Justifying the Administrative State -- and Congress

by Michael C. Dorf

Today I have the pleasure of attending an all-day conference at Cornell Law School focusing on the forthcoming book The Reasoning State by my colleague Prof Edward ("Jed") Stiglitz. I look forward to hearing the comments of various panelists who have come from near and far. I'm commenting on two of the chapters. Here I want to preview some of my remarks on Chapter 2 -- Reasoning and Distrust: State Architecture in Advanced Societies.

The book as a whole is terrific. It is methodologically diverse, including historical materials, legal and institutional analysis, formal mathematical models, and the reports of some psychological experiments, all integrated into a coherent whole. I won't attempt to summarize the entire book here, however, both because that's too daunting a task and also because Chapter 2 is the core of the book.

In both my oral comments later today and in this preview, I'll begin with a summary of the chapter (and thus the book's core argument), and then raise some questions. Readers who want a deeper understanding should consult Stiglitz's 2018 article in the University of Pennsylvania Law Review, which sets forth an earlier version of the core argument, and/or his 2017 article in the Supreme Court Economic Review, which reports on one of his experiments. And of course they should buy the book when it's available.


The Reasoning State is a qualified defense of the administrative state. It offers a somewhat unconventional account of how the administrative state arose that simultaneously justifies the basic shape of administrative law.

Two views of the administrative state dominate the conventional narrative. In the public-regarding view, agency administration arose because of the complexity of modern life. Congress could no longer write the detailed rules necessary to govern, so it created and delegated power to expert agencies to adopt public-regarding rules.

In the alternative, more cynical, view--closely associated with the work of Morris Fiorina, John Hart Ely, and David Schoenbrod--the administrative state arose as part of a con game: Congress delegates power to agencies so it can claim credit for solving problems--"See, we passed the Clean Air Act, so now you get clean air"--while not having to take the political heat for failures--"What? You're choking on pollution. Damn that EPA for subverting our good work."

Each view has problems. The public-regarding view leads to a puzzle: If Congress lacks adequate capacity to address details, why doesn't Congress simply hire a much larger staff to write and update statutes? Why would Congress give up power to agencies that it doesn't control? And why should we assume that members of Congress care about the public good rather than their electoral prospects?The public-regarding view seems insufficiently cynical about congressional motives.

Stiglitz shares the cynics' views about motives, but he thinks that the cynical view is too cynical about the public. You can't fool all of the people all of the time and you can't even fool enough of the people enough of the time to get re-elected. The cynical view fails to address the question why the public would accept the transparently self-serving ploy that the cynics consider delegation to be.

Stiglitz thinks that delegation arose to address an information problem. At the Founding, it may have been possible for the People to figure out whether their representatives were serving the public good as the People saw it. The electorate was much smaller, allowing for much greater knowledge about the elected officials, and, crucially, life was simpler then. But with the growth in the population, expansion of the franchise, and industrialization, came opacity.

Here's an example Stiglitz uses: the People could understand that it's appropriate for the government to set rates for regulated monopolies, but they couldn't know enough to know whether their representatives were setting rates too high (perhaps due to a corrupt bargain with the regulated monopolies) or too low (perhaps to pander to populist sentiment). The People couldn't monitor their representatives' lawmaking efforts, and so the People couldn't trust their representatives.

Accepting the cynics' view, Stiglitz explains that politicians need and want to be trusted because an electorate that distrusts its elected officials will tend not to vote to re-elect them. Delegation, Stiglitz contends, does not fool the People. Delegation enables the People to trust the outputs of the administrative process and thus to trust (and therefore re-elect) the legislators who did the delegating.

How so? Here Stiglitz contrasts the legislative process with the administrative process. The latter is subject to numerous procedural constraints, including (under the APA) all of the elements of a list drawn from a description by Judge Friendly in a 1975 speech:
(1) notice of proposed action; (2) an opportunity to argue against the proposed action; (3) the right to know the basis of the action; (4) the right to have the decision based "on the record;" (5) the right to counsel; (6) the formation of a formal record; (7) a statement of reasons for the decision; (8) an open and public hearing; and (9) judicial review of agency action.
Congressional lawmaking sometimes complies with some of these requirements -- such as when Congress holds open hearings -- but the features on the list are not required of Congress, which frequently fails to satisfy any of these elements. Most crucially, given the co-equal status of the legislative and judicial branches, the courts could not through judicial review impose these elements of legislative due process on Congress.

