Today I’d like to sketch some lessons for the deference debates from the regulatory process as viewed from the inside. My main goal is to convince you that the regulatory process has important insights to offer to the deference debates.

That point has often been overlooked, especially by the courts. As now-Justice Kagan and Professor David J. Barron explained back in 2001,

The Court’s approach [to deference] treats agencies as unitary actors—each an undifferentiated ‘black box,’ from which decisions issue impersonally. But agencies are multi-faceted organizations, made up of diverse actors with diverse attributes and orientations. . . . Perhaps only the courts, among those concerned with administration, routinely neglect this aspect of internal agency structure.¹David J. Barron & Elena Kagan, Chevron’s Nondelegation Doctrine, 2001 Sup. Ct. Rev. 201, 234–35 (2001) (emphasis added). Barron and Kagan specifically argue that the Court disregards agencies’ hierarchic structure. My point is more general: the regulatory process has important lessons for the deference debates, whether we consider its vertical or horizontal relationships.

This distorted view of the agencies, Kagan and Barron argued, introduces errors into the Court’s deference jurisprudence.

Today I’d like to build on and broaden Kagan and Barron’s insight. The Court’s deference jurisprudence views the agencies as monads, each a simple and solid whole unto itself. But agencies are not at all monads; rather, they are made of many parts, and they are themselves parts of the broader executive machine. The Court’s view of the agencies as monads means that, in the deference cases, it has not been asking the right questions about agency expertise and political accountability, which are the two primary bases for the presumption that Congress intends agencies to resolve ambiguity in statutes and regulations they administer. My goal here is not to argue for any particular position on deference but to sketch some important questions that should, but presently do not, feature in the Court’s considerations in light of the realities of the regulatory process.

Let’s take up the expertise rationale for deference first. That rationale goes something like this: courts should defer to agencies when agencies have and use expert knowledge that enables a better interpretation of an ambiguous provision. In Chevron U.S.A., Inc. v. Natural Resources Defense Council,²467 U.S. 837 (1984). the Court described the expertise that warrants deference as “more than ordinary knowledge respecting the matters subjected to agency regulations.”³Id. at 844 (quoting United States v. Shimer, 367 U.S. 374, 382 (1961)). The point is one of institutional competence: agencies with expertise are better situated to interpret ambiguous text than inexpert courts. As the Court put it in Kisor v. Wilkie,⁴139 S. Ct. 2400 (2019) (plurality opinion). “[w]hen the agency has no comparative expertise in resolving a regulatory ambiguity, Congress presumably would not grant it that authority.”⁵Id. at 2417.

There is an implicit condition embedded in the expertise rationale. It is this: deference is appropriate only when the interpreting agency cannot reasonably transfer its expertise to the reviewing court. After all, if it could transfer its expertise by educating the court, then the court would have no reason to defer; it would simply evaluate the agency’s interpretation with the benefit of the expertise the agency has helped it to acquire. Doing so would avoid deference’s indisputable costs, in particular the cost of making the government in some measure the judge in its own cause. Perhaps that is a cost we are willing to bear for the sake of the benefits expert interpretation offers—but we shouldn’t bear it if we don’t have to, and we don’t have to if agencies can impart their expertise to courts. So, courts should defer only when agencies cannot reasonably transfer their expertise.⁶See Paul J. Ray, Lover, Mystic, Bureaucrat, Judge: The Communication of Expertise and the Deference Doctrines (The C. Boyden Gray Center for the Study of the Administrative State, Working Paper No. 2332, 2023), https://perma.cc/ET7K-JZV8.

Curiously, the courts seem to assume that expertise is incommunicable: agencies know something that judges do not, and between the two a great gulf is fixed. We see this, for instance, in the plurality opinion in Kisor. The plurality tells us in an example that the Transportation Security Administration (“TSA”) occupies a better position than a court to know why “TSA ban[s] liquids and gels in the first instance” and what “makes them dangerous.”⁷Kisor, 139 S. Ct. at 2413 (plurality opinion). That is fair enough, but the TSA could surely educate a reviewing court in these matters. The plurality simply assumes that because this special knowledge about liquids and gels originates in the TSA, it must remain there.

Let’s reflect for a moment on why courts assume agencies cannot share their expertise. It cannot be because agency expertise is incommunicable in principle, as is, say, knowledge of what it is like to see the color red. On the contrary, as Max Weber wrote, “in principle a system of rationally debatable ‘reasons’ stands behind every act of bureaucratic administration.”⁸Max Weber, 2 Economy and Society 979 (Guenther Roth & Claus Wittich eds., 1978). Perhaps, then, it is because the costs of agencies to express, and of judges to understand, agency expertise are just too high, higher than the costs of deference. This sounds plausible. After all, often in daily life we rely on others—doctors, engineers, lawyers—whose judgments are based on special knowledge that is sharable in principle but that neither we nor they wish to spend the time and effort to share.

Here is where the monadic error creeps in. For an expert agency is very little like a doctor, engineer, lawyer, or any other expert person. A person with specialized knowledge can form a plan of action or recommendation without sharing his special knowledge with anyone. That is because the expertise resides in the same self as the will that makes a decision or renders a recommendation. But agencies are not like that. Their expertise is distributed across the minds of their staff. The agency’s decisions are also made in the minds of the staff. But unlike in a person, those minds may not be the same. The Environmental Protection Agency (“EPA”) has deep knowledge about the Clean Air Act, and it also makes decisions about implementing the Clean Air Act. But the minds with the expertise may well be different than the minds that make the decisions.

Indeed, that will very often be the case. The chief decision-makers within EPA—the Administrator, the various Assistant Administrators, and other leadership—often have backgrounds in environmental policy, but very rarely do they have deep expertise relevant to a given regulation. That expertise usually belongs to the subject-matter experts, or SMEs in executive parlance, who specialize in particular programs. For the SMEs’ expertise to influence regulations, it must make the journey from the SMEs to the decision-makers.

But it cannot stop there. The SMEs’ expertise must also be transmitted to the rest of the agency team working on the regulation. And it must also make it to all the other officials, at the Office of Information and Regulatory Affairs (“OIRA”) and the Department of Justice (“DOJ”), at the EPA’s sister agencies and in the West Wing, whose agreement or at least non-opposition is needed for the regulation to go forward. Many of these officials will have significant environmental policy experience and education, but again, few will have the deep expertise that the specialist SMEs do. The SMEs must therefore explain their special knowledge to others, including many non-specialists, within the executive branch. As Anya Bernstein and Cristina Rodriguez put it recently, the regulatory process is “permeated with norms of justification, explanation, and persuasion” and agency experts must comply with these norms if they are to shape the content of regulations.⁹Anya Bernstein & Cristina Rodriguez, The Accountable Bureaucrat, 132 Yale L.J. 1600, 1639 (2023).

To be sure, sometimes these norms must give way to executive exigencies; busy officials cannot afford to plumb every interpretation in every rulemaking down to its roots. But for an important interpretation, a critical mass of officials have strong incentives to understand why they should support and how they can defend it. And this means they demand, and the SMEs provide, explanations of the SMEs’ expertise on which the interpretation rests.

Agencies, then, are not monadic individuals who make interpretive decisions on the basis of expertise lodged deep in their own selves; they are instead groups of people who deliberate and who must therefore share their expertise with each other. The lesson for Chevron deference (and for Auer deference, for that matter) is this: even if it is plausible that the costs for an individual expert to share his expertise with a reviewing court would be unduly high, that fact tells us little about whether it would be unduly costly for an expert agency to do so. For the expertise that shapes important regulations is already being shared. That agency experts already communicate their expertise to non-experts within the executive branch is a reason to believe the additional costs for them to communicate it to non-expert judges would be modest. And that executive non-experts already acquire the expertise they need to evaluate arguments about interpretations is a reason to suspect that non-expert judges could often do the same.

