wouldn’t—and CMS had clearly put that binary in play. The court nonetheless concluded that CMS had “pull[ed] a surprise switcheroo” on the hospitals. 174 Id. at 1108 (internal quotation marks omitted) (quoting Envtl. Integrity Project v. EPA, 425 F.3d 992, 996 (D.C. Cir. 2005)). According to the court, “The hospitals should not be held to have anticipated that the Secretary’s ‘proposal to clarify’ could have meant that the Secretary was open to reconsidering existing policy.” 175 Id. The court therefore vacated the rule and remanded it to the agency. 176 Id. at 1111.

Even if the court was correct about the logical-outgrowth violation—and that’s debatable—it’s very hard to see how that violation harmed the plaintiffs. As it happened, CMS received lots of comments both for and against the particular classification decision. One representative of the hospital industry, for example, submitted four pages of close analysis in support of CMS’s proposed approach, drawing on the text, structure, and purpose of the Medicare statute. 177 Sw. Consulting Assocs., Comment Letter on Proposed Rule Regarding Medicare Program’s Prospective Payment Systems for Hospital Inpatient Rates (July 12, 2004), as reprinted in Joint Appendix at 84–87, Allina Health Servs., 746 F.3d 1102 (Nos. 13-5011, 13-5015) [hereinafter Allina Joint Appendix]. Several hospitals offered comments to the same effect. 178 See Mercy Health Sys., Comment Letter on Proposed Rule Regarding Medicare Program’s Prospective Payment Systems for Hospital Inpatient Rates (July 3, 2003), as reprinted in Allina Joint Appendix, supra note 177, at 96–97; Univ. of Pa. Health Sys., Comment Letter on Proposed Rule Regarding Medicare Program’s Prospective Payment Systems for Hospital Inpatient Rates (July 2, 2003), as reprinted in Allina Joint Appendix, supra note 177, at 101–02. On the other side, the Association of American Medical Colleges, a group representing more than 400 teaching hospitals, submitted a comment stating that it “disagree[s]” with what it saw as a “proposed change” in CMS’s practice. 179 Ass’n of Am. Med. Colls., Comment Letter on Proposed Rule Regarding Medicare Program’s Prospective Payment Systems for Hospital Inpatient Rates (July 8, 2003), as reprinted in Allina Joint Appendix, supra note 177, at 95. Enrollees in private plans, it wrote, “are just as much Medicare beneficiaries as beneficiaries enrolled in the fee-for-service program,” and “CMS’ proposal is unneces­sary, unwise, and should be abandoned.” 180 Id. A number of hospitals submitted similar comments 181 See NYU Med. Ctr., Comment Letter on Proposed Rule for Changes to the Hospital Inpatient Prospective Payment Systems and Fiscal Year 2004 Rates (July 7, 2003), as reprinted in Allina Joint Appendix, supra note 177, at 99–100; Greater N.Y. Hosp. Ass’n, Comment Letter on Proposed Rule for Changes to the Hospital Inpatient Prospective Payment Systems and Fiscal Year 2004 Rates (July 3, 2003), as reprinted in Allina Joint Appendix, supra note 177, at 107. —including two that later sued CMS for doing what they asked it to do. 182 Franklin Hospital Medicare System and North Shore Health System initially supported the approach that CMS took in its final rule. See Franklin Hosp. Med. Ctr., Comment Letter on Proposed Rule for Changes to the Hospital Inpatient Prospective Payment Systems and Fiscal Year 2004 Rates (July 8, 2003), as reprinted in Allina Joint Appendix, supra note 177, at 120; N. Shore Univ. Hosp., Comment Letter on Proposed Rule for Changes to the Hospital Inpatient Prospective Payment Systems and Fiscal Year 2004 Rates (July 8, 2003), as reprinted in Allina Joint Appendix, supra note 177, at 123. (Now that’s chutzpah.) Because CMS heard and had the chance to consider comments both for and against the proposal, any failure of the agency to emphasize its openness to abandoning its informal policy appears harmless.

Nonetheless, the D.C. Circuit dismissed the comments as too vague to amount to “focused opposition to the final rule.” 183 Allina Health Servs. v. Sebelius, 746 F.3d 1102, 1110 (D.C. Cir. 2014). Given the specificity of the submitted comments, the court’s decision is difficult to defend. 184 For just one example among many, see Southwest Consulting Associates, supra note 177, at 87 (objecting to the approach adopted in the final rule on the ground that it “violates the Medicare statute and is otherwise arbitrary and capricious because it deflates the resulting disproportionate patient percentage and results in systemic underpayment”). But it makes complete sense as a reflection of the court’s avowed commitment to remedial purity: “We have not been hospitable to government claims of harmless error in cases in which the government violated § 553 of the APA by failing to provide notice.” 185 Allina Health Servs., 746 F.3d at 1109. The court observed without deciding that the harmless error exception may be inconsistent with the judicial review provisions of the Medicare statute. See id.

One final example. To prevent mining accidents, the Mine Safety and Health Administration (MSHA) proposed a rule prescribing a minimum velocity—300 feet per minute—for the air flow over conveyer belts that hauled coal. 186 69 Fed. Reg. 17,480 (Apr. 2, 2014) (codified at 30 C.F.R. pt. 75); 68 Fed. Reg. 3936 (proposed Jan. 27, 2003) (codified at 30 C.F.R. pt. 75). In response to comments, the Agency’s final rule also established a maximum velocity—500 feet per minute—for air flow. 187 69 Fed. Reg. at 17,495. In 2005, in International Union v. Mine Safety & Health Administration, the D.C. Circuit ruled that the maximum-velocity cap was not a logical outgrowth of the proposed rule: “MSHA did not afford . . . public notice of its intent to adopt, much less an opportunity to comment on, such a cap.” 188 407 F.3d 1250, 1259–61 (D.C. Cir. 2005).

In so ruling, however, the court ignored that MSHA in fact received extensive feedback from a number of sources both for and against a maximum-velocity cap. The mines, for example, submitted comments in support of the Agency’s decision “that there should be no upper limits on the velocity of belt air.” 189 Brief for the Secretary of Labor at 37, Int’l Union, 407 F.3d 1250 (No. 04-1164), 2004 WL 3130624. Coal mines’ primary antagonist, the United Mine Workers of America, submitted an extensive comment, along with a concrete recommendation, urging MSHA to incorporate a cap into its final rule. 190 See, e.g., id. (recording a formal comment from the United Mine Workers of America that “it is not sufficient to make a determination regarding minimum velocity of air allowed to be coursed through the conveyor belt entry without also looking at what maximum should also be placed on it”). During the comment period, the Agency also held five public hearings at which union members vociferously objected to the absence of a cap. 191 See id. Indeed, representatives of the very petitioner that later brought the challenge in the D.C. Circuit—Jim Walter Resources, Inc. (JWR)—attended a hearing at which it was called out by name for failing to do enough to reduce air velocity in its mines. 192 Id. at 35. “I would invite each and every one of you,” one union member said, “to come to Jim Walter Number 5 Mine and let me show you what belt air can do in high pressure situations.” 193 Id.

More significantly, JWR was no newcomer to the rulemaking pro­ceedings. By its own description, JWR “played a prominent role” in the rulemaking process “over a sixteen-year period.” 194 Brief for Jim Walters Resources, Inc. at 7, Int’l Union, 407 F.3d 1250 (No. 04-1164), 2005 WL 139674. MSHA had first proposed rules for mine ventilation, including velocity rules, in 1988. 195 69 Fed. Reg. 17,480, 17,481–83 (Apr. 2, 2004) (codified at 30 C.F.R. pt. 75). After accepting comments and holding six public hearings, the agency called for a formal agency review of its ventilation rules. 196 Id. Having conducted that review, the Agency reopened the proposed rule for another round of notice and comment. 197 See 54 Fed. Reg. 35,356, 35,356 (Aug. 25, 1989). To solicit still more feedback on ventilation rules, MSHA then appointed an outside advisory commit­tee to study the question. 198 57 Fed. Reg. 57,078, 57,078 (Dec. 2, 1992). To the dismay of coal mines, the committee’s 1992 report—issued after holding six public meetings over six months—specifically recommended “both minimum and maximum” velocity rules. 199 Id. When MSHA finalized the proposed rule in 1996, however, it declined to impose velocity rules, instead reserving them for a later rule­making. 200 69 Fed. Reg. at 17,481–82. Only in January 2003 did the Agency launch the rulemaking that culminated, in 2004, in the final rule establishing maximum velocity caps. 201 Id. The preamble to the rule noted that the agency had received comments both in favor and against the caps and explained at length why it had been “persuaded that there is a need for a velocity cap.” 202 Id. at 17,495.

Given this iterative, lengthy process, it beggars belief that JWR lacked a meaningful opportunity to communicate its objections to a velocity cap. Only by ignoring how diligently MHSA sought to involve the public in its decisionmaking process could the court conclude that the proper remedy was to vacate the rule and give the company yet another chance to lobby the agency.

The court’s mistake is endemic to logical-outgrowth cases. 203 See, e.g., Chamber of Commerce v. SEC, 443 F.3d 890, 905–06 (D.C. Cir. 2006) (holding that the Chamber of Commerce, which was closely involved in the rulemaking, received inadequate notice of the SEC’s reliance on a publicly available report); Sprint Corp. v. FCC, 315 F.3d 369, 375–77 (D.C. Cir. 2003) (vacating a portion of a rule even though Sprint likely received notice of the rule and identified nothing the company could have said that might have changed the Agency’s mind); Shell Oil Co. v. EPA, 950 F.2d 741, 752 (D.C. Cir. 1991) (rejecting EPA’s claim that “it had considered and rejected the points raised by petitioners” and relieving petitioners of the responsibility “to show that they would have submitted new arguments” had the Agency been more forthcoming). For agency decisions of any importance—which is to say, for any decision worth suing over—federal agencies will typically work closely with regulated industries and other interested groups to fashion a rule that is both workable and broadly acceptable. 204 See supra note 87 and accompanying text. When a party challenging an agency rule has taken advantage of that opportunity, it’s very hard to see the point of forcing the agency back to the drawing board.

C. Good Cause

With some regularity, agencies adopt legislative rules without afford­ing an opportunity for notice and comment, most frequently because they decide they have “good cause” for thinking that “notice and public procedure thereon are impracticable, unnecessary, or contrary to the public interest.” 205 Administrative Procedure Act, 5 U.S.C. § 553(b)(3)(B) (2012); see also U.S. Gov’t Accountability Office, supra note 84, at 15 (finding that agencies invoked the good-cause exception seventy-seven percent of the time for major rules and sixty-one percent of the time for nonmajor rules). The courts can and do review agency judgments about good cause; indeed, the D.C. Circuit has said that it owes agencies “no particular deference” on that question. 206 See Mack Trucks, Inc. v. EPA, 682 F.3d 87, 93 (D.C. Cir. 2012). The circuits are split on whether to review good-cause determinations de novo or for arbitrariness. See United States v. Reynolds, 710 F.3d 498, 506–09 (3d Cir. 2013) (examining the split at length). The difference may be more apparent than real: Even those circuits that employ a standard that is nominally more deferential say that the good-cause exception must be narrowly construed. Id. As such, the courts will sometimes countermand an agency’s good-cause determina­tion, espe­cially given their insistence that good cause “is to be narrowly construed and only reluctantly countenanced.” 207 Util. Solid Waste Activities Grp. v. EPA, 236 F.3d 749, 754 (D.C. Cir. 2001) (internal quotation marks omitted) (quoting Tenn. Gas Pipeline Co. v. FERC, 96 F.2d 1141, 1144 (D.C. Cir. 1992)). Vacatur is the conven­tional remedy. 208 E.g., Mack Trucks, Inc., 682 F.3d at 95.

In a substantial fraction of cases, however, the punishment doesn’t seem to fit the crime. The flexibility of the good-cause exception means that an agency will sometimes believe that good cause exists, only to learn after the fact that the courts disagree. An agency’s mistake may thus reflect an error of judgment, not a deliberate effort to evade procedural restraints. That’s especially so when an agency solicits comments on a rule after its publication and respond in writing to those comments. 209 See U.S. Gov’t Accountability Office, supra note 84, at 24 (finding that agencies sought postpublication comment for sixty-three percent of a sample of major rules issued without a notice of proposed rulemaking and responded to comments sixty-six per­cent of the time). Might the postpublication comment period cure the procedural deficiency?