But precisely because administrative agencies are inferior institutions, no constitutional obstacles prevent robust judicial enforcement of procedural constraints on agency lawmaking. Through a kind of jujitsu, Stiglitz turns what is frequently seen as a problem with agencies--that the Constitution does not establish them and thus the courts eye them somewhat skeptically--into a feature. Judicially enforceable procedural regularity facilitates trust in the outputs of agencies, which explains why legislators who know their electorate couldn't otherwise trust them would delegate to agencies.

So much for my summary. Now onto a couple of lines of questions.


(1) The basic picture of agencies that underwrites the possibility of congressional delegation to enhance trust is one of agencies as experts. Ordinary citizens cannot and do not know what rates should be set for railroads or the maximum parts per billion of some chemical that a power plant should be permitted to emit, so they cannot trust their elected officials to get it "right." However, they can be assured that an agency will dot its i's and cross its t's, so they do trust the agency to get it right.

Yet what assurance does the public have that the agency is trying to get it right in the sense of responding to the best available scientific evidence about how to achieve some vague policy objective such as ensuring that approved drugs are "safe" and "effective" or that grantees of radio licenses serve the "public interest"? In many circumstances, these will not be scientific questions subject to the exercise of expertise. Accordingly, it has long been common knowledge that, as Prof Peter Strauss observed in an important 1996 speech describing trends that go back at least to the 1960s, the justification for agency decision making now rests at least as much on politics -- the fact that agency heads answer ultimately to the president -- as on expertise.

For evidence, one need look no further than the facts of the one case students remember from their administrative law courses -- Chevron USA v. NRDC. There, the Reagan EPA adopted a more permissive definition of "stationary source" as used in the Clean Air Act than had its predecessors in the Carter administration. The Court's opinion contains various nods to expertise but recognizes that the science didn't change in 1981; the administration did. That was sufficient to justify the changed rule, because an agency may "properly rely upon the incumbent administration's views of wise policy to inform its judgments. While agencies are not directly accountable to the people, the Chief Executive is, and it is entirely appropriate for this political branch of the Government to make such policy choices."

It is easy to see why a citizen who voted for Reagan would trust an EPA decision justified on such grounds. It is much harder to see why citizens who did not vote for Reagan should.

Accordingly, one question I shall pose for Stiglitz is how the political justification for judicial deference to administrative agencies squares with his theory. In chapter 5, he cites the Strauss speech I've linked above, but only in the context of discussing empirical evidence that the requirement of reason giving leads to greater trust; he does not discuss the broader implications of Strauss's paper. In chapter 7 (on which I'm also commenting), Stiglitz describes the increased politicization of agency administration -- including the supplanting in importance of independent agencies by executive agencies and greater use of mechanisms of presidential control, such as OIRA and politicized appointments--as undercutting the delegating-for-trust rationale.

That seems right to me as an analytical matter, but where does it leave the theory? We are unlikely to see much in the way of a reversal of these trends. Does that imply that while Congress once delegated for trust it now delegates for some other reason? And if so, what is that other reason?

(2) Stiglitz's account is chiefly descriptive and interpretive. He provides us with a story that better fits the facts of the administrative state than does either the public-regarding view or the purely cynical view. But his account is also somewhat normative. He aims to show that the administrative state we have -- while hardly perfect -- is broadly speaking justified.* Yes, he says, the risk of agency capture and related pathologies are real, but given the procedural constraints, the outputs of agencies are less likely to reflect pure rent-seeking than are the outputs of the unconstrained legislative process.

Yet Stiglitz's normative defense raises a question: What is the point of Congress in his view? Congress cannot be trusted, so it delegates to the more trustworthy agencies, which are subject to judicial review. Maybe that makes the best of a bad constitutional design, but if so, does Stiglitz's view imply that we would be better off with a government consisting only of administrative agencies restricted by the APA and courts that apply those restrictions? If he were writing on a blank slate, would his new Constitution make room for an elected legislature at all?

I suspect Stiglitz will respond roughly along the following lines: (1) Even without a legislature to delegate power to them, agencies have a substantial democratic pedigree in light of the public's rights of participation; and (2) there is still an important role for Congress in making the top-level policy choice of setting objectives for the agency to follow.

If that's the response, then I'll concede that (1) does some work but not enough and then push back pretty hard on (2). Congress often delegates authority in ways that are so vague as to leave the policy choice to the agency -- as in the Clean Air Act addressed in Chevron. The public's limited ability to monitor Congress may indeed be sufficient to judge whether they approve of, say, an environmental policy that balances conservation against cost, or a railroad policy that requires rates to be set at a "fair and reasonable" level (an example Stiglitz helpfully provides from the Hepburn Act of 1906). But candidates for office will not be able to distinguish themselves by voting for such measures if everyone is willing to do so.

That leaves two possible escapes, both of which seem problematic. One is that our politics becomes trivialized. With the real decