Embedded Rules

Rules are rules and orders are orders, and never the twain shall meet. Generations of scholars and practitioners were taught back in law school that the Administrative Procedure Act (APA) divides the universe of agency action into two exclusive and exhaustive categories: “rulemaking,” which is used for promulgating “rules,” and “adjudication,” which is used for issuing “orders.” Each of those modes of agency action has its formal and informal versions, and some statutes mandate “hybrid” procedures with an intermediate level of formality. But the starting point for analyzing a given agency action is to decide whether that action falls into the “rule” box or the “order” box, which are separate and distinct. That is what then-Professor, now-Justice Elena Kagan taught me back when I took her Administrative Law class as a 2L, and it’s what I’ve taught my students for the last fifteen years.

But it’s not quite right. “Rules” and “orders” are not, in fact, completely separate and non-overlapping categories. Sometimes an administrative action that is properly classified as an order contains within it—usually in the portion explaining the order’s legal basis—a statement that qualifies as a rule and ought to be treated as such. The fact that such a rule is embedded within an order does not make it any less of a rule. And that means that the process for formulating an embedded rule counts (or ought to count) as a “rulemaking” under the APA.

Does that mean that agencies are routinely and flagrantly violating the APA by issuing orders that contain embedded rules? No, it doesn’t. Most of the rules embedded within agency orders can properly be characterized as interpretive rules (agency statements that explain what some existing statute or regulation means) or policy statements (declarations of how the agency intends to exercise its discretionary authority) rather than legislative rules (which create new rights, duties, or prohibitions). The APA explicitly exempts interpretive rules and policy statements—sometimes referred to collectively as non-legislative rules—from notice-and-comment requirements.11. 5 U.S.C. § 553(b)(A). See also John F. Manning, Nonlegislative Rules, 72 Geo. Wash. L. Rev. 893 (2004). Indeed, a non-legislative rule can usually be issued without any formal procedure at all, save what is necessary to ensure that the agency’s action is not arbitrary, capricious, or an abuse of discretion.22. 5 U.S.C. § 706(2)(A). So an agency that uses proper adjudicative procedures to issue an order that contains an embedded non-legislative rule has not violated the APA. Even though a rule is inherently the product of a rulemaking—whether or not the rule is embedded in an order—a non-legislative rule can be issued as part of an order without any additional process.

For this reason, recognizing that embedded rules are rules would not throw the administrative state into chaos, nor would it require jettisoning well-established administrative law doctrines. Yet resolving the conceptual confusion regarding the appropriate classification of embedded rules would have meaningful and desirable consequences. For starters, certain doctrinal anomalies and uncertainties might be easier to resolve if we recognize the agency orders often contain embedded rules, rather than persisting in the idea that orders and rules are entirely non-overlapping categories. Perhaps more importantly, in some cases seeing embedded rules for what they are might lead to a markedly different scope for certain administrative law requirements. And in other cases, this recognition would help put existing doctrines on a more solid foundation.

The case for recognizing that agency orders may have rules embedded within them starts with the APA’s text. As relevant here, the APA defines a “rule” as “the whole or a part of an agency statement of general or particular applicability and future effect designed to implement, interpret, or prescribe law or policy”;33. 5 U.S.C. § 551(4). the APA further defines a “rule making” as an “agency process for formulating, amending, or repealing a rule.”44. 5 U.S.C. § 551(5). An agency “order” is a final disposition in a matter other than a rulemaking,55. 5 U.S.C. § 551(6). and the process for formulating an order is an “adjudication,” whether or not it resembles what we ordinarily think of as an adjudicative process.66. 5 U.S.C. § 551(7). Many agency orders include statements about the legal and policy considerations that determined the agency’s resolution of the case that is the subject of the order. Sometimes these explanations merely restate clearly established law before proceeding to apply that law to the particular facts of the case at hand. But in many other cases, the agency’s statements of law or policy go further than what the agency had previously declared. These statements, which announce the agency’s view of what the law requires or the policy considerations the agency will take into account when exercising its discretion, look a lot like rules, especially if one takes the APA’s statutory definition of “rule” seriously.