As Professors Kristin Hickman and Mark Thomson have recently shown, there’s considerable diversity among the circuit courts in how to treat postpublication comment periods. 210 See Kristin E. Hickman & Mark Thomson, Open Minds and Harmless Errors: Judicial Review of Postpromulgation Notice and Comment, 101 Cornell L. Rev. 261, 276–81 (2016). Some courts have staked out positions suggesting an absolute unwillingness to accept any kind of postpublication comment period as a cure. 211 See U.S. Steel Corp. v. EPA, 595 F.2d 207, 215 (5th Cir. 1979) (holding an agency’s failure to observe notice and comment is harmless only “when a mistake of the administrative body is one that clearly had no bearing on the procedure used or the substance of decision reached” (emphasis added) (internal quotation marks omitted) (quoting Braniff Airways, Inc. v. Civil Aeronautics Bd., 379 F.2d 453, 466 (D.C. Cir. 1967))). The Fifth Circuit, for example, has said that doing so would “make the provisions of § 553 virtually unenforceable. An agency that wished to dispense with pre-promulgation notice and comment could simply do so, invite post-promulgation comment, and republish the regulation before a reviewing court could act.” 212 Id. The D.C. Circuit is mildly more flexible: Post­publication notice and comment can save a rule if an agency makes a “compelling showing” that it kept an open mind upon receipt of postpublication comments. 213 Air Transp. Ass’n of Am. v. Dep’t of Transp., 900 F.2d 369, 379 (D.C. Cir. 1990), 933 F.2d 1043 (D.C. Cir. 1991) (per curiam) (mem.); see also McClouth Steel Prods. Corp. v. Thomas, 838 F.2d 1317, 1323 (D.C. Cir. 1988) (holding “defects in an original notice may be cured by an adequate later notice, but that curative effect depends on the agency’s mind remaining open enough at the later stage” (citation omitted)). On the whole, however, the consensus in the courts is skepticism: When an agency mistakenly skips notice and comment, the presumptive remedy is vacatur, whether or not the agency has solicited further comments. Hickman and Thomson share that skepticism. They argue for a “strong presumption” that postpublication notice and comment can’t cure notice-and-comment errors. 214 Hickman & Thomson, supra note 210, at 311. Professor Michael Asimow has advanced the contrary argument: “The public’s opportunity to comment on the interim-final rule was sufficient to meet the APA’s standards; the error relating to adoption of the interim-final rule makes little difference to anyone once the final-final rule has supplanted the interim-final rule.” Michael Asimow, Interim-Final Rules: Making Haste Slowly, 51 Admin. L. Rev. 703, 726 (1999).

That skepticism can lead to odd results. Take the recent fracas over the Sex Offender Registration and Notification Act (SORNA), which established a national system for registering sex offenders in 2006 and criminalized interstate travel for those who fail to register. 215 18 U.S.C. § 2250(a) (2012). The Supreme Court has held that the statute does not, by its own force, apply to sex offenders convicted prior to the statute’s adoption. 216 Reynolds v. United States, 132 S. Ct. 975, 984 (2012). But SORNA contains a provision delegating to the Attorney General “the authority to specify the applicability of [SORNA’s] requirements . . . to sex offenders convicted before the enactment of this [Act].” 217 42 U.S.C. § 16913(d) (2012). In 2007, the Attorney General exercised that authority to issue—without going through notice and comment—an interim final rule concluding that SORNA would apply to sex offenders with pre-2006 convictions. 218 72 Fed. Reg. 8894 (Feb. 28, 2007) (codified at 28 C.F.R. pt. 72). The Attorney General concluded that good cause existed for skipping notice and comment: “Delay in the implementation of this rule would impede the effective registration of . . . sex offenders and would impair immediate efforts to protect the public from sex offenders who fail to register . . . .” 219 Id. at 8896.

The Attorney General’s conclusion that public safety demanded urgent action was not unreasonable; indeed, the Fourth and Eleventh Circuits agreed with the Attorney General’s good-cause determination. 220 See United States v. Dean, 604 F.3d 1275, 1278–82 (11th Cir. 2010); United States v. Gould, 568 F.3d 459, 469–70 (4th Cir. 2009). But the Third, Fifth, Sixth, Eighth, and Ninth Circuits did not. 221 See United States v. Brewer, 766 F.3d 884, 890 (8th Cir. 2014); United States v. Reynolds, 710 F.3d 498, 509 (3d Cir. 2013); United States v. Johnson, 632 F.3d 912, 927–30 (5th Cir. 2011); United States v. Valverde, 628 F.3d 1159, 1165–68 (9th Cir. 2010); United States v. Cain, 583 F.3d 408, 419–24 (6th Cir. 2009). Among other things, the courts read the delegation to the Attorney General to reflect Congress’s judgment that the retroactivity question demanded careful agency deliberation. After all, “Congress could have expressly waived the APA procedural requirements in SORNA if it feared those requirements would produce significant harm or excessive delay.” 222 Johnson, 632 F.3d at 928.

Of the five courts to hold that the Attorney General inappropriately invoked the good-cause exception, four vacated the interim final rule. They did so even though the Attorney General anticipated and addressed a number of objections in the interim final rule itself; 223 72 Fed. Reg. at 8896–97. even though he formally invited post-promulgation comments; 224 Id. at 8895. even though, within three months, he issued a notice of proposed rulemaking recommitting to the interim final rule; 225 72 Fed. Reg. 30,210 (proposed May 30, 2007). and even though he then issued a final rule that responded to all the comments he received on SORNA’s application to pre-2006 offenders. 226 75 Fed. Reg. 81,849 (Dec. 29, 2010) (codified at 28 C.F.R. pt. 72). All of this activity, the courts reasoned, couldn’t cure the Attorney General’s improper invocation of the good-cause exception. The courts spoke the language of remedial purity. The Third Circuit, for example, reasoned that holding the error harmless would “eviscerate” the notice-and-comment requirements. 227 United States v. Reynolds, 710 F.3d 498, 509 (3d Cir. 2013). And the Eighth Circuit lambasted the Attorney General for suggesting that the interim final rule would have been no different had he accepted comments before issuing it: “The Attorney General’s attempt to foreclose the possible claims of pre-Act offenders seems incompatible with his duty seriously to consider whether SORNA applies to those offenders, and if so, which ones.” 228 United States v. Brewer, 766 F.3d 884, 892 (8th Cir. 2014).

Only the Fifth Circuit declined to vacate the interim final rule. As the court saw it, the goals of notice and comment “may be achieved in cases where the agency’s decision-making process ‘centered on the identical substantive claims’ as those proposed by the party asserting error, even if there were APA deficiencies.” 229 United States v. Johnson, 632 F.3d 912, 931 (5th Cir. 2011) (quoting Friends of Iwo Jima v. Nat’l Capital Planning Comm’n, 176 F.3d 768, 774 (4th Cir. 1999)). And the Fifth Circuit recognized that the Attorney General anticipated the defendant’s objections in the interim rule and responded again to those objections in the final rule. “There is no suggestion that, if given the opportunity to comment, [the defendant] would have presented an argument the Attorney General did not consider in issuing the interim rule.” 230 Id. at 932. The court was cautious about its invocation of harmless error—it flagged “the limited role of the harmless error doctrine in administrative law” 231 Id. at 931. —but its decision is a rare instance of a court that refused to indulge in the easy pieties of remedial purity.

Still, SORNA is unusual: The courts might reasonably be reluctant to find harmless error in a case in which a procedurally defective rule leads to a criminal conviction. 232 See Reynolds, 710 F.3d at 515 (declining to find harmless error in a criminal conviction case). The D.C. Circuit’s decision in Utility Solid Waste Activities Group v. EPA is more typical—and reflects the same impulse to vacate without sufficient regard to the functions that notice and comment is meant to serve. 233 236 F.3d 749, 752–53 (D.C. Cir. 2001) (analyzing whether EPA could amend its regulation without undergoing notice and comment). At issue in the case was EPA’s effort to correct a mistake in a rule that establishes stringent conditions for clean­ing up porous surfaces (like concrete) that are contaminated with PCBs. 234 40 C.F.R. § 761.30(p) (2016). In a technical revision—adopted without notice and comment—EPA amended the rule. 235 Util. Solid Waste Activities Grp., 236 F.3d at 755.

As a result of the amendment, the stringent clean-up conditions would apply to a larger number of porous surfaces than they would have under the text of the original rule. 236 Id. Industry groups sued. EPA asserted that it had good cause for avoiding notice and comment, arguing that further comment was “unnecessary” because it was just trying to correct a mistake. 237 Id. at 754. The D.C. Circuit disagreed, reasoning that the change mate­rially altered the obligations of the regulated firms and that the “amendment was, without doubt, something about which these members of the public were greatly interested.” 238 Id. at 755. The rule was therefore vacated. 239 Id.

All of this looks entirely conventional. Digging deeper, however, the court’s decision becomes very hard to defend. When EPA first issued its notice of proposed rulemaking, the Agency had included no accom­modation at all for porous materials that had been contaminated with PCBs. 240 56 Fed. Reg. 26,738 (June 10, 1991) (codified at 40 C.F.R. pt. 761 (2016)). If that proposal had been finalized, all such materials would have had to be discarded. The regulated community—including Utility Solid Waste Activities Group (USWAG) and GE—feared that such an approach would prove needlessly costly. 241 63 Fed. Reg. 35,384, 35,398 (June 29, 1998) (codified at 40 C.F.R. 750, 761). Sensitive to the concern, EPA worked closely with the industry during the original notice-and-comment period to devise an alternative. 242 Id.

It was here that the Agency erred. Its final rule stated that PCB contamination of porous materials would be measured with reference to their surface concentration. 243 40 C.F.R. § 761.30(p) (1998). To those in the know, this was obviously a mistake. (An agency official had apparently screwed up using the find-and-replace command. 244 Util. Solid Waste Activities Grp., 236 F.3d at 752. ) Because PCBs can leak into porous materials, surface concentration doesn’t speak to overall contamination. 245 See 40 C.F.R. § 761.3 (2015) (defining a porous surface as one “that allows PCBs to penetrate or pass into itself”). That was why EPA needed a specific rule governing porous surfaces in the first place.

No neophytes to PCBs, USWAG and GE understood as much—and were likely unsurprised when EPA caught its mistake. More importantly, the agency gave them notice of its desire to fix the mistake and a chance to comment. Nine months before amending its rule, 246 64 Fed. Reg. 33,755, 33,760 (June 24, 1999) (codified at 40 C.F.R. pt. 761). EPA published an internet bulletin containing a list of anticipated technical corrections—including the change that became the subject of Utility Solid Waste Activities Group—and invited additional contributions to the list. 247 See Brief for Respondent at 11–12, Util. Solid Waste Activities Grp., 236 F.3d 749 (Nos. 99-1372, 99-1374), 2000 WL 35585189 (noting “[t]he Agency quickly compiled a list of . . . errors and their corrections and posted the list on its Internet website”). EPA then held a number of meetings with both USWAG and GE; at one of those meetings, USWAG asked specifically about the change to the porous-surface rule and was told that the original rule was an error. 248 Id. at 13. After all this, it’s difficult to accept USWAG’s and GE’s claims that they never had a chance to comment on the correction.

For a final example, consider Action on Smoking & Health v. Civil Aeronautics Board. 249 713 F.2d 795 (D.C. Cir. 1983). In 1981, the Civil Aeronautics Board (CAB) relaxed its existing rules offering certain protections on airplanes to nonsmokers  who wished to avoid other passengers’ cigarette smoke. 250 See id. at 797 n.1. Although CAB had adhered to the conventional notice-and-comment procedure, the D.C. Circuit originally vacated the rule because the Agency offered a “palpably inadequate” response to the many comments that it received. 251 See Action on Smoking & Health v. Civil Aeronautics Bd., 699 F.2d 1209, 1215–17 (D.C. Cir. 1983). On remand, CAB reissued the same rule—but without walking through the notice-and-comment process. In the Agency’s view, “the applicable notice of proposed rulemaking and record are still outstanding,” leading it to “doubt that further comments would produce any additional light.” 252 Action on Smoking & Health, 713 F.2d at 800 (internal quotation marks omitted) (quoting Part 252, Smoking Aboard Aircraft, Order No. 83–5–101, 1983 WL 35235 (CAB May 19, 1983) (order denying stay)). Instead, the agency offered a more extensive response to the comments that it had already received. 253 Id.

The D.C. Circuit invalidated the rule again. The court brusquely dismissed CAB’s argument that putting the rule out for comment a second time was “unnecessary” within the meaning of the good-cause exception: “Bald assertions that the agency does not believe comments would be useful cannot create good cause to forgo notice and comment procedures.” 254 Id. But this wasn’t just a bald assertion. The plaintiff in the lawsuit—an antismoking organization—had participated fully in the ini­tial notice-and-comment process. Although the organization claimed that it had new comments that it wished to submit, the court acknowledged that it “may have had other opportunities to bring some or all of this information to the attention of the Board.” 255 Id. at 801 n.6. Nonetheless, the D.C. Circuit never asked whether it might excuse strict compliance with § 553 on the ground that the plaintiff could not demonstrate prejudice from the Agency’s error. 256 For other examples in which vacatur in response to an inappropriate invocation of the good-cause exception appears disproportionate, see, e.g,. Nat. Res. Def. Council, Inc. v. Evans, 316 F.3d 904, 909–13 (9th Cir. 2003); Buschmann v. Schweiker, 676 F.2d 352, 355–58 (9th Cir. 1982). Remedial purity demanded invalidation.




The notice-and-comment process is one way to encourage public feedback, but it is by no means the only one, or even the most effective one. As these examples demonstrate, agencies routinely maintain close working relationships with the regulated community as they pull through difficult sets of issues. Those relationships should make a difference in deciding whether a party challenging an agency action has suffered actual prejudice. If the party has had a full and fair opportunity to voice its objections, if the agency has heard and considered the substantive concerns that the party would have raised, or if there’s no substantial reason to think that the party’s comments would have led the agency to change its mind, it becomes difficult to see how the party has been harmed by the agency’s failure to adhere to § 553.

At times, the courts have acknowledged as much. For a ripped-from-the-headlines example, consider the D.C. Circuit’s “net neutrality” deci­sion, in which the court upheld the FCC’s decision to classify internet providers as common carriers. 257 See U.S. Telecom Ass’n v. FCC, 825 F.3d 674, 689, 725–26 (D.C. Cir. 2016). For a similarly recent example of remedial restraint, see Pub. Emps. for Envtl. Responsibility v. Hopper, 827 F.3d 1077, 1083–84 (D.C. Cir. 2016), in which the court ordered the Bureau of Ocean Energy Management to redo an inadequate environmental impact statement but nonetheless upheld the Bureau’s approval of an offshore lease to a wind farm operator. For another example, see U.S. Telecom Ass’n v. FCC, 400 F.3d 29, 41–42 (D.C. Cir. 2005) (holding any notice-and-comment error harmless because the agency offered adequate notice and “invited and received comment from the industry”). The providers had argued (among many other things) that the FCC supplied inadequate notice of its intent to redefine a particular regulatory term. 258 See U.S. Telecom Ass’n, 825 F.3d at 712. Without addressing the merits of that claim, the D.C. Circuit held that any error was harmless. As the court explained, the challengers “raised and fiercely debated all of the same arguments they now raise before us, thus demonstrating not only the presence of actual notice, but also the absence of new arguments they might present to the Commission on remand.” 259 See id. at 726. The decision offers a heartening example of remedial sensitivity with respect to notice-and-comment rulemaking. But it’s an example that is conspicuous for its rarity.

More often, the courts treat notice-and-comment failures like struc­tural trial errors—the sorts of mistakes that require automatic reversal, without any opportunity to demonstrate lack of prejudice. The purist approach follows naturally from the harmlessness standard that many lower courts apply: that the “failure to provide notice and comment is harmless only where the agency’s mistake clearly had no bearing on the procedure used or the substance of decision reached.” 260 See Cal. Wilderness Coal. v. Dep’t of Energy, 631 F.3d 1072, 1090 (9th Cir. 2011) (internal quotation marks omitted) (quoting Riverbend Farms, Inc. v. Madigan, 958 F.2d 1479, 1487 (9th Cir. 1992)). Read literally, the standard eliminates any role for the rule of prejudicial error: Skip­ping notice and comment necessarily bears “on the procedure used,” so improperly skipping it can never be harmless. 261 See id. at 1112 (Ikuta, J., dissenting) (noting the test “suggests that a procedural error is prejudicial per se”).

The standard also pinions agencies on the horns of a dilemma. To demonstrate the absence of prejudice, they must argue that they would “clearly” not have changed their minds in response to any comments the challenging party might have submitted. But that argument carries with it the implication that the agency would have treated the formal notice-and-comment process as a charade. Given the judicial conceit that agencies must keep an open mind during the notice-and-comment period, agencies can’t say that there is zero chance they would have changed their minds. The standard thus disables agencies from making a full-throated harmlessness argument.

Instead of presuming harm, why not insist on a demonstration of prejudice before invalidating agency rules? There’s no magic in strict adherence to notice-and-comment formalities. As a matter of due process, notice and comment isn’t required. 262 See Bi-Metallic Inv. Co. v. Colorado, 239 U.S. 441, 445 (1915). The “good cause” exception, available when “notice and public procedure thereon are impracticable, unnecessary, or contrary to the public interest,” 263 Administrative Procedure Act, 5 U.S.C. § 553(b) (2012). affords courts one reason to excuse compliance with the notice-and-comment rules. But the existence of one excuse doesn’t imply the nonexistence of others. To the contrary, the rule of prejudicial error applies, per § 706, to judicial review of any agency action, including informal rulemaking. 264 See id. § 706. Just as courts in the 1970s intensified the rigors of notice and comment to accord with their views about what it ought to accomplish, so too could courts today adjust what the rule of prejudicial error entails. 265 For a defense of administrative common law, see generally Gillian E. Metzger, Embracing Administrative Common Law, 80 Geo. Wash. L. Rev. 1293 (2012). For notice-and-comment cases, the rule could easily be understood to require courts to undertake a context-sensitive inquiry into prejudice.

What might an invigorated rule of prejudicial error look like? Consider the review standard incorporated into a portion of the Clean Air Act, which instructs courts to invalidate an agency action for procedural errors “only if the errors were so serious and related to matters of such central relevance to the rule that there is a substantial likelihood that the rule would have been significantly changed if such errors had not been made.” 266 42 U.S.C. § 7607(d)(8) (2012). The instruction appears to oblige courts to exercise remedial restraint in a particular subset of notice-and-comment cases. To date, the courts have shirked that responsibility. In 1983, the D.C. Circuit somehow concluded that the (relaxed) Clean Air Act stan­dard merely restates the (stringent) remedial rule that typically applies in APA cases. 267 See Small Refiner Lead Phase-Down Task Force v. EPA, 705 F.2d 506, 521–23 (D.C. Cir. 1983) (“[F]ailure to observe the basic APA procedures, if reversible error under the APA, is reversible error under the Clean Air Act as well.”). The opinion is curious: Far from restating a rigid presump­tion of vacatur, the statutory instruction is tailor-made to undo it. But the important point for present purposes is that nothing in the APA prevents the courts from adopting the Clean Air Act standard as a gloss on what it means to take “due account . . . of the rule of prejudicial error.” 268 5 U.S.C. § 706. The common law of administrative law could easily accommo­date an approach that required a discriminating look into prejudice. 269 The D.C. Circuit may already apply a heightened prejudice standard when the challenging party objects that the agency relied on studies that it failed to disclose. See, e.g., Pers. Watercraft Indus. v. Dep’t of Commerce, 48 F.3d 540, 544 (D.C. Cir. 1995); Small Refiner Lead Phase-Down Task Force, 705 F.2d at 540–41.

Invigorating the prejudice inquiry would sometimes require courts to look beyond the documents compiled during notice and comment. But nothing in the APA precludes that approach; to the contrary, the APA instructs reviewing courts to review “the whole record.” 270 5 U.S.C. § 706. At least outside the context of formal proceedings, that record can consist of whatever influenced the agency’s decisionmaking. 271 See Citizens to Pres. Overton Park, Inc. v. Volpe, 401 U.S. 402, 420 (1971) (“That review is to be based on the full administrative record that was before the Secretary at the time he made his decision.”); 39 Fed. Reg. 23,033, 23,044 (June 26, 1974) (providing recommendation from the Administrative Conference of the United States that a rulemaking record include, among other things, “factual information . . . that was considered by the authority responsible for promulgation of the rule”); William F. Pedersen, Jr., Formal Records and Informal Rulemaking, 85 Yale L.J. 38, 61–66 (1975) (discussing the APA’s silence on what constitutes a rulemaking record). True, the courts will generally confine their inquiry to the record generated during notice and comment in deciding whether an agency has violated § 553. 272 See, e.g., Solite Corp. v. EPA, 952 F.2d 473, 484 (D.C. Cir. 1991) (noting, in general, “[a]n agency commits serious procedural error when it fails to reveal portions of the technical basis for a proposed rule in time to allow for meaningful commentary” (quoting Conn. Light & Power Co. v. Nuclear Regulatory Comm’n, 673 F.2d 525, 530–31 (D.C. Cir. 1982))). Otherwise, the agency could base its decision on undisclosed materials, undermining notice and comment. 273 See United States v. N.S. Food Prods. Corp., 568 F.2d 240, 251–52 (2d Cir. 1977). But once a procedural deficiency has been identified, the question is no longer whether the agency made a procedural error. The court has already concluded that it has. The question, instead, is what the court ought to do about it. 274 Id. at 249. At that point, the categorical objection to reviewing “the whole record” boils down to the objection that it’s never appropriate to let a notice-and-comment violation slide. In other words, it reflects a commitment to remedial purity at all costs. If that commitment is misplaced, so too is the categori­cal objection.

To be clear, there are costs associated with shifting away from a rigid, prophylactic approach. It might license agency carelessness with respect to notice and comment. Courts would face the taxing respon­sibility of sifting an expansive record to determine whether the parties challenging the rule suffered any real harm from the procedural viola­tion. And the indeterminacy of that counterfactual inquiry would yield uncertainty about the appropriate remedy, as Part IV will discuss. For now, it’s just important to notice that the costs of remedial purity can be large and its benefits elusive.

III. Inadequate Explanations

Apart from failures of notice and comment, agency decisions can be vacated when the reasons agencies proffer for their decisions are incorrect, confused, or inadequate. Even when agency rationales are deficient, however, vacating and remanding will at times appear excessive. In such cases, the appropriate course may be to uphold the agencies’ actions on the ground that the complaining parties have suffered no prejudice.

But didn’t the Supreme Court foreclose that course of action in SEC v. Chenery Corp.? 275 318 U.S. 80 (1943). There, the Court rejected the justification that the Securities and Exchange Commission (SEC) proffered for approving a company’s merger plan only on the condition that the managers of the company surrender certain preferred shares that were acquired while the merger was under consideration. 276 See id. at 93. In so doing, the Court declined to pass on an alternative justification that the SEC proffered in litigation, instead confining its review “to a judgment upon the validity of the grounds upon which the Commission itself based its action.” 277 Id. at 88. The Court reasoned that upholding the decision on the alternative ground would be inappropriate:

If an order is valid only as a determination of policy or judgment which the agency alone is authorized to make and which it has not made, a judicial judgment cannot be made to do service for an administrative judgment. For purposes of affirming no less than reversing its orders, an appellate court cannot intrude upon the domain which Congress has exclusively entrusted to an administrative agency. 278 Id.

Not unreasonably, this passage—with its absolutist refusal to “intrude upon the domain” of agency decisionmaking 279 Id. —is widely understood to reject all but the most limited role for harmless error in arbitrariness review.

But notice how curious the passage is. The SEC never asked the Supreme Court to substitute its own “determination of policy or judgment” for the agency’s. 280 Id. The SEC instead asked the Court to uphold the order on the strength of the agency’s alternative justification. Deferring to that alternative wouldn’t have intruded on the agency’s domain or otherwise shifted executive power to the judicial branch. Quite the opposite: It would have marked the Court’s refusal to interfere in an exercise of executive power.

Chenery’s real objection was that the SEC’s alternative justification was made in the wrong manner and at the wrong time. The Commission could have—but did not—adopt a “general rule of which its order here was a particular application.” 281 Id. at 92. In the absence of such a rule, the Court reasoned, the SEC’s imposition of conditions on a merger was improper. 282 See id. at 92–93 (“[B]efore transactions otherwise legal can be outlawed or denied their usual business consequences, they must fall under the ban of some standards of conduct prescribed by an agency of government authorized to prescribe such standards . . . .”). What’s more, the Commission’s alternative justification came too late. Why? Because “the orderly functioning of the process of review requires that the grounds upon which the administrative agency acted be clearly disclosed and adequately sustained.” 283 Id. at 94.