At the very least, it should be clear that the costs of disclosing the expertise that is the subject of deliberations within the regulatory process are different than the costs of disclosing expertise hidden within a single mind. Perhaps those costs are still higher than the costs of deference. My point is just that the courts have not asked that question. But it is a question they need to ask and answer if Chevron’s expertise rationale is to be complete and persuasive.

Now let’s turn to Chevron’s accountability rationale. The Chevron Court placed great weight on the fact that the agencies are supervised by the President, who is in the Court’s words “directly accountable to the people.”¹⁰Chevron U.S.A., Inc. v. Nat. Res. Def. Council, Inc., 467 U.S. 837, 865 (1984). “Federal judges,” the Court explained, “have a duty to respect legitimate policy choices made by those” who are responsible to the people.¹¹Id. at 866. So, courts must not second-guess the executive branch’s interpretation so long as it does not contravene the unambiguous statutory text, even if it is not the interpretation “the court would have reached if the question initially had arisen in a judicial proceeding.”¹²Id. at 866 n.11.

Now, it is not immediately obvious why Chevron deference promotes accountability. After all, Congress too has a democratic mandate—indeed, a superior one for at least two reasons. First, the Constitution the people made vests legislative authority in Congress rather than the executive. When courts and agencies follow Congress’s lead, they respect the people’s allocation of power in the Constitution. And second, Congress, but not the executive, labors under a set of protections designed to prevent factionalism. These protections increase Congress’s responsiveness to popular reason even as—indeed, because—they decrease responsiveness to popular passion and private interest.¹³See The Federalist No. 49, at 285 (James Madison) (Clinton Rossiter ed., 1961) (“[I]t is the reason, alone, of the public, that ought to control and regulate the government. The passions ought to be controlled and regulated by the government.”); see generally The Federalist No. 10 (James Madison). But the executive lacks the benefit of these protections from faction.¹⁴See The Federalist No. 70, at 394–95 (Alexander Hamilton) (Clinton Rossiter ed., 1961) (contrasting the plurality of the proposed legislative branch and its anti-factionalistic effect with the unity of the proposed executive).

The upshot is that the best way to promote accountability is to give effect to Congress’s policy decisions in the statutes it enacts. I don’t take the Chevron Court to disagree. Rather, I think its reasoning goes something like this: Congress has the strongest democratic pedigree, but the President’s is both real and greater than the courts’. So courts should not stand athwart the presidential will unless they are relatively certain that, by doing so, they are giving effect to Congress’s will. And this means courts should be confident that Congress has indeed spoken to a question before interfering with an interpretation adopted under the President’s auspices.

Implicit in this line of reasoning is the assumption that the agencies under the President’s supervision are not significantly more likely than courts to depart from congressional intent. After all, if they are, then deference to the President’s democratic mandate comes at the expense of Congress’s superior one. What basis do we have to believe that agencies under presidential supervision are not more likely than courts to depart from congressional intent in the zone of freedom that Chevron opens up for them?

Of course courts, no less than executive officials, can make mistakes. And courts, no less than executive officials, can willfully pursue their own priorities. Chevron’s attraction for many, it seems to me, has been its recognition that judges are not exempt from these behaviors, and that judges are therefore not likely, simply by dint either of skill or integrity, to serve congressional intent better than executive officials.

But executive officials do differ in some important ways from judges. Among the most important is that, while judges are immersed in particularity—they are called upon to interpret this particular provision in this statute governing this program—agencies are parts of a greater whole headed by the President. And presidents, because they sit atop dozens of agencies administering thousands of statutes, confront possibilities and pressures different than those judges face. Agencies’ incentives and chances to depart from congressional intent therefore differ from the incentives and chances of judges to do so.

We can begin to see the difference this distinction makes by looking back to Chevron itself. There, the Court understood that Congress meant to accommodate certain interests but did not clearly balance them with respect to the bubble concept. Rather than balance these interests itself, the Court chose to let the EPA do so, since it is accountable to the people through the President.¹⁵Chevron, 467 U.S. at 865–66. But of course, to the extent the President influences a rulemaking, he does not simply step into the shoes of the EPA Administrator, with a focus restricted to the Clean Air Act or even to environmental issues more broadly. He is tasked with the pursuit of many, many interests that are irrelevant to those Congress sought to accommodate in the program at issue in Chevron, and we may wonder whether the President might use the program to pursue these other interests rather than those Congress intended to accommodate in the Clean Air Act.

Several of the major questions cases seem to arise from agency action under a statute repurposed by the President to pursue interests far from those Congress had in mind when it enacted the statute. Consider West Virginia v. EPA.¹⁶597 U.S. 697 (2022). That case’s Clean Power Plan pursued a presidential goal of national energy resource management beyond Congress’s horizon when it enacted the Clean Air Act amendments. I think we find the same thing in Biden v. Nebraska¹⁷143 S. Ct. 2355 (2023). and Alabama Association of Realtors v. Department of Health and Human Services,¹⁸141 S. Ct. 2485 (2021) (per curiam). among others.

The concern I have raised becomes even more plausible if we reflect on the situation in which modern presidents find themselves. Neither the President nor his constituents can have any firm idea of the limits on agency and thus presidential powers, which are spelled out in thousands of statutes enacted over the decades and whose outer reaches are discovered on an as-needed basis. More and more, presidents seem to their followers and to themselves to be all-purpose problem-solvers, within the scope of whose ample powers all the day’s most pressing crises fall. To address these problems, presidents are tempted to formulate policy first and then grasp at any authority remotely useful to accomplish their goals. For presidents, the means by which a given high-level objective is pursued is a matter of relative indifference; the important thing is to achieve the objective. Presidents and their teams thus have strong incentives to strain interpretations to the breaking point. These are incentives that judges lack.

Presidents differ from judges in another important way. A judge’s ability to alter policy through a judgment about a regulation is restricted by the reach of the regulatory program itself. But presidents, because they coordinate policy across many agencies, are not so limited. They can set a general objective to be pursued by agencies across the federal government; President Biden’s climate executive orders are examples.¹⁹See, e.g., Exec. Order No. 14,030, Climate-Related Financial Risk, 86 Fed. Reg. 27967 (May 25, 2021). By doing so, presidents can shape domestic policy in a much more far-reaching way than can a judicial decision on a single regulation. And this means presidents can far more comprehensively displace congressional decisions.

It also means presidents can assume the prime policy-setting initiative, pursuing the kinds of across-the-board solutions that Congress can enact but that courts almost never can effectuate. Further, the public can see that presidents possess the initiative. The Presidency thus may come to displace Congress in the minds of many as the main origin and arbiter of domestic policy. This state of affairs would undermine Congress’s democratic mandate in a wholly different way than any judge can.

These concerns all flow from the status of agencies as parts of a regulatory machine helmed by the President. To be sure, nothing I have said establishes that these concerns are so powerful as to justify overturning Chevron. My point is just that, once we see that the President does not make decisions as if he were a monadic agency pursuing a limited mission but rather in light the many agencies he directs and the manifold priorities he pursues, we must recognize that presidents and courts are differently situated with respect to giving effect to congressional intent. And if that is so, then it’s far from clear that Chevron promotes accountability. It may in fact create space in which presidents can, and have incentives to, impede Congress’s own democratic mandate. It therefore becomes incumbent on the courts to ask whether deference promotes or hinders accountability. But courts have not asked that question so far.

So Kagan and Barron were right. The way the federal regulatory process works matters for the deference battles. The intra-executive habit of demanding and giving explanations raises questions about the expertise rationale; after all, if executive experts must explain their expertise to executive non-experts, how hard would it be to explain it also to judicial non-experts? And the realities of political direction by presidents, with their universal perspective and authority, should make us uncomfortable with the assumption underlying the accountability rationale that the executive branch is no less likely than the judicial branch to pursue congressional purposes. I’m sure there are many other lessons the regulatory process holds for the deference debates, but perhaps these are enough to get us started.