Now, the fact that a statement contained within an order meets the APA’s definition of a rule is not necessarily dispositive. The APA, as interpreted by courts, has developed in ways that the enacting Congress likely did not foresee, and that in some cases are hard to square with the statute’s text. And while critics have objected to this more freewheeling approach to interpreting the APA,77. See Jeffrey A. Pojanowski, Neoclassical Administrative Law, 133 Harv. L. Rev. 852 (2020). others have insisted that the APA is a kind of “super-statute”88. See Kathryn E. Kovacs, Superstatute Theory and Administrative Common Law, 90 Ind. L.J. 1207 (2015), or “quasi-constitutional” framework statute99. See Bruce Ackerman, The Emergency Constitution, 113 Yale L.J. 1029, 1077–78. that may legitimately evolve over time in a common-law fashion. If there are strong functional or policy arguments for treating a general statement interpreting or prescribing law or policy as something other than a rule if that statement is embedded within an order, then perhaps we should overlook the fact that such a characterization is in tension with the APA’s text.

But in fact the failure to recognize that a rule is a rule even when it is embedded within an order produces conceptual confusion and anomalous results that fail to advance the APA’s core purposes. This confusion is perhaps most evident with respect to the relationship between two separate and distinct lines of administrative law doctrine, both of which concern the question of when an agency may issue a rule (or the practical equivalent of a rule) without going through the APA’s notice-and-comment rulemaking process.

First, as noted above, agencies often include in administrative orders statements that look and function like rules. When is this permissible? The short answer, under the prevailing doctrine, is “almost always.”1010. See M. Elizabeth Magill, Agency Choice of Policymaking Form, 71 U. Chi. L. Rev. 1383 (2004). The leading cases on this issue, the Supreme Court’s decisions in Chenery II and Bell Aerospace, embrace a forgiving standard that allows an agency to proceed by adjudication rather than rulemaking unless doing so would amount to an abuse of discretion.1111.  NLRB v. Bell Aerospace Co., 416 U.S. 267, 294 (1974); SEC v. Chenery Corp. (Chenery II), 332 U.S. 194, 201–03 (1947).

Second, agencies frequently issue standalone statements that the agencies characterize as non-legislative rules (interpretive rules or general statements of policy) that need not go through notice and comment. But which agency pronouncements are in fact eligible for this exemption, and which ones are actually legislative rules? The doctrine on this question, which has developed mainly in the courts of appeals rather than the Supreme Court, is murky and not entirely consistent. Still, the key test that the courts have developed for distinguishing non-legislative rules from legislative rules is the “force of law” test. That test asks whether the rule creates new legal rights or obligations—in which case the rule is legislative—or whether the rule merely explains legal rights or obligations created by some pre-existing statute, regulation, or other law (in which case the rule is interpretive) or announces the agency’s general priorities or objectives in exercising its discretion (in which case the rule is a general statement of policy).1212.  See, e.g., Am. Mining Cong. v. Mine Safety & Health Admin., 995 F.2d 1106, 1109 (D.C. Cir. 1993); Pac. Gas & Elec. Co. v. Fed. Power Comm’n, 506 F.2d 33, 38–39 (D.C. Cir. 1974). In deciding whether an agency statement qualifies as an interpretive rule, courts typically consider whether the agency’s alleged interpretive rule could be derived from the pre-existing law through a process that could reasonably described as “interpretation”—filling in the gaps or fleshing out the implications of an existing legal text—or whether the supposed interpretive rule seems more like an act of law-creation.1313.  See, e.g., Hoctor v. U.S. Dep’t of Agric., 82 F.3d 165, 170 (7th Cir. 1996). And in considering whether an agency statement qualifies as a general statement of policy, courts will often take into account additional factors such as whether the policy is framed in absolute rather than tentative terms, and whether as a practical matter the rule is likely to have a coercive effect on regulated parties.1414.  See, e.g., Chamber of Com. v. U.S. Dep’t of Lab., 174 F.3d 206, 212–13 (D.C. Cir. 1999).

The case law, commentary, and standard teaching texts treat these two lines of doctrine as distinct, though perhaps thematically related. And that separation suggests the possibility, which most commenters seem to treat as mildly puzzling but doctrinally correct, that the same agency declaration might be unlawful if issued as a standalone non-legislative rule but lawful if embedded in an agency order.1515.  See, e.g., David L. Franklin, Legislative Rules, Nonlegislative Rules, and the Perils of the Short Cut, 120 Yale L.J. 276, 316 (2010). In other words, there might be situations in which, if an agency were to announce that “statute X requires result Y” in a standalone rule issued without notice and comment, a court might decide that this is invalid because the connection between X and Y is not sufficiently tight for the rule to qualify as interpretive and that the rule is actually a legislative rule that requires notice and comment. However, if the agency were to resolve an individual adjudication by issuing an order that justified the result with the identical statement that “statute X requires result Y,” then the court would likely uphold the agency’s action on the grounds that it is not an abuse of discretion for the agency to proceed by adjudication rather than rulemaking, and the agency may therefore resolve the case before it by announcing what looks like a general rule.

That divergence in the doctrinal tests produces odd results. For one thing, it would seem to discourage an agency from announcing in advance the legal principles that it will use to decide individual cases. If an agency plans to invoke and apply a particular view of what the law requires when resolving individual disputes in adjudication, then it seems intuitive that the agency should—and at the very least should be allowed—to announce that view ahead of time. But if such a declaration, though permissible in the order itself, would be viewed by a court as “legislative” rather than “interpretive,” then it would be unlawful for the agency to announce the interpretation in advance unless the agency goes through the onerous and time-consuming notice-and-comment process. What is gained from such a prohibition on quick-and-easy advance notice? Sure, clever lawyers (and overly clever law professors) have concocted arguments as to why it makes sense to bar an agency from announcing ahead of time the legal positions that the agency plans to apply in concrete cases. But those arguments have the flavor of strained attempts to rationalize a feature of the doctrine that doesn’t make a whole lot of intuitive sense.

The disharmony in these doctrinal tests also looks odd when approached from the other direction. Suppose an agency tries to couch a legal conclusion as nothing more than an interpretation of existing law, but the agency’s declaration is in fact legislative in nature—it establishes a new requirement, rather than simply explaining what some existing statute, regulation, or precedent already requires. If an agency promulgated such a statement as a standalone rule, without a notice-and-comment process, then a court would properly invalidate it. And this is not just a formality. As courts have emphasized, the APA’s notice-and-comment requirements are rightly understood as vital safeguards to ensure that new rules are carefully considered and properly vetted.1616.  See, e.g., Hoctor, 82 F.3d at 170–71. That isn’t to deny the ongoing debate about the value of notice-and-comment procedures. But the doctrine in this area presumes that notice and comment is necessary for legislative rules because when an agency is imposing new obligations or creating new entitlements, a more elaborate process is required to ensure broad public input into the rulemaking process and broad public scrutiny of the agency’s proposal. Why should this change if the agency embeds the legislative rule within an adjudicative order? If we assume that the rule is sufficiently legislative in character to require notice-and-comment procedures in the ordinary course, why should the rule suddenly qualify for an exemption from notice and comment simply because the rule is inserted into an order?