In other words, Chenery is best understood to reflect the Court’s pragmatic judgment that the refusal to accept the Agency’s alternative justification would yield two systemic benefits. First, vacating would encourage agencies to proceed by rulemaking instead of ad hoc adju­dication. Second, vacating would force agencies to offer nonarbitrary reasons at the time of decision. Agencies might otherwise grow careless, confident they could rehabilitate a defective action by articulating a better set of reasons in the event they faced a lawsuit. These systemic benefits, in the Court’s judgment, amply justified what the dissent saw as a wasteful remand. 284 See id. at 99 (Black, J., dissenting) (“The Court can require the Commission to use more words; but it seems difficult to imagine how more words or different words could further illuminate its purpose or its determination.”).

Upon reconsideration, the SEC, without adopting a rule, reinstated its decision based on its alternative justification. In Chenery II, the Court upheld that decision, discarding the portion of Chenery that had suggested the SEC had to proceed by rulemaking before applying a new rule. 285 See SEC v. Chenery Corp. (Chenery II), 332 U.S. 194, 202 (1947). What’s left of Chenery is a prophylactic rule that creates incentives for agencies to offer valid reasons for their decisions when they make them. That incentive function is critically important; indeed, it explains why Chenery has become so firmly stitched in administrative law. And it explains, too, why Chenery requires the vacatur of some agency decisions even when there’s little doubt that agencies will reinstate them on remand.

But Chenery need not be absolute to serve that incentive function. In other enclaves of the law, prophylactic rules are relaxed when strict adherence would yield an especially senseless result and offer exiguous systemic benefits. Consider Miranda. 286 Miranda v. Arizona, 384 U.S. 436 (1966). Even when the risks of uncon­stitutionally coercing a confession are slight, the general rule is that a suspect’s statements are inadmissible in court unless that suspect received a Miranda warning. 287 See Evan H. Caminker, Miranda and Some Puzzles of “Prophylactic” Rules, 70 U. Cin. L. Rev. 1, 1 (2001). Otherwise, police officers might skip the warning and attempt to persuade a court to admit a suspect’s confession anyhow. In this, both Miranda and Chenery use the risk of judicial invalid­ation to encourage state actors (police departments and federal agencies, respec­tively) not to cut corners.

Yet the Supreme Court has allowed the introduction of unwarned statements in discrete categories of cases in which the costs of exclusion appear to outweigh the benefits: “Fidelity to the doctrine announced in Miranda requires that it be enforced strictly, but only in those types of situations in which the concerns that powered the decision are impli­cated.” 288 Berkemer v. McCarty, 468 U.S. 420, 437 (1984). Miranda, for example, does not apply to a suspect’s responses to questions asked at a routine traffic stop 289 See id. or to questions asked in an effort to protect the public safety. 290 See New York v. Quarles, 467 U.S. 649, 657 (1984). Unwarned responses to police questioning can also be used to impeach a defendant’s testi­mony 291 See Harris v. New York, 401 U.S. 222, 225 (1971). or to develop admissible evidence. 292 See Michigan v. Tucker, 417 U.S. 433, 446 (1974).

In each of these categories, the Court has reasoned that the costs of excluding probative evidence outweigh whatever systemic benefits a closer adherence to Miranda might entail. The same contextually sen­sitive approach can and should hold for administrative law. In his seminal article, for example, Judge Friendly argued that Chenery’s rule should be confined to those cases in which an agency supplied a wrong or unexplained reason in announcing its decision. In those cases, he wrote, vacating and remanding “permits a court to in effect say to an agency, ‘Do you really mean it?’” 293 Friendly, supra note 28, at 207–08. In Judge Friendly’s view, however, reversal is unwarranted when an agency errs with respect to subsidiary factual findings but adequate, alternative facts support the agency’s decision and there is no reason to believe the agency would reconsider if apprised of its mistake. 294 Id. at 211. “[C]ourts should not be obtuse,” he wrote, “to reasons implicit in the determination itself and thereby cause needless expense and delay.” 295 Id. at 222. Judge Friendly’s argument had special resonance for the ratemaking cases that characterized much agency law in earlier genera­tions. In complex cases involving the finances of large railroad concerns, subsidiary errors were common—indeed, were to be ex­pected. 296 See id. The wastefulness of vacating every decision infected with that kind of error was a powerful argument for declining to read Chenery for all it was worth.

The conventional wisdom is that this friendly amendment to Chenery has not been warmly received. 297 See Kevin M. Stack, The Constitutional Foundations of Chenery, 116 Yale L.J. 952, 964 (2007) (“The courts, however, have generally not heeded [Friendly’s] distinctions.”). Indeed, fealty to Chenery has often led the courts to invalidate agency actions even in cases in which the costs of vacatur appear to outweigh any plausible calculation of benefits. 298 See infra section III.A. At the same time, however, the courts have quietly developed two remedial approaches—one primarily for orders and the other for rules 299 See infra sections III.B–.C. —to soften the blow of Chenery. Because the Supreme Court has never ad­dressed whether the approaches are consistent with Chenery, and because they are to some extent embattled in the courts, they hold a tenuous position in administrative law. But these doctrinal innovations are not aberrations. They respond to the courts’ tacit recognition that, in some cases, the costs of remedial purity outweigh the benefits. More generally, they suggest that administrative law should start taking remedial discre­tion more seriously—both the discretion that courts already exercise and the discretion that they should consider exercising more frequently.

A. Failures to Resolve Statutory Ambiguities

Under Chevron, agencies may ascribe a determinate meaning to a statute even when the statute could reasonably be construed dif­ferently. 300 Chevron U.S.A., Inc. v. Nat. Res. Def. Council, Inc., 467 U.S. 837, 843–44 (1984). In so doing, they’re expected to explain why they selected the one interpretation and discarded the other. Sometimes, however, an agen­cy doesn’t see any ambiguity to resolve. The agency interprets the statute in a particular way not because that interpretation makes the most sense in light of the statute’s objectives and the agency’s goals but because it thinks the statute can’t be read any other way. When a court later determines that the statute is ambiguous, the agency’s failure to exercise its interpretive discretion amounts to a Chenery violation. As the D.C. Circuit explained in Prill v. NLRB, “[J]udicial deference is not accorded a decision . . . when the [agency] acts pursuant to an erroneous view of law and, as a consequence, fails to exercise the discretion dele­gated to it by Congress.” 301 755 F.2d 941, 942 (D.C. Cir. 1985). Perhaps, the court mused, the agency would have selected a different interpretation had it understood that the statute was amenable to that interpretation. 302 Id.

But three features of Prill cases distinguish them from a mine-run Chenery situation. First, as Professors Jerry Mashaw and Chris Walker have documented, agencies tend to take a pragmatic approach to statutory construction. 303 See Jerry L. Mashaw, Norms, Practices, and the Paradox of Deference: A Preliminary Inquiry into Agency Statutory Interpretation, 57 Admin. L. Rev. 501, 516 (2005); Christopher J. Walker, Inside Agency Statutory Interpretation, 67 Stan. L. Rev. 999, 1006 (2015). If the most straightforward reading of a statute best enables agencies to achieve their objectives, they won’t strain to identify ambiguity that might enable them to construe the statute differently. The very fact that an agency has read the statute in a particular way is thus evidence—not conclusive evidence but certainly probative—that it pre­fers the interpretation it adopted to the one that it did not adopt. That being the case, there’s reason to credit an agency’s protestation that it would have adopted the same interpretation had it formally wrestled with the statute’s ambiguity.

That’s especially so when Prill claims are presented in connection with agency enforcement proceedings. In such cases, the target of the enforcement action often argues that the agency failed to consider an alternative interpretation that would have precluded it from enforcing. 304 See PDK Labs. Inc. v. DEA, 438 F.3d 1184, 1197 (D.C. Cir. 2006). But when an enforcement action is available only under a particular interpretation of the statute, the very fact that the agency has exercised its discretion to bring the action is powerful evidence that the agency prefers the interpretation that allows such an action to be brought. When the agency represents as much to a court, it’s not just cheap talk or a convenient litigating position. The agency’s actions confirm that it believes the statute ought to be read to allow it to act in the manner that it did.

Second, an agency will nearly always be able to reinstate its previous decision on remand. After all, the problem in Prill cases is not that the agency’s interpretation is improper but that the agency hasn’t explained why it rejected an alternative. 305 Prill, 755 F.2d at 942. That’s different from the typical Chenery case, in which the agency may or may not be able to rehabilitate a flawed decision.

Third, deciding whether a statute is ambiguous enough to support a competing interpretation can be tricky. Take, for example, an interpret­ation that is linguistically plausible but clashes with statutory structure and context. An agency and a reviewing court might reasonably disagree about whether the statute forecloses the agency from adopting that interpretation. When that happens, the agency’s conclusion that the statute compels a particular interpretation may not be the product of “an erroneous view of law” 306 Id. so much as a difference of opinion over how to read the statute at hand. The question is not black and white, and rebuking an agency for not distinguishing finely enough between shades of gray won’t have much incentive value. To the contrary, agencies will read Prill to require them to entertain and reject arguments about statutory ambiguities that they don’t believe actually exist. Insulating a decision from Prill could become an exercise in boilerplate: “We do not believe the statute is amenable to an alternative reading, but even if it is, we reject it in favor of a reading that better advances our goals.”

But these three features of Prill cases are best shown by example. Start with the D.C. Circuit’s 2004 decision in PDK Laboratories Inc. v. U.S. DEA. 307 362 F.3d 786 (D.C. Cir. 2004). To crack down on domestic methamphetamine production, Congress armed the Drug Enforcement Administration (DEA) with the power to prevent the importation of certain “listed chemicals” into the United States if “the chemical may be diverted to the clandestine manu­facture of a controlled substance.” 308 21 U.S.C. § 971(c)(1) (2012). Pursuant to that authority, DEA blocked a pharmaceutical company from importing drugs containing ephedrine and pseudoephedrine—two “listed chemicals”—into the United States. 309 PDK Labs. Inc., 362 F.3d at 789–90. As DEA documented, the company’s drugs were regu­larly diverted to methamphetamine labs. 310 Id.

The D.C. Circuit found a Prill violation for two reasons. First, in the court’s view, the statute could be read to cover the importation of “listed chemicals” in their pure form but not drugs that happened to contain those chemicals. 311 Id. at 796. Second, the court thought the phrase “may be diverted” could be read to cover diversion during importation—hijack­ings at sea, say—but not diversion that occurred after the chemical had been imported. 312 Id. at 797. Because DEA failed to grapple with either ambiguity in concluding that it had the authority to suspend PDK’s shipments, the court vacated and remanded.

The ambiguities—if they existed at all—were marginal. From a functional perspective, why should DEA’s power to suspend an ephed­rine shipment depend on whether that ephedrine was mixed with something else? And why read the statute to restrict DEA’s authority to ocean hijackings? Rebuking DEA for failing to ventilate these unlikely possibilities seems excessive.

What’s more, and as then-Judge John Roberts explained in a sepa­rate opinion, DEA explained persuasively why it would be appropriate to read the statute to enable the agency to prohibit the shipment in question. Far from a “confession of powerlessness” 313 See id. at 808 (Roberts, J., concurring in part and concurring in the judgment). —an agency admis­sion that its hands were tied because the statute could mean only one thing—DEA’s explan­ation amply demonstrated its conviction about how the statute should be read. Judge Roberts continued:

DEA wanted to suspend PDK’s imports. We know this because it did suspend the imports. If it did not want to, the agency had discretion to choose otherwise. Given its manifest desire to suspend PDK’s imports, it is fanciful to suggest that the agency—when presented on remand with an opportunity to choose “any permissible construction” of [the statute]—will choose an interpretation that diminishes its discretion to an extent that places PDK’s imports beyond its reach. 314 Id.

Quoting an opinion from Judge Friendly, Judge Roberts closed with the observation that “‘Chenery does not require that we convert judicial review of agency action into a ping-pong game.’” 315 Id. at 809 (quoting Time, Inc. v. U.S. Postal Serv., 667 F.2d 329, 335 (2d Cir. 1981) (Friendly, J.)).

But ping-pong is what the court got. On remand, DEA again suspended PDK’s shipments, this time explicitly rejecting alternative readings of the statute. When the case went back up to the D.C. Circuit, the court, having already resolved that the statute was ambiguous, brushed aside the statutory challenge. 316 See PDK Labs. Inc. v. U.S. DEA, 438 F.3d 1184, 1197 (D.C. Cir. 2006). What exactly did this round trip accomplish?

