The Reasoning State: Theory, Interpretation, and Evidence, by Jed Stiglitz
*This is the ninth and final post in a symposium on Jed Stiglitz’s “The Reasoning State.” For other posts in the series, click here.
My main response to this symposium is gratitude. I thank Yale Journal on Regulation and the Notice & Comment editors, and Bridget Dooling especially, for the opportunity to discuss The Reasoning State with such amazing participants. And I am incredibly grateful to each of the contributors for taking the time to engage so thoughtfully and generously with the book, particularly at this very busy time of the semester. I also greatly enjoyed learning more about how many of the themes of the book live in the works of the contributors.
I cannot respond to all the excellent points raised by the commentors, but I see the entries as speaking to three main categories of issues: matters of theory, matters of interpretation, and matters of evidence. I agree with many of the points raised by my interlocutors, though in places I also seek to clarify my perspective and to gently resist on several margins. I also continue to consider the thoughtful contributions as I develop further work in this line.
Matters of theory
At the core of this book is a positive theory of institutions—a theory that seeks to describe the world and explain why it takes the form it does. The book also contains normative and empirical aspects, some of which I discuss below, but the main thrust of the book is to articulate a positive theory of the modern state and those empirical or normative/doctrinal aspects that either support or derive from the theory.
The main axis of the theory is a breakdown in representative institutions due to problems of information. I argue the following: In a complex society, we often cannot judge policy outputs themselves (by ourselves or as assisted by heuristics, though they help), nor typically can we know whether our representatives act in our interest. Yet we know that lobbyists, economic elites and others misaligned with our interests actively attempt to subvert our representative institutions. So we have reason to distrust our representatives. This problem of distrust drives down rates of re-election and thereby generates an incentive for representatives to resolve it. Though we might imagine legislative regimes that allow members to “show” their faithfulness to the electorate, the legislature cannot commit to procedural integrity and lacks an external enforcement technology (such as, e.g., effective judicial review). So they seek out other arrangements, and around the turn of the last century they stumbled upon the modern state: delegation of lawmaking authority to a class of entities that can be made to “show” the nexus between means and publicly stated ends—that is, to reason credibly—facilitated by institutional conversations between agencies and courts. What is distinctive of the administrative state, in this theory, is therefore not its ability to develop expertise—a common justification for agencies—but instead its ability to commit to procedural integrity, to reason credibly, and thereby to ameliorate problems of trust between the public and our public institutions. That ability to credibly reason is central to its political value as an institution and therefore to the continued viability of delegated authority and state capacity.
This theory relates most closely to the now-common idea that delegation exists so that members can avoid responsibility for their choices, well represented in Gorsuch’s Gundy opinion and in many prominent academic accounts. Both theories see electoral foundations as key to the shape of the modern state. What is different is that the theory of this book sees voters in a more charitable light, i.e., that they’re not easily duped—they face an information problem, of course, but they update beliefs rationally where they can. And the theory recognizes the crucial role of administrative law in supporting the modern state. The upshot is a theory that, despite similar electoral foundations to a Gorsuch-like theory, affirms the administrative state.
Bernstein raises deep questions about the role of theory in her thoughtful review. This is a positive theory, it is true, and seeks to explain why the world looks as it does. It posits a set of actors responding to the incentives established by their institutional context and information environment. But this does not commit the theory to an intentionalist understanding of human behavior. I agree with Bernstein that “actions often have effects that actors neither wanted nor predicted”—the metaphor I try to stitch into the project is one of explorers rather than engineers. Political actors often did not know what they were doing—but they sensed what worked for their interests and what did not, and they leaned into the former and away from the latter. The story is one of the political actors as a class surviving, but without necessarily full awareness let alone intent as to action, as against structural constraints that threaten their well-being.
The differential ability of some institutions to credibly reason sits at the core of this theory. Shah pushes on the distinction between “credible reasoning and the mere application of expertise.” Under the theory, expertise is often necessary to credibly reason. But it is also often not sufficient. To credibly reason, one must not only apply expertise, but also be able to show to others that one applied expertise in connection with a stated end. To take a simple example, suppose I have two coins, one in each fist, one fake gold and the other real gold. I know which is fake and which is real (you do not). In this game, you keep the coin in the hand you pick; I keep the coin in the other hand. I say to you, “pick my left hand to win the real gold coin.” You would have good reason to distrust my statement that the coin is in my left hand—you would probably wrap yourself into a Princess Bride “battle of the wits”-like circle over the hand that truly contained the gold coin. But note that it is without question I applied expertise, that is, my specialized knowledge of the hand containing the real gold coin. But to the stated end? No. I failed to credibly reason entirely. To credibly reason, I might instead allow you or an independent third party to inspect the coins, show you studies of their weights or chemical compositions, etc., such that if I say you win the gold coin by selecting the left hand, you have some assurance of a nexus between stated means (picking the left hand) and stated ends (winning the gold coin). This is of course a simple example, but it illustrates the crucial difference between expertise and credible reasoning that motivates the project.
It also relates to a motivating puzzle—why doesn’t Congress increase its capacity? Observers have long wondered why, if expertise is what is gained by delegation, the legislature cannot instead acquire expertise natively. I agree with Reynolds and Walker that Congress might increase its capacity. It is within its authority and powers to do so. But I also argue that it is not in members’ interest to do so. Why? Because expertise is not the binding constraint on legislative institutions that leads to delegation of authority. The binding constraint, instead, is the ability to credibly reason and to be trusted. As I explain in the book, they cannot commit to credible reasoning; nor is external enforcement feasible. So investing in expertise promises few (electoral) returns to members or to the institution more broadly. Even with expertise, members would be rather like the person saying the real gold coin is in their left hand. Expert? Yes. Credible and trusted? No. Today’s meager investment in legislative information processing capacity reflects these constraints.
Lewis questions whether administrative agencies still possess the capacity to engage in reasoned behavior. Their capacity is lower than it might be, he observes, due to structural incentives and to an ideological push from the right. I sympathize greatly with this concern. The only qualification that I would add is that the most radical proposals to “deconstruct” the administrative state do not succeed, even when the ideological alignment is most favorable: the Regulatory Accountability Act of 2017 failed; the Separation of Powers Restoration Act of 2017 failed; the repeated campaign promises to eliminate entire prominent agencies, such as the EPA, Energy, or Education fail. My reading of these failures is that, save the exceptionally ideological or right-pandering, most sense the political value in an administrative state of some substantial reasoning and policy-making capacity.
The reviewers further point to several theoretical themes that deserve emphasis. Livermore observes that part of the message of the book is that supporters of the administrative state ought to be less apologetic about its democratic credentials. A theme I attempt to thread throughout is that Congress is not as “democratic” as it is often taken to be in the literature—at least in the particularities, it is often subverted by narrow concerns, without effective broader representative linkages. Noll thoughtfully observes that the book bridges normative and positive theories of administrative law, providing a positive account for the normative values traditionally associated with the field. I view this harmonization as a core outcome of the book. Walker, too, identifies a foundational theme of the project: that many features of our system that, on first contact, appear to constrain effective government, in fact enable and advance effective government. It is the apparent constraints of administrative law—the regularity of reasoning, the openness to participation and contestation, both by the public and courts—that is the distinctive margin of the administrative state and produces its great political and institutional value, empowering agencies to solve pressing public problems that would otherwise be unmet.
Matters of interpretation
As several commenters note, the theory interfaces most directly with aspects of arbitrariness review. This form of review often serves to amplify reasoning capacities. It therefore often, though of course not always, services the enlightenment values at the foundation of this project. It also often, though of course not always, promotes progressive values, either directly, or indirectly, by supporting the possibility of faith in administrative institutions. As Stack, Livermore, and Walker note, I spend less time discussing statutory interpretation.
Stack is right to see the theory’s emphasis on objectives as embracing an idea of a purposivist agency—an agency, that is, pursuing ends identified in statutes. Agencies should be given room to run, then, under this theory, and this may be most easily accomplished through forms of deference. And for much of the post-New Deal period at least, they were generally given this room, either through doctrines of deference or purposivist methods of interpretation. But that tide is turning.
The major questions doctrine (MQD) interacts with the theory in subtle ways, and had I been writing a few months later I would have spent more time on this important topic. On the one hand, Walker and Livermore correctly see the book as articulating a rationale for the MQD: it is precisely on these “major” questions where our concern about information is weakest and the apparatus of elections works best. Downgrading deference in this context may be sensible—where election work, we should put faith in them.
But much of course turns on how the doctrine is applied. The doctrine as articulated focuses on the political/economic significance of the authority asserted by the agency. If it crosses some undefined threshold of major-ness, the doctrine calls on courts to find “clear authorization” in the statute before concluding that the agency possesses the authority in question. Among the critical issues that must be resolved in application: what is major-ness? And what is clear authorization? The book’s theory suggests some guiding thought experiments on these questions. On the relevant threshold for major-ness: is delegation of authority is one we might reasonably imagine representatives being held accountable for given the applicable information problems? The response to this question will depend on the particularity, salience, and complexity of the issue. To give an initial sense of application on this margin: I would see King v. Burwell and WV v. EPA as failing that test; Brown & Williamson, Gonzales v. Oregon, and Alabama Association of Realtors v. HHS as plausibly passing it. Whether “best system of emissions reduction” includes generation shifting, for example, is a question that is hard to see voters parsing; whether the FDA has the authority to regulate cigarettes as a “drug delivery device” is one we might more easily see voters parsing. On the question of clear authorization: I would ask only whether the asserted authority is within the meaning of the statute, examining text and context. So no deference; but also no implausible level specificity to application. Brown and Williamson, for instance, would pass on this margin as the relevant tobacco products fell within the statutory definitions. Likewise, the vaccine mandate of NFIB v. OSHA would pass. However, the eviction moratorium at issue in Alabama Association of Realtors would fail. But critically, I see no concern in finding authority to confront new problems in broadly worded statutes—it is indeed important to affirm that authority for agencies find the room to run with their statutory objectives in the face of challenges that cannot be anticipated at the time of drafting. Of course, this perspective only very partially hews to the hazy lines drawn by the justices.
Matters of evidence
On the empirical side, several commenters observe that the experiments leave many questions open. I could not agree more. Yet let me first observe that the book takes a broad view of what counts as evidence. Two chapters engage in what I refer to as field harmonization: exercises that ask us whether the theory helps us understand other important debates in public law and American political institutions. The first exercise relates to a theory of administrative law; the second to the “origins” of the administrative state in the progressive era. Echoing the theme of Noll’s comments, the principal theory implicates a positive theory of administrative law that embraces values traditionally associated with the field. Likewise, the main parameters of the theory provide an interpretation for the proliferation and growing importance of agencies during the progressive era. I regard these harmonization exercises as a form of evidence that supports the theory.
The experiments themselves follow this core setup: give a person a sum of money and instructions about how to distribute it that condition on private information; then manipulate whether they must provide reasons for their choice and whether the choice is reviewed by a third party. The person in this position is meant to roughly mimic that of a bureaucrat implementing a “statute” that establishes a normativity as to possible courses of action. In brief, I find that requiring reasons increases compliance with the statutory instructions, as against their self-interest (which is to keep the money), particularly if the action is made in expectation of third-party review. A working paper using observational rather than experimental data echoes these findings.
Although the experiments support the main claims of the book, they represent only a first step in many ways. Stack points to a useful path forward: to examine the constraint that reasons provide on future (rather than contemporaneous) actions. Reynolds suggests another: introducing elements of partisanship. These both represent margins that I am pursuing in follow-up work. The objective is to push on the parameters of the original experiments to the point that reason-giving breaks down. The results in the book sit in tension with parts of the psychology literature (and with the presuppositions of strong legal realists and fellow travelers), and I am not committed to the claim that reason-giving always affords constraint or induces pro-social behavior. My hope in this follow-up line of work is to begin to map out more systematically the terrain over which reason-giving might exert more or less constraint, and, in turn, to be more or less credible.
In conclusion, I simply want to thank again the journal and editors, Bridget in particular, and the incredible interlocutors for their engagement with the project.
Jed Stiglitz is a Professor of Law at Cornell Law School.
 This is of a piece with a more evolutionary perspective on the analysis; indeed, many elected members lost their jobs. I sympathize with Alchian’s classic thesis that (in a different context) emphasized selection rather than intentional objectives as a key to appreciating actor behavior, even as we use those objectives in an institutional analysis, and understood an economic (or political) “system as an adaptive mechanism which chooses among exploratory actions generated by the adaptive pursuit of ‘success.’” Armen A. Alchian, Uncertainty, Evolution, and Economic Theory, 58 J. Pol. Econ. 211 (1950).
 I also wonder if investment in information-processing capacity would raise the same political complications as increasing member salaries. Congress might also employ various adjuncts to increase information-processing capacity; these might avoid the same self-serving appearance of spending money on oneself.
 The votes in the House on some of these measures should be viewed in light of their dim prospects in the Senate—they were expressive votes.
 I note that the doctrine is more inchoate than sometimes suggested in the literature and that there may be substantial differences remaining even within the conservative bloc. To provide one example: Roberts refers to the doctrine as requiring “clear authorization”; Gorsuch is eager to say a “clear statement” is required. This is a potentially consequential difference that speaks to the evidence each is willing to admit, with Roberts sweeping in more contextual factors.
 Another set of experiments in the project examines how the views of those on the receiving end of this transaction condition on reason-giving.