The most plausible answer to this question is that the procedural safeguards of formal adjudication are an adequate substitute for the procedural safeguards of notice-and-comment rulemaking. If an agency wants to make a legislative rule, this argument implies, it can do so either by going through the APA’s rulemaking process or by going through the APA’s formal adjudication process, unless the agency’s organic statute restricts that choice. But the notion that the APA’s adjudicative procedures are an appropriate substitute for notice-and-comment rulemaking procedures is questionable. For one thing, not all of the orders that contain embedded rules are the product of formal adjudications; some are the produce of informal (or less-formal hybrid) adjudications. But even putting that to one side, adjudicative procedures—even formal ones—are different from rulemaking procedures. Adjudicative procedures are mainly about ensuring fairness to the parties whose interests are most directly affected (usually those who are seeking some benefit or who are trying to avoid some sanction) and guaranteeing those parties’ right to be heard. Rulemaking procedures, as they have been interpreted and developed by the courts, are designed to facilitate broad-ranging public input. So while formal adjudications obviously entail substantially more procedural formality than do standalone interpretive rulemakings, it is odd to suggest that the procedures associated with formal adjudication are appropriate for promulgating legislative rules, when the APA and administrative law doctrine treat rulemaking and adjudication as such different creatures in so many respects.

In sum, there is little formal or functional justification for treating embedded rules differently from standalone rules with respect to the question whether they can be issued without observance of notice-and-comment requirements. If an agency’s statement of law or policy would be deemed an invalid legislative rule if it were issued as a standalone pronouncement, then the agency should not be permitted to incorporate that statement into an adjudicative order. But if a given agency statement of law or policy could be lawfully incorporated into an adjudicative order as the grounds for decision, then the agency ought to be permitted to issue that same declaration as a standalone interpretive rule or policy statement, taking advantage of the APA’s express exemption of non-legislative rules from notice-and-comment requirements.

Declaring that these two lines of doctrines should be harmonized naturally invites the question of how they should be harmonized. Should the doctrine on rulemaking via adjudication become more stringent, refusing to countenance orders that contain embedded rules unless those rules could be upheld as non-legislative under the prevailing test applied to standalone rules? Or should the doctrine on distinguishing legislative from non-legislative standalone rules become more forgiving, allowing agencies to issue, without notice and comment, any rule that the agency could properly include as part of an adjudicative order under the prevailing understanding of Chenery II and Bell Aerospace? Or should the doctrines meet somewhere in the middle? I lean toward some version of that last option, though I do not pursue the question further here. My main argument is that, whatever the right test is, it should be consistent. An embedded rule is a rule. If a rule is a legislative rule, it should have to go through the notice-and-comment process, even if it is embedded. If a rule is not a legislative rule, then it should not have to go through the notice-and-comment process, even if it is promulgated as a standalone rule.

There’s an important qualification here, one that highlights another way in which understanding embedded rules as rules would help clarify the doctrine. Under some circumstances, an agency conducting an adjudicative proceeding might recognize that the question before it has never come up before, and that resolving that question properly requires articulating and applying a new legislative rule, one that could not be characterized as a mere interpretation or general policy statement. Notwithstanding the above discussion, under some circumstances the agency should be able to do this—but not because embedding a legislative rule in an order somehow transforms the rule into something other than a rule. Rather, the APA includes another set of exemptions from the usual notice-and-comment requirements, which apply when “the agency for good cause finds” that notice and comment would be “impracticable, unnecessary, or contrary to the public interest.”1717.  5 U.S.C. § 553(b)(B). In light of this provision, if an agency determines that failing to resolve an issue of first impression by announcing and applying a legislative rule would be “contrary to the public interest” (or, in a slightly different formulation, that it would have been “impracticable” to have formulated and announced the rule prior to the adjudicative proceeding), then the agency should be permitted to issue an order containing that legislative rule. But the agency ought to be required to defend this invocation of the “good cause” exemption to the same extent that the agency would have to do so if it had invoked the exemption to issue the rule in a standalone rulemaking proceeding: the agency should be required to expressly acknowledge and defend its invocation of the good cause exemption, and meet the more demanding judicial scrutiny of such claims. And insofar as courts require agencies to treat rules adopted pursuant to the good cause exemption as interim rules, which are to be followed by full notice-and-comment proceedings,1818.  See, e.g., Mid-Tex Elec. Coop., Inc. v. FERC, 822 F.2d 1123, 1132 (D.C. Cir. 1987); Am. Fed’n of Gov’t Emps. v. Block, 655 F.2d 1153, 1156–59 (D.C. Cir. 1981). that same requirement ought to apply when agencies rely on the good cause exemption to justify the issuance of legislative rules embedded in adjudicative orders.