Or consider another D.C. Circuit case, this one involving the Postal Regulatory Commission’s denial of a request from the United States Postal Service (USPS) to raise postage rates above a presumptive statutory threshold. 317 See U.S. Postal Serv. v. Postal Regulatory Comm’n, 640 F.3d 1263 (D.C. Cir. 2011). In the Commission’s judgment, the size of USPS’s proposed postage increase wasn’t tethered closely enough to the “extraordinary or exceptional circumstances”—here, the recession that began in late 2007—that, by statute, were necessary to justify the higher rates. 318 Id. at 1264 (internal quotation marks omitted) (quoting 39 U.S.C. § 3622(d)(1)(E) (2012)). The postage increase was instead an “‘attempt to address long-term structural problems not caused by the recent recession.’” 319 Id. at 1268 (quoting Order Denying Request for Exigent Rate Adjustments, Docket No. R2010-4 (Postal Regulatory Comm’n Sept. 30, 2010), []).

The court agreed with the Commission that it could approve the rate increase only in response to extraordinary circumstances like the Great Recession. But the court thought that the statute was ambiguous as to whether the size of the rate increase had to be tethered to those circumstances. Under the statute, the court mused, the Commission could perhaps approve rate increases that were larger than necessary to cope with the particular exigent circumstance. The court therefore vacated and remanded so that the Commission could “exercise its discretion to construe the ambiguous language.” 320 Id. at 1264.

In so doing, the court ignored that the Commission, when it rejected the proposed rate increase, had offered an extensive, twenty-two-page discussion of the statute’s text, background, and purpose. 321 Order Denying Request for Exigent Rate Adjustments, Docket No. R2010-4 (Postal Regulatory Comm’n Sept. 30, 2010), []. In painstak­ing detail, the Commission explained that the statute was meant to put financial pressure on USPS to operate “in a more business-like fashion.” 322 Id. at 7. Allowing USPS to increase postage rates beyond what was necessary to cover emergency shortfalls, the Commission reasoned, would work at cross purposes to that objective.

On remand, the Commission did exactly what might be expected given its earlier discussion: It reinstated its prior order. This time, it explained that it was resolving the statutory ambiguity to require a nexus between the amount of a rate increase and the exigent circum­stances. 323 Order Resolving Issues on Remand, Docket No. R2010-4R (Postal Regulatory Comm’n Sept. 20, 2011), []. The Commission justified its interpretation for the same reasons that it offered in its prior order—indeed, its analysis relied almost exclusively on citations to that prior order. 324 See id. at 34–35. On review, the D.C. Circuit upheld the interpretation. 325 See All. of Nonprofit Mailers v. Postal Regulatory Comm’n, 790 F.3d 186, 191, 193–94 (D.C. Cir. 2015).

As in PDK, the ambiguity that the court identified was far from obvious. And, as in PDK, the Agency explained why it believed it was good policy to read the statute in the manner that it did. What sense did it make to punish the Commission for failing to discuss and reject an alternative, halfway-plausible interpretation? Yet such needless punish­ment is the norm in Prill cases, not the exception. Examples abound. 326 Here are six. In Noble Energy, Inc. v. Salazar, the D.C. Circuit invalidated an agency’s discretionary decision to require an oil well to be permanently plugged because the agency failed to explain at the time of decision that its authority stemmed from an interpretation of an existing regulation. 671 F.3d 1241, 1245–46 (D.C. Cir. 2012). In Menkes v. Department of Homeland Security, the court vacated an agency action that lacked “a forthright agency interpretation of the statute,” notwithstanding the “implication” that the agency interpreted it a particular way. 486 F.3d 1307, 1313–14 (D.C. Cir. 2007). In Arizona v. Thompson, the court held that the Department of Health and Human Services (HHS) improperly read a statute to preclude states from using Temporary Assistance for Needy Families (TANF) funding to cover the joint administrative costs of running the TANF, Medicaid, and the Food Stamp programs. 281 F.3d 248, 259 (D.C. Cir. 2002). In Transitional Hospitals Corp. v. Shalala, the court held that HHS improperly believed it was statutorily obligated to refuse certification to new long-term-care hospitals for a six-month period. 222 F.3d 1019, 1029 (D.C. Cir. 2000). In Phillips Petroleum Co. v. FERC, the court invalidated a FERC order on the ground that the Agency wrongly interpreted its statute to compel a particular outcome. 792 F.2d 1165, 1171–72 (D.C. Cir. 1986). And in Ewing v. NLRB, the Second Circuit reversed the NLRB on the grounds that it erroneously believed itself to be statutorily bound to a view about the meaning of “concerted activities.” 768 F.2d 51, 54 (2d Cir. 1985).
In each of these six cases, the agency reached the same result on remand. Four of the cases were appealed; the courts upheld the agency decision in each. See Menkes v. U.S. Dep’t of Homeland Sec., 637 F.3d 319, 333 (D.C. Cir. 2011); Midwest Energy v. FERC, 870 F.2d 660, 661–62 (D.C. Cir. 1989); Ewing v. NLRB, 861 F.2d 353, 359, 362 (2d Cir. 1988); Noble Energy, Inc. v. Jewell, 110 F. Supp. 3d 5, 15–16 (D.D.C. 2015); 73 Fed. Reg. 42,718, 42,719 (July 23, 2008) (codified at 45 C.F.R. pt. 263) (reinstating the prior HHS decision challenged in Arizona v. Thompson); 66 Fed. Reg. 39,828, 39,917 (Aug. 1, 2001) (codified in scattered sections of 42 C.F.R.) (reinstating certification procedure for long-term-care hospitals challenged in Transitional Hospitals Corp. v. Shalala).
Cases in which an agency changes its mind in response to Prill remands are rare, and the circumstances tend to be unusual. In Alarm Industry Communications Committee v. FCC, for example, the court opined that the Agency’s original interpretation was “senseless,” 131 F.3d 1066, 1068 (D.C. Cir. 1997), which may explain why the Agency changed its view on remand. Ameritech Corp., 13 FCC Rcd. 19,046, 19,047 (1998). And in GTE Service Corp. v. FCC, 224 F.3d 768, 775–76 (D.C. Cir. 2000), the D.C. Circuit issued a Prill remand in another FCC case. The agency eventually changed its mind, but only after eleven more years had elapsed. See S. Seas Broad., Inc., 26 FCC Rcd. 4164, 4167–69 (2011).

To what end? Rigid adherence to Prill won’t make agencies better at spotting latent ambiguities. In all likelihood, agencies will carry on much as they would in the absence of Prill, deaf to its marginal incentive effects. To the extent agencies do pay attention, their decisions will become bloated with boilerplate legal analysis. Chenery does not demand—and should not be read to demand—that kind of waste. 327 Professors Daniel Hemel and Aaron Nielson have developed an ingenious defense of Prill, one premised on a dark picture of agencies that are riven by internal factions and often seek to shirk public responsibility. See Daniel J. Hemel & Aaron L. Nielson, Chevron Step One-and-a-Half, 84 U. Chi. L. Rev. (forthcoming 2017) (manuscript at 34–35) (on file with the Columbia Law Review). Even Hemel and Nielson acknowledge, however, that, “in most cases, an agency that says in the first instance that ‘the statute compels X’ will, following a [Prill] remand, adopt position X as an exercise of discretion.” Id. (manuscript at 42). They thus conclude that, “given the high costs of vacatur, a context-dependent approach [to remedy] makes the most sense here—with vacatur reserved for cases that cannot be chalked up to innocent agency error.” Id. (manuscript at 54). If the category of cases involving noninnocent errors is a small one—and the frequency with which agencies reinstate their decisions suggests that it is—Hemel and Nielson’s approach to Prill errors may differ little in practice from the one advanced here. When an agency has adopted a reasonable construction of a statute, when its actions indicate that it prefers that interpretation to the alternative, and when it represents to a reviewing court that it would stick to that interpretation even if the statute could be read differently, the rule of prejudicial error suggests that the interpretation should stand.

B. Harmless Error

The excessiveness of Prill-type cases notwithstanding, the lower courts have exercised more remedial flexibility than Chenery appears to display. In an echo of Judge Friendly, every one of the federal courts of appeals has made a practice of upholding unsound agency decisions when they are confident that the agency would reach the same decision on remand. 328 See Grossmont Hosp. Corp. v. Burwell, 797 F.3d 1079, 1086 (D.C. Cir. 2015); Gillum v. Comm’r, 676 F.3d 633, 646 (8th Cir. 2012); Li Hua Yuan v. Attorney Gen., 642 F.3d 420, 427 (3d Cir. 2011); Japarkulova v. Holder, 615 F.3d 696, 701 (6th Cir. 2010); Parker v. Astrue, 597 F.3d 920, 924 (7th Cir. 2010); Nadal–Ginard v. Holder, 558 F.3d 61, 70 n.7 (1st Cir. 2009); Cao He Lin v. U.S. Dep’t of Justice, 428 F.3d 391, 401 (2d Cir. 2005); Ngarurih v. Ashcroft, 371 F.3d 182, 190 n.8 (4th Cir. 2004); In re Watts, 354 F.3d 1362, 1369 (Fed. Cir. 2004); Beltran-Resendez v. INS, 207 F.3d 284, 287 (5th Cir. 2000); Nazaraghaie v. INS, 102 F.3d 460, 464 (10th Cir. 1996); Curry v. Sullivan, 925 F.2d 1127, 1131 (9th Cir. 1991); Lucas v. Sullivan, 918 F.2d 1567, 1574 (11th Cir. 1990). These harmlessness cases are especially prominent in the immigration and social security disability context in which front-line adjudicators make serious errors but the record, as a whole, strongly suggests that the claimant could not prevail if given a second bite at the apple.

The courts sometimes acknowledge the tension between Chenery’s apparent absolutism—“an administrative order cannot be upheld unless the grounds upon which the agency acted in exercising its powers were those upon which its action can be sustained” 329 SEC v. Chenery Corp., 318 U.S. 80, 95 (1943). —and a decision to characterize an agency’s error as harmless. 330 See, e.g., Illinois v. Interstate Commerce Comm’n, 722 F.2d 1341, 1348–49 (7th Cir. 1983). When courts do, they usually assert that an evaluation of harmlessness is “an exception to the Chenery doctrine” 331 Parker, 597 F.3d at 924. without offering much in the way of analysis. The most sophisticated discussions tend to bolster the court’s reasoning with a citation to the Supreme Court’s 1964 decision in Massachusetts Trustees of Eastern Gas & Fuel Associates v. United States. 332 377 U.S. 235 (1964). For an example of such a discussion, see, e.g., Cao He Lin, 428 F.3d at 401–02. There, the Court ex­plained the point of cases like Chenery:

[They] are aimed at assuring that initial administrative determinations are made with relevant criteria in mind and in a proper procedural manner; when a mistake of the administrative body is one that clearly had no bearing on the procedure used or the substance of decision reached, as in this instance (assuming there was such a mistake), the sought extension of the cases cited would not advance the purpose they were intended to serve. 333 Mass. Trs., 377 U.S. at 248. The lower courts will also sometimes refer to Justice Abraham Fortas’s opinion for four Justices in NLRB v. Wyman-Gordon, Co., in which he wrote that “Chenery does not require that we convert judicial review of agency action into a ping-pong game . . . . There is not the slightest uncertainty as to the outcome of a proceeding before the Board, whether the Board acted through a rule or an order. It would be meaningless to remand.” 394 U.S. 759, 766 n.6 (1969); see also Cao He Lin, 428 F.3d at 401 (citing Wyman-Gordon, 394 U.S. at 766 n.6).

Writing a few years later, Judge Friendly called Massachusetts Trustees a “true indentation of Chenery” but “an altogether sound one.” 334 Friendly, supra note 28, at 210–11. A remand, he wrote, “is necessary only when the reviewing court concludes that there is a significant chance that but for the error the agency might have reached a different decision.” 335 Id. at 211 (footnote omitted).

The lesson of Massachusetts Trustees has been lost to administrative law. The case isn’t cited in the leading casebooks 336 See Stephen G. Breyer et al., Administrative Law and Regulatory Policy (7th ed. 2011); Mashaw et al., supra note 104. or treatises, 337 See, e.g., 1 Richard J. Pierce, Administrative Law Treatise 559–60 (5th ed. 2010) [hereinafter Pierce, Administrative Law]. and it’s rarely mentioned in the academic literature. In part because of that omission, most administrative law scholars would be surprised to learn that the lower courts regularly hold Chenery errors harmless. The oversight of the rule of prejudicial error persists notwithstanding the Supreme Court’s 2009 decision in Shinseki v. Sanders, which confirmed that the rule “is intended to ‘su[m] up in succinct fashion the “harmless error” rule applied by the courts in the review of lower court decisions as well as of administrative bodies.’” 338 556 U.S. 396, 406 (2009) (alteration in original) (quoting AG’s Manual, supra note 35, at 110). Sanders has largely been ignored, 339 See, e.g., Breyer et al., supra note 336 (including no mention of the case); Mashaw et al., supra note 104 (same); Pierce, supra note 337, at 559–60 (including only a brief discussion of Shinseki v. Sanders). One rare exception is a 2010 student note. See Craig Smith, Note, Taking “Due Account” of the APA’s Prejudicial-Error Rule, 96 Va. L. Rev. 1727, 1764 (2010) (noting harmless error “appears to be an afterthought in many opinions”). perhaps because the Court didn’t so much as cite Chenery, much less resolve the tension between its categorical rule and the harmlessness standard. The same failure to grapple with the tension was on display in the Supreme Court’s decision in National Ass’n of Home Builders v. Defenders of Wildlife, 340 551 U.S. 644 (2007). which excused as harmless a minor error in an EPA order without mentioning Chenery 341 Id. —even though the dissent charged the Court with “ignor[ing] this hoary principle of administrative law.” 342 Id. at 684 (Stevens, J., dissenting).

Instead of passing unnoticed, National Ass’n of Home Builders and Sanders should have sparked a conversation about the proper role of remedial discretion in administrative law. Can Chenery be squared with harmlessness review? If so, how? Should reviewing courts deploy the same harmlessness standard in reviewing agency decisions as they do in reviewing trial court decisions? Or does the standard need to be adapted for administrative law? For decades now, the lower courts have put into action their tacit belief that, in a subset of cases, the costs of vacating flawed agency actions outweigh the systemic benefits. But the lower courts have done so haphazardly, with little guidance on how to recon­cile the practice with Chenery.

Yet some patterns emerge from the case law. Harmless error tends to crop up in connection with agency orders that result from relatively formal adjudicatory processes. Errors are most commonly held harmless in two circumstances: either when the agency has made a factual mistake of peripheral significance 343 See, e.g., PDK Labs. Inc. v. DEA, 438 F.3d 1184, 1196 (D.C. Cir. 2006) (holding that minor factual inaccuracies were de minimis); Curry v. Sullivan, 925 F.2d 1127, 1131 (9th Cir. 1990) (disregarding as irrelevant mistakes about a claimant’s age and her GED status). or (more controversially) when the agency’s error is serious but the evidence in the record so strongly supports the result that the court is confident the agency would reach the same decision on remand. 344 See, e.g., In re Watts, 354 F.3d 1362, 1370–71 (Fed. Cir. 2004) (refusing to remand to the Board of Patent Appeals when the Agency’s rejection of certain patents implied that it would have similarly denied two others); Nazaraghaie v. INS, 102 F.3d 460, 464 (10th Cir. 1996) (finding error in the rejection of an alien’s claim that he was subject to persecution but refusing to remand on the alternative ground, not relied on by the Board of Immigration Appeals, of changed country conditions). These aren’t the only circumstances in which the rule of harmless error will salvage an arbitrary agency decision, 345 See, e.g., Fed. Express Corp. v. Mineta, 373 F.3d 112, 120 (D.C. Cir. 2004) (holding a failure to grapple with a legal argument in a rule was harmless). but they are the most common. Perhaps unsurprisingly, then, harmless error looms largest in the immigration and disability contexts. The quality of the underlying decisions is sometimes poor and the weaknesses in the claimants’ cases are often apparent from a well-developed record.

Consider an immigration example. An alien from Kyrgyzstan sought asylum on the ground that she was persecuted in her home country. 346 See Japarkulova v. Holder, 615 F.3d 696, 698 (6th Cir. 2010). She testified that she had discovered that the wife of the then-president was mishandling a private foundation’s funds. 347 Id. When she tried to expose the irregularities, the head of the national security apparatus told her “that the government would arrange a fatal ‘accident’ for [her] if she did not desist.” 348 Id. The immigration judge concluded that the death threat did not amount to persecution because it was just verbal harass­ment. 349 Id. at 700. The Sixth Circuit found that conclusion to be arbitrary and acknowledged that “[i]n the ordinary case” Chenery would require a re­mand. 350 Id. at 701. Here, however, the court found the error harmless. 351 Id. The threat was not “immediate and menacing” enough, standing alone, to count as persecution. 352 Id. The alien didn’t flee Kyrgyzstan until eight years after the threat was made, which “lessens the severity of the threat.” 353 Id. And she kept up her antigovernment activities, suggesting the persecu­tion wasn’t all that substantial. 354 Id. “Under these circumstances, we see no reasonable prospect that ‘remand might lead to a different result.’” 355 Id. (quoting Shkabari v. Gonzales, 427 F.3d 324, 328 (6th Cir. 2005)).

Or take a similar disability case. 356 See Batson v. Comm’r, Soc. Sec. Admin., 359 F.3d 1190 (9th Cir. 2004). After losing his job, a claimant filed for disability benefits on account of a degenerative disease of the neck. 357 Id. at 1193. An administrative law judge (ALJ) discredited the claimant’s testimony about his pain and the extent of his functional limitations in part because the claimant said that he watched between six and ten hours of television a day, which suggested to the ALJ that he could sit for extended periods. 358 Id. at 1197. The Ninth Circuit held that the ALJ erred in assuming, without evidence, that the claimant sat while he watched television. “[I]t is possible [he] at times watched television while standing or reclining, or that he changed positions from time to time.” 359 Id. The court nonetheless found the error harmless given other evidence discussed in the ALJ’s opinion that raised questions about the claimant’s credibility. 360 Id. The court didn’t mention Chenery.

It’s possible to quarrel with these two decisions, each of which provoked a dissent. 361 See Japarkulova v. Holder, 615 F.3d 696, 703–04 (6th Cir. 2010) (Martin, J., dissenting); Batson, 359 F.3d at 1198 (Graber, J., dissenting). But their impulse to salvage the agencies’ decisions can’t be dismissed out of hand. Sure, it’s possible that agencies might become careless if courts too readily excuse their mistakes, diminishing faith in administrative adjudication and increasing the risk of error. It’s thus possible that vacating every defective agency opinion will, over time, increase the quality of decisionmaking. But the reverse is also possible. Front-line adjudicators may not be especially sensitive to the risk that an appellate court might someday vacate their decisions. After all, agencies are all but precluded from rewarding or penalizing ALJs for their job performance. 362 See 5 U.S.C. § 4301(2)(D) (2012) (excluding ALJs from the definition of “employees” subject to an agency’s performance appraisal system); 5 C.F.R. § 930.206(a) (2016) (“An agency may not rate the job performance of an administrative law judge.”). Especially in the context of mass adjudication, agency adjudicators may focus more on the day-to-day imperatives of the job, including the need to process large numbers of cases with an acceptable degree of rigor. To the extent that adjudicators’ mistakes are a function of inadequate resources, further straining those resources could slow the adjudicatory process and generate mistakes. A refusal to hold errors harmless could thus be counterproductive.

The challenge isn’t in identifying these tradeoffs but in determining their magnitude. That’s hard, but courts should at least be trying—not ignoring the tradeoffs on the false assumption that Chenery commands but one approach to administrative error. To that end, some of the circuit courts have attempted to supply guardrails to confine the rule of prejudicial error. The Second Circuit, for example, has held that an error cannot be harmless unless it is tangential to the result or “overwhelming evidence supporting the administrative adjudicator’s findings makes it clear that the same decision would have been reached in the absence of the errors.” 363 Cao He Lin v. U.S. Dep’t of Justice, 428 F.3d 391, 401 (2d Cir. 2005). The Seventh, Ninth, and D.C. Circuits have offered variations on that theme. 364 See Grossmont Hosp. Corp. v. Burwell, 797 F.3d 1079, 1086 (D.C. Cir. 2015) (holding that a court can affirm an agency decision for reasons that don’t appear in that decision only “when there is not the slightest uncertainty as to the outcome of a proceeding on remand” (internal quotation marks omitted) (quoting Manin v. Nat’l Transp. Safety Bd., 627 F.3d 1239, 1243 n.1 (D.C. Cir. 2011))); Spiva v. Astrue, 628 F.3d 346, 353 (7th Cir. 2010) (“If it is predictable with great confidence that the agency will reinstate its decision on remand because the decision is overwhelmingly supported by the record though the agency’s original opinion failed to marshal that support, then remanding is a waste of time.”); Stout v. Comm’r, Soc. Sec. Admin., 454 F.3d 1050, 1055–56 (9th Cir. 2006) (requiring reversal in disability cases whenever an ALJ fails to discuss competent testimony supporting the claimant’s position unless the court “can confidently conclude that no reasonable ALJ, when fully crediting the testimony, could have reached a different disability determination”).

Maybe that’s the right approach, maybe it’s not. Maybe the answer is contingent on the general diligence (or negligence) of the agency, the interest at stake, or the type of error at issue. 365 Cf. Spiva, 628 F.3d at 353 (criticizing the government for trying “to dissolve the Chenery doctrine in an acid of harmless error” and worrying about the quality of ALJ decisions in disability cases). But that discussion isn’t happening, even as the lower courts keep reminding us that Chenery in the trenches is not as absolute as Chenery in the casebooks. Their practice suggests, too, that Judge Friendly’s views about Chenery’s limits have made more inroads in the courts than is commonly appreciated.

C. Remand Without Vacatur

Although the practice of holding errors harmless in agency adjudication has been largely overlooked, a different doctrinal innova­tion—remand without vacatur—has received considerable attention. 366 See, e.g., Tatham, supra note 60; Levin, Vacation, supra note 43; Daugirdas, supra note 23. In the late 1980s, the D.C. Circuit began to leave agency rules in place, notwithstanding the presence of error, when the costs of vacating appeared disproportionate to the agency’s mistake. 367 Pierce, Administrative Law, supra note 337, at 686–88 (describing the early practice). That practice of “remand without vacatur” was formalized in Allied-Signal, Inc. v. U.S. Nuclear Regulatory Commission, when the D.C. Circuit held that “[t]he decision whether to vacate depends on ‘the seriousness of the order’s deficiencies (and thus the extent of doubt whether the agency chose correctly) and the disruptive consequences of an interim change that may itself be changed.’” 368 988 F.2d 146, 150–51 (D.C. Cir. 1993) (quoting Int’l Union, United Mine Workers v. Fed. Mine Safety & Health Admin., 920 F.2d 960, 967 (D.C. Cir. 1990)). Since 2000, the D.C. Circuit has left agency rules intact and remanded them to agencies at the rate of about three times every year. 369 See Tatham, supra note 60, at 22. Other circuits—including the First, Third, Fifth, Eighth, Ninth, Tenth, and Federal—have followed the D.C. Circuit’s lead. 370 See id. at 27.

But the practice is controversial. Two judges on the D.C. Circuit have publicly registered concerns with its legality, arguing that the APA, which says that a reviewing court “shall . . . set aside agency action[s] . . . found to be . . . arbitrary, capricious, [or] an abuse of discretion,” 371 5 U.S.C. § 706(2)(A) (2012). strips courts of the remedial authority to decline to vacate. 372 See Milk Train, Inc. v. Veneman, 310 F.3d 747, 757 (D.C. Cir. 2002) (Sentelle, J., dissenting); Checkosky v. SEC, 23 F.3d 452, 491 (D.C. Cir. 1994) (Randolph, J., concurring); see also Brian S. Prestes, Remanding Without Vacating Agency Action, 32 Seton Hall L. Rev. 108, 109 (2001). Judges and commentators sometimes pair the formal statutory argument with a functional concern: Agencies have little or no incentive to respond to an order to cure the error that the court has identified if that rule remains intact on remand. 373 See Nat. Res. Def. Council v. EPA, 489 F.3d 1250, 1262–64 (D.C. Cir. 2007) (Randolph, J., concurring) (“A remand-only disposition is, in effect, an indefinite stay of the effectiveness of the court’s decision and agencies naturally treat it as such.”). That concern has been borne out in practice. As Professor Kristina Daugirdas has documented, agencies sit on their hands in response to remands, taking as long as a decade (or more) to fix inadequacies that the reviewing court identified. 374 See Daugirdas, supra note 23, at 301–05. On occasion, agency dilatoriness has so frustrated the D.C. Circuit that it has entered writs of mandamus to demand prompt action from agencies. 375 See, e.g., In re People’s Mojahedin Org. of Iran, 680 F.3d 832, 838 (D.C. Cir. 2012); In re Core Commc’ns, 531 F.3d 849, 861 (D.C. Cir. 2008).