Another context in which recognizing that embedded rules are nonetheless rules might matter concerns the application of other statutorily-mandated procedures that apply to agency “rules.” If embedded rules are rules in the relevant legal sense, then these statutory mandates apply to those rules that agencies announce within their adjudicative orders. And recognizing that might make an important difference, at least in some cases.

Consider, by way of illustration, the Congressional Review Act (CRA).1919.  5 U.S.C. § 801 et seq. The CRA requires agencies to report their “rules” to Congress,2020.  5 U.S.C. § 801(a)(1)(A)–(B). and gives Congress sixty legislative days during which Congress may use special fast-track procedures to pass a joint resolution disapproving the agency’s rule.2121.  5 U.S.C. § 802(a). The joint resolution is then presented to the president just like any other statute. If a joint resolution of disapproval is enacted into law, then not only may the agency not maintain or implement the disapproved rule,2222.  5 U.S.C. § 801(b)(1). but the agency is also barred from issuing a new rule that is substantially the same as the disapproved rule, unless Congress enacts new legislation authorizing that rule.2323.  5 U.S.C. § 801(b)(2).

The CRA expressly adopts the APA’s definition of a rule, subject to a few exceptions.2424.  5 U.S.C. § 804(3). Most importantly for present purposes, interpretive rules and general statements of policy are considered rules for CRA purposes: agencies must report them, and they can be disapproved via CRA resolutions. But agencies, commentators, and Members of Congress all seem to have assumed, without much careful consideration, that administrative orders fall entirely outside the scope of the CRA’s coverage.2525.  See, e.g., Valerie C. Brannon & Maeve P. Carey, Cong. Rsch. Serv., R45248, The Congressional Review Act: Determining Which “Rules” Must Be Submitted to Congress 11 (2019). At first glance, that makes sense. The CRA, after all, applies explicitly and exclusively to agency rules, not to agency orders. But if every agency statement that meets the CRA’s definition of rule in fact counts as a rule, even if the rule is embedded within an order, then these embedded rules should also be covered by the CRA.

That latter approach is not only more consistent with the CRA’s text (which, again, simply borrows the APA’s definition of “rule”), but it would also more effectively advance the CRA’s purposes and objectives. For one thing, in those cases where agencies have a choice whether to proceed by adjudication or by (standalone) rulemaking, treating embedded rules as non-rules for CRA purposes gives agencies even stronger incentives to shift their rulemaking activity to individual adjudications. This evasion of the CRA comes at a cost, at least to the extent that one believes that it is usually better for agencies to announce their rules in advance. Now, this concern should not be exaggerated. There isn’t much evidence of agencies taking the CRA into account when deciding whether to proceed by adjudication or rulemaking. But the CRA was not used much until recently,2626.  See Bridget C.E. Dooling, Into the Void: The GAO’s Role in the Regulatory State, 70 Am. U. L. Rev. 387, 399 (2020). and if the possibility of CRA reversal becomes more salient, agencies may become more inclined to use adjudication rather than rulemaking. Moreover, even if we discount the possibility that the incentives created by this doctrinal anomaly would produce notable changes in agency behavior, it still seems rather arbitrary that those agencies that do most of their significant policymaking through adjudication are insulated from the possibility of CRA reversal, while those agencies that proceed through rulemaking are not.