Because the Supreme Court has declined to address its legality, remand without vacatur has become a routine part of administrative law even as it remains legally vulnerable. There’s some reason to think the Court might not smile on the practice. In its recent decision in Perez v. Mortgage Bankers Ass’n, the Court took the D.C. Circuit to task for requiring an agency to conduct notice-and-comment rulemaking before deviating from a prior interpretation of its own regulation. 376 135 S. Ct. 1199, 1206 (2015) (abrogating Paralyzed Veterans of Am. v. D.C. Arena, 117 F.3d 579 (D.C. Cir. 1997)). That doctrinal innovation, the Court held, ran counter to the “clear text” of the APA. 377 Id. It “may be wise policy. Or it may not. Regardless, imposing such an obligation is the responsibility of Congress or the administrative agencies, not the courts.” 378 Id. at 1207. Perez may presage a renewed era of skepticism about other doctrinal innovations that likewise have no obvious home in the APA—including, perhaps, remand without vacatur. 379 I’m indebted to Adam White for this point. See Email from Adam White, Research Fellow, Hoover Inst., to Nicholas Bagley, Professor, Univ. of Mich. Law Sch. (Mar. 17, 2015, 9:41 pm) (on file with the Columbia Law Review).

Embattled though it may be, remand without vacatur is the only enclave of administrative law in which remedy has been taken seriously. And the remedy has grown organically as a response to the courts’ lived experience of adjudicating cases in which the defects in an agency rule, though real, were not so serious as to warrant the disruption that vacating can entail. That in itself is suggestive. It’s unlikely that a practice that salvages rules with minor defects exhausts the field of remedial discretion in administrative law.

Even with respect to remand without vacatur, however, discomfort with remedial discretion has warped the discussion. Strictly as a matter of statutory construction, the controversy is inexplicable. Although the APA says that a reviewing court “shall . . . hold unlawful and set aside” arbi­trary agency action, 380 5 U.S.C. § 706(2)(A)(2012). it also says that “due account shall be taken of prejudicial error.” 381 Id. § 706. So far as the APA is concerned, reviewing courts are authorized to hold agency errors harmless, much as they can hold trial errors harmless. There’s nothing to the argument that the APA, by its terms, strips courts of the authority to leave procedurally defective agency rules intact. 382 Professor Levin has developed this argument at length. See Levin, Vacation, supra note 43, at 309–15. The argument is taken seriously, I think, only because the norm of remedial purity is so embedded in administrative law. It feels right to read remedial inflexibility into the APA, even if that’s not what the APA says.

To see what’s truly anomalous about remand without vacatur, it’s important to tease apart its two elements: a judgment sustaining the agency action and an order requiring the agency to correct its mistake. Think about how odd a similar remedy would be in the context of civil litigation. When an appellate court concludes that a trial court has erred, it either vacates the judgment or holds the error harmless. 383 See Fed. R. Civ. P. 61. It doesn’t order the trial court to fix its mistakes even as its judgment takes effect. Courts have better things to do with their time.

Yet it’s taken for granted in the case law and the commentary that it’s proper to insist that agencies rectify errors—and that it’s a problem when they don’t. 384 See In re Core Commc’ns, 531 F.3d 849, 862–63 (D.C. Cir. 2008) (Griffith, J., concurring) (encouraging future courts “to consider the alternative to open-ended remand without vacatur”); Daugirdas, supra note 23, at 301–05. Why? I know of no instance in which an agency, on remand, has substantially revised an action that the courts have left in place. And forcing an agency to supply better reasons for doing what it’s already done is unlikely to mollify those who brought the initial chal­lenge or to inspire agencies to craft higher-quality rules in the first place.

An agency response to a remand order is often little more than a formal ritual signifying obeisance to the reviewing court’s authority. Giving low priority to that kind of ritual is completely reasonable—and not only from the agency’s perspective. Fixing minor mistakes is no trivial matter, especially when the agency has to run the arduous rulemaking gantlet for a second time after staffers have been reassigned to other projects. Insisting on a response will deplete agency resources and could distract from matters of greater urgency. When an agency dallies in responding to a remand order, that might be because it thinks that it should put its scarce resources to more productive ends. What the courts see as laziness or disdain for judicial authority may be nothing of the kind.

As such, the problem with remand without vacatur may not be that agencies don’t respond expeditiously to remand orders. The problem may be that they’re ordered to respond at all. The prominence of remand without vacatur has displaced a forthright discussion of the possibility that some errors should be excused as harmless. 385 The courts will from time to time hold trivial errors in agency rulemaking harmless. See, e.g., Inv. Co. Inst. v. CFTC, 720 F.3d 370, 375 (D.C. Cir. 2013); Cape Code Hosp. v. Sebelius, 630 F.3d 203, 212 (D.C. Cir. 2011). They usually don’t, however, consider the possibility of holding more substantial errors harmless, even when there’s no realistic chance that the agency will revisit its decision.

Consider Michigan v. EPA, in which a fractured Supreme Court invalidated EPA’s determination that regulating coal- and oil-fired power plants was both “appropriate and necessary” within the meaning of the Clean Air Act. 386 135 S. Ct. 2699, 2701 (2015). In the Court’s view, EPA misconstrued its statutory authority when it purported to make that threshold determination with­out reference to the costs of regulation. 387 Id. at 2712.  Justice Elena Kagan disagreed: “Over more than a decade,” she wrote, “EPA took costs into account at multiple stages and through multiple means as it set emissions limits for power plants.” 388 Id. at 2714 (Kagan, J., dissenting). For Justice Kagan, EPA could reasonably construe the open-ended direction—regulate when “appropriate”—to allow it to ignore costs at the threshold so long as it considered costs at later stages. 389 Id. at 2714–15. To support the reasonableness of EPA’s interpretation, she recounted at length the myriad ways that the agency infused cost into its decisionmak­ing process. 390 Id. at 2718–22.

Notice, though, that Justice Kagan’s Chevron defense could have done double duty as an explanation of why EPA’s error was harmless. She hinted as much in a retort to the Court’s refusal to accept EPA’s after-the-fact consideration of costs. “Of course a court may not uphold agency action on grounds different from those the agency gave,” she wrote, citing Chenery. 391 Id. at 2725. “But equally, a court may not strike down agency action without considering the reasons the agency gave.” 392 Id. Those reasons included a careful weighing of costs and benefits—suggesting that the agency’s actions washed clean any legal error at the threshold.

On remand, the D.C. Circuit picked up on that hint of harmlessness and ordered the parties to file motions to govern future proceedings. 393 See White Stallion Energy Ctr., LLC v. EPA, No. 12-1100 (D.C. Cir. Aug. 11, 2015) (order to file motions). Even then, EPA had no doctrinal toehold to argue that the court should just hold its error harmless. To the contrary, the agency had to argue for a remand, saying that “EPA believes, based on the cost data that is already in the record, that there is a ‘serious possibility’ that EPA will reaffirm the finding.” 394 Respondent’s Motion to Govern Future Proceedings at 10, White Stallion Energy Ctr., LLC v. EPA, No. 12-1100 (D.C. Cir. Sept. 24, 2015). There’s more than a measure of insincerity here: EPA knew full well it would reaffirm its findin. Is it healthy for administrative law to force agencies into this kind of doublespeak? What purpose does it serve to require EPA on remand to pretend to think about costs in deciding whether to regulate categories of power plants that it has already decided it is “appropriate” to regulate?

Among commentators, too, the rise of remand without vacatur has crowded out discussion of harmless error. Professor Daniel Rodriguez, for example, has argued that remand without vacatur may lead courts to increase the stringency of hard-look review precisely because they need not vacate arbitrary agency rules. 395 See Daniel B. Rodriguez, Of Gift Horses and Great Expectations: Remands Without Vacatur in Administrative Law, 36 Ariz. St. L.J. 599, 617–24 (2004). As a result, he worries that courts will force agencies to engage in senseless exercises of reason giving when, if remand weren’t an option, the courts might have ruled in the agencies’ favor. 396 Id. But Professor Rodriguez’s argument is built on the false premise that identifying an error requires the courts either to vacate or to remand with instructions to cure the error. Courts have a third option: They can hold the error harmless. If they did so more regularly, Professor Rodriguez’s concern would go away. More errors might be identified but fewer would require action from the agency.

Remand without vacatur need not be discarded altogether. Some­times there’s value in asking an agency to clarify an issue that may recur. And sometimes there’s reason to think that the agency might reconsider. In particular, remanding might be appropriate in cases brought by proregulatory interest groups that want stricter standards for private industry. Such groups are likely to prefer a lax rule to no rule at all, even if the errors they identify are so serious that the agency is unlikely to be able to rehabilitate the rule in its current form. Insisting that the agency respond to a remand order under those circumstances is probably appro­priate. In North Carolina v. EPA, for example, the D.C. Circuit declined to vacate an EPA rule that had been held to be contrary to law. 397 550 F.3d 1176 (D.C. Cir. 2008). Leaving the rule intact, the court reasoned, “would at least temporarily preserve the environmental values covered” by the program, even though the court’s judgment meant that the Agency had no choice but to revise the rule. 398 Id. at 1178. Alternatively, a reviewing court could invalidate the agency rule but withhold its mandate for a fixed period to give the agency an opportunity to revisit the rule. See, e.g., Friends of the Earth v. EPA, 446 F.3d 140, 148 (D.C. Cir. 2006) (citing Daugirdas, supra note 23, at 307 & n.141).

At a minimum, though, reviewing courts should be more deliberate about the choice of whether to require agencies to rectify errors. There is often nothing to be gained, and something to be lost, in assigning make-work.

IV. Rules, Standards, and Remedial Discretion

In both the courts and the casebooks, administrative law admits few exceptions to its commitment to invalidating procedurally defective or substantively arbitrary agency action. Letting agencies off the hook for their errors would risk puncturing what has become a central tenet of American administrative law: that agencies cannot be trusted to responsibly wield their vast discretionary powers without close judicial supervision. 399 See Norton E. Long, Bureaucracy and Constitutionalism, 46 Am. Pol. Sci. Rev. 808, 808 (1952) (“Because bureaucracy is often viewed as tainted with an ineradicable lust for power, it is alleged that, like fire, it needs constant control to prevent its erupting from beneficient servitude into dangerous and tyrannical mastery.”). On this picture, only courts stand between the American public and a federal bureaucracy that would, if given the chance, disregard legal constraints and run roughshod over individual liberties. Spare the rod and spoil the agency.

This court-centric vision of administrative law allows courts to opine that the APA’s procedural requirements “obviously would be eviscerated” if the courts relented to an agency’s request for remedial flexibility. 400 Sugar Cane Growers Coop. v. Veneman, 289 F.3d 89, 96 (D.C. Cir. 2002); see also United States v. Reynolds, 710 F.3d 498, 509 (3d Cir. 2013) (reaching a similar conclusion). That same vision tacitly underwrites much of the scholarship in admini­strative law. Professors Hickman and Thomson, for example, are unusually attentive to the benefits of remedial flexibility. Even they, however, argue for a “strong presumption” against allowing post-promulgation notice and comment to make up for a rule’s procedural deficiencies. 401 Hickman & Thomson, supra note 210, at 35. “If agencies see no disadvantage to relying on post-promulgation notice and comment,” they write, “they will more frequently disregard § 553’s pre-promulgation requirement and rely on § 706’s harmless error doctrine to sustain rules against procedural objections.” 402 Id.

In this, the courts and commentators make a common mistake. They identify an unfortunate incentive effect that doctrine creates for agencies and then call for eliminating that incentive—without defending the position that the incentive is large enough to warrant the costs of its elimination. It’s the equivalent of worrying about a vector’s direction without noting its magnitude. Professors Cass Sunstein and Adrian Vermeule characterize this as a sign fallacy, 403 See Cass R. Sunstein & Adrian Vermeule, The Unbearable Rightness of Auer, 84 U. Chi. L. Rev. (forthcoming 2017) (manuscript at 3–4) (on file with the Columbia Law Review). and it’s a pervasive problem in administrative law. The incessant worry that modest changes in judicial doctrine will exert large effects on agency behavior may also reflect what Professor Daniel Kahneman calls the focusing illusion, in which people focus disproportionately on a salient aspect of a large and complex problem: “Nothing in life is as important as you think it is, while you are thinking about it.” 404 Daniel Kahneman, Thinking Fast and Slow 402 (2011). For courts and legal commentators, judicial review’s hypersalience likely outstrips its influence on agency conduct.

The sign fallacy and the focusing illusion help explain the enduring appeal of remedial purity in administrative law. It’s true that excusing more agency mistakes would give agencies greater latitude to make those mistakes. If you assume that the incentive effect is large, cutting agencies slack will lead directly to the rise of procedurally defective and poorly reasoned rules. To discourage bad behavior, it’s essential to swiftly and severely punish agencies when they err. Courts thus embrace a rigid remedial rule, confident that they need not worry unduly about the costs of disrupting agency business. Those costs must pale in comparison to the damage that unrestrained agencies could inflict.

But a rule-bound approach to remedy looks more questionable if you relax the assumption that judicial review is the primary reason that agencies adhere to procedural rules, offer reasons for their decisions, or conform to law. If agencies have other reasons for behaving well, the adoption of a flexible remedial standard might not affect their behavior. Agencies will still desire information on the technical feasibility and political acceptability of their proposals. They will still need to cultivate their reputations with private industry, members of Congress, and the public. 405 See generally Daniel Carpenter, Reputation and Power 33–70 (2010) (highlight­ing the importance of agency reputation). They will still be full of lawyers who care about fidelity to law and will resist efforts to skirt it. 406 See Neal Kumar Katyal, Internal Separation of Powers: Checking Today’s Most Dangerous Branch from Within, 115 Yale L.J. 2314 (2006) (discussing internal checks in the executive branch). They will still want to adopt procedures that yield accurate, fair, and defensible decisions. 407 See Vermeule, supra note 81, at 1925 (describing why agencies have reasons independent of the law to promote accuracy). And agencies will still be staffed by civil servants who have a professional commitment to serving the public interest without bias or favor. 408 See Long, supra note 399, at 814–15 (emphasizing how the institutions of bureaucracy can enable the pursuit of the public interest).

The claim here is not that agencies are angels. The claim, instead, is modest: that holding more agency errors harmless might not much affect agency incentives. And that’s the margin that matters. A relaxed reme­dial approach is not the same as the elimination of assertive judicial review. Agencies that fear judicial review will still fear it, even if the rule of prejudicial error becomes somewhat more prominent. An agency official would be foolish to put much stock in the uncertain prospect that litigators could perhaps salvage a defective rule down the line. Yes, that response might be “rational” in a narrow economic sense. But agency officials are unlikely to calculate their legal exposure with such refine­ment that the downstream possibility of remedial flexibility will change their behavior. Competing incentives—to do the job right while preserv­ing agency resources, to preserve credibility, to assuage interest groups and congressional overseers, to avoid litigation if at all possible—will usually swamp the incentives created by modest adjustments to remedial doctrine. The story shouldn’t be overdrawn: Such adjustments might sometimes matter to some agencies. But there’s no evidence that they’ll make any difference most of the time.

There’s an opening, then, for administrative law to rethink its commitment to remedial purity. If holding more errors harmless is unlikely to systematically decrease the quality of agency decisions, the benefits of remedial purity dissipate. At that point, the costs come into sharper focus. Every vacated agency action wastes government resources—when it comes to rulemaking, a tremendous amount of resources. Bene­ficial agency projects can be delayed or derailed. Instead of reflexively forcing the government to bear those costs, the courts should be more willing to ask: To what end?

Without question, shifting from an inflexible rule to a remedial standard would force courts to confront a series of complex counter­factuals. Would the agency have made a different choice if it had corrected its error at the time of decision? Will letting the agency off the hook embolden it to ignore procedures or offer slipshod reasons for future actions? Might it embolden other agencies to do so? The questions are so intractable that it’s tempting to embrace a rule that doesn’t depend on answering them.

But the questions can’t be avoided. A rule-like approach to remedy is defensible only if the benefits of additional agency scrupulousness, over time, outweigh the costs of invalidating agency actions that would have been left undisturbed under a relaxed remedial standard, taking into account the decision costs associated with a standard. Remedial purists must therefore believe that agencies are quite sensitive to modest changes in remedial approaches. They must believe that adhering strictly to procedural rules has a big and salutary effect on the substance of agency decisions. They must believe that the costs of invalidation are small relative to those benefits since agencies can always try again on remand. In other words, purists’ embrace of a rule rests on implicit wholesale answers to the very questions that a remedial rule allows them to avoid answering at retail. What if those wholesale answers are wrong?

At any rate, courts routinely make judgments under conditions of uncertainty about the systemic effects of withholding a remedy for a legal violation. Courts can only guess, for example, at whether relaxing the exclusionary rule for Fourth Amendment violations will make police officers more cavalier about their constitutional obligations. Answering that question with any confidence requires intimate knowledge of how police departments work. How do officers learn about the rules governing exclusion? How (if at all) do they discover if evidence they collected is eventually suppressed? Do officers face consequences when evidence is excluded? How much does the downstream prospect of exclusion affect their behavior on the streets? Courts know next to nothing about these matters. Yet they do their best with the information they have, sensitive to the costs of excluding hard-earned evidence when the systemic benefits are elusive. 409 See, e.g., United States v. Leon, 468 U.S. 897, 922 (1984) (holding that “the marginal or nonexistent benefits produced by suppressing evidence obtained in objectively reasonable reliance on a subsequently invalidated search warrant cannot justify the substantial costs of exclusion”). They muddle through. Why not here?

The answer, I suspect, is not because the harmlessness inquiry is more tractable for the exclusionary rule than it is for administrative law. The answer, instead, is because the courts are not motivated to revisit the implicit policy judgments that underwrite remedial purity. Judges of whatever political stripe are acculturated into a legal community that views the administrative state as a leviathan that can be tamed only through zealous judicial oversight. 410 See supra section I.A (discussing the role of prejudicial error). There’s no substantial constituency clamoring about the serious costs of that attitude. Matters are different for the Fourth Amendment because part of the legal community has rallied to the argument that the costs of the rigid application of the exclusionary rule—that criminals walk free because the constable blundered—are too high. That policy argument found a sympathetic audience in the courts and harmlessness became the tool for striking a different balance between governmental efficiency and individual rights.

In administrative law, too, the call for remedial restraint will res­onate only to the extent that voices in the legal community sharpen the argument that automatically vacating defective agency actions imposes needless costs on effective governance. If that argument gains traction, the rule of prejudicial error will be at hand to sand off the harsh edges of judicial review.

Courts may also become more creative about acquiring information about what might have happened if the error had never occurred. Why aren’t split proceedings more common in administrative law, for one example? 411 See Daugirdas, supra note 23, at 309 (recommending bifurcated hearings). Such bifurcation is common in traditional civil cases, in which trial courts decide the merits of a case before turning to a separate damages phase. 412 See Fed. R. Civ. P. 42(b) (allowing bifurcated trials); Charles Alan Wright et al., 9A Federal Practice & Procedure: Civil Rules § 2390, Westlaw (database updated Apr. 2016) (“[A] significant number of federal courts, in many different kinds of civil litigation, have ordered the questions of liability and damages to be tried separately.”). Michigan v. EPA 413 135 S. Ct. 2699 (2015). offers an accidental example of the benefits of bifurcation. Once the Supreme Court cleared away the merits arguments, the D.C. Circuit on remand could focus on the remedial inquiry: whether the agency’s errors were sufficiently grave as to warrant invalidation of the rule. 414 White Stallion Energy Ctr., LLC v. EPA, No. 12-1100, 2015 WL 11051103, at *1 (D.C. Cir. Dec. 15, 2015) (per curiam) (order remanding proceeding to EPA). That question is unlikely to get adequate attention at the merits stage—especially from agencies that may be initially reluctant to press the argument that they were so committed to doing what they did that additional procedures couldn’t possibly have made a difference.

The unwillingness to characterize agency errors as harmless is especially anomalous given the frequency with which trial errors are held harmless. That remedial practice countenances and encourages proce­dural sloppiness from trial courts, at least on the margins. And it also requires reviewing courts to ask difficult counterfactual questions about how the trial would have progressed without the error. Yet reviewing courts don’t adopt a rigid remedial rule to keep trial courts in line or because the inquiry is hard. Even in the criminal context, in which the stakes for individual liberty are high, errors that require automatic reversal are uncommon: 415 See Arizona v. Fulminante, 499 U.S. 279, 309–10 (1991) (discussing the narrow circumstances in which automatic reversal is appropriate). They include a denial of self-representation, 416 See Faretta v. California, 422 U.S. 806, 834 (1975). denial of counsel, 417 See Gideon v. Wainwright, 372 U.S. 335, 344–45 (1963). and judicial bias. 418 See Tumey v. Ohio, 273 U.S. 510, 522–23 (1927). For all other errors, appellate courts do their best, full in the knowledge that they may err. 419 See, e.g., Haddad v. Lockheed Cal. Corp., 720 F.2d 1454, 1459 (9th Cir. 1983) (“The danger of the harmless error doctrine is that an appellate court may usurp the jury’s function, by merely deleting improper evidence from the record and assessing the sufficiency of the evidence to support the verdict below.”). Agency mistakes, in contrast, seem to be treated as structural. Why the dif­ference? After all, what are trial courts but specialized agencies with unusual tenure rules that exercise congressionally delegated authority to adjudicate disputes?


Administrative law should be more attentive to the possibility of holding agency errors harmless, at least in those categories of cases in which picayune procedural violations can yield costly invalidation. This Article identifies a few such categories. An agency, for example, might skip notice and comment because it thinks it’s not required: Maybe the agency’s rule is not a legislative rule, maybe it’s the logical outgrowth of a proposed rule, or maybe the agency has good cause for skipping the procedures. A reviewing court, drawing on malleable administrative law standards, might disagree with the agency’s judgment. But when the mistake arises from reasonable disagreement about the application of an open-ended standard, invalidation won’t make the agency better at applying that standard in the future. As importantly, agencies that skip notice and comment often solicit considerable public feedback outside the formal process. Yet the courts tend to credulously accept challengers’ claims that the absence of the formal process meant that they never had a chance to offer their views—even if the facts show otherwise. When a challenger knew what the agency meant to do, when it had an oppor­tunity and means to offer feedback, and when the agency re­sponded to the feedback it received, the proper attitude toward the challenger’s claim is not credulousness. It’s skepticism.

In the Chenery context, too, harmless error should come into play more often. The lower federal courts, for example, routinely invalidate agency actions infected by a Prill-type error. In those cases, however, agencies are very unlikely to revise their views about statutory meaning on remand, partly because their policy priorities already inflect their views about statutory meaning. Vacating is wasteful, at least when the agency confirms to a reviewing court that it prefers its interpretation to other available options. The courts should also hold errors harmless more often when agency decisions rest on inappropriate or incorrect factual determinations, but the weight of the record evidence dem­onstrates that the agency’s error did not affect its ultimate decision. Although the courts engage in harmless error review more often than is commonly appreciated, they do so haphazardly and in the shadow of Chenery. The practice should be acknowledged, defended, and regularized.

At the same time, not all agency errors can or should be treated as harmless. When an agency has acted beyond its legal authority, for example, there’s no point in asking whether it might have reached the same decision had it adhered to the proper procedures or offered a different explanation. Such errors should presumptively be treated as prejudicial. The same should hold if an agency skips notice and comment to avoid public scrutiny. Similarly, Chenery’s core—“assuring that initial administrative determinations are made with relevant criteria in mind and in a proper procedural manner” 420 Mass. Trs. of E. Gas & Fuel Assocs. v. United States, 377 U.S. 235, 248 (1964). —can and should remain intact. The risk that agencies might otherwise fail to supply valid reasons at the time of decision is substantial.

In general, however, the courts should abandon the appealing but unlikely assumption that mechanically rebuking agencies for their errors will improve agency decisionmaking so much that the costs are worth bearing. This is a place for a standard, for all the messiness and uncertainty that entails. Judge Friendly concluded as much almost a half-century ago:

Although, when I began my labors, I had the hope of discovering a bright shaft of light that would furnish a sure guide to decision in every case, the grail has eluded me; indeed I have come to doubt that it exists. Determination when to reverse and remand a decision that an administrative agency had power to make, and sufficient evidence to support, is, I fear, perhaps more an art than a science. 421 See Friendly, supra note 28, at 199–200.

Administrative law would do well to remember these words.