Additionally, treating embedded rules as something other than rules makes it too easy for agencies to circumvent the CRA’s bar on reissuing a disapproved rule. As noted above, a CRA disapproval resolution not only prohibits the agency from adopting or maintaining the disapproved rule, but it also prohibits the agency from reissuing that rule or issuing a substantially similar rule. But if rules embedded in adjudicative orders are not considered rules for CRA purposes, then in those contexts where agencies have the discretion to proceed through rulemaking or adjudication, CRA disapproval resolutions may be quite easy to sidestep. Consider the following hypothetical, based on the facts of Bell Aerospace. Suppose the National Labor Relations Board (NLRB) went through a full notice-and-comment process and issued a legislative rule declaring that a certain class of employee is permitted to unionize, but Congress passed, and the President signed, a CRA resolution disapproving that rule. Let’s suppose that the very next day, in the context of resolving a dispute over whether a group of these employees at a particular facility may unionize, the NLRB declares that they can, and issues an order that adopts verbatim the legal reasoning and arguments contained in the legislative rule that Congress disapproved. That seems like a pretty blatant circumvention of the CRA. But the CRA’s prohibition on reissuance applies only to rules that are the same or substantially similar as the disapproved rule. So if the legal conclusion articulated in the order is not a rule in the relevant legal sense, then the Board’s action would be permissible. Now, it might be possible to contrive a rationalization for why the CRA would bar an agency from reissuing a disapproved rule as a standalone rule but permit the agency to issue a substantively identical rule embedded in an order. But it would be much more straightforward to recognize that a rule is a rule, in which case there is no principled statutory basis for exempting such embedded rules from the CRA.

That view, if accepted, would have a few important consequences for how the CRA operates. First, it would mean that in addition to needing to report their legislative rules, interpretive rules, and guidance documents to Congress, agencies would also need to report adjudicative decisions that contain embedded rules—or, more accurately, the agencies would need to report their embedded rules, whether or not the agencies also choose to report the orders in which those rules appear. The CRA’s fast-track procedures could then be deployed to disapprove embedded rules.2727.  Because the CRA’s sixty-legislative-day clock for passing disapproval resolutions does not start until a rule is reported to Congress, there might be many older rules embedded in adjudicative orders that are still potentially susceptible to CRA review, as they have not yet been properly reported. Furthermore, any enacted CRA resolutions ought to be treated not only as barring an agency from reissuing a disapproved rule in the form of a standalone rule, but also as barring the agency from embedding a disapproved rule in an order. Of course, those who think that the CRA is a misguided law might embrace any loopholes that narrow the act’s scope and effect, no matter how unprincipled those loopholes might be. But if we focus on trying to faithfully implement the CRA—both attending to the statute’s text and trying to give effect to its general purpose—then it is much more plausible to treat rules as rules, rather than arbitrarily exempting from the CRA’s coverage those rules that happen to be embedded in orders.

Though the discussion above focuses on the CRA, the same line of argument would apply to any statute that imposes special requirements for agency “rules” or “rulemakings.” Agencies should not be able to evade such requirements by embedding rules within orders. As a formal matter, an embedded rule meets the definition of a “rule”; as a functional matter, the underlying rationales for imposing additional requirements on agency rules would typically apply with equal force to those rules issued in adjudicative orders. Of course, special requirements for legislative rulemaking might not apply to embedded rules, because many such rules could likely be classified as non-legislative. But requirements such as those contained in the CRA, which apply to both legislative and non-legislative rules, ought to apply, at least presumptively, to all embedded rules.

The current understanding of rules and orders as mutually exclusive, non-overlapping categories is misleading. Some rules are standalone rules, but other rules are embedded in orders. Embedding a rule within an order does not change its status as a rule. All of the APA’s procedural requirements for rulemaking, as well as additional statutory mandates concerning agency rules (such as the CRA), ought to apply to embedded rules in the same way that they apply to standalone rules. Such a reconceptualization of the relationship between rules and orders would not upend the doctrine. The basic frameworks could remain in place, and most current agency practices, including the practice of regularly embedding rules in adjudicative orders, could likely continue. Nevertheless, embracing this alternative understanding would require some substantial doctrinal adjustments. Those adjustments would be welcome, as they would resolve uncomfortable anomalies that have long bedeviled administrative law and would more effectively advance the underlying policies and purposes of various administrative rulemaking requirements.

Eli Goldston Professor of Law, Harvard Law School. I am grateful to Ben Eidelson, Todd Rakoff, and Dan Tarullo for helpful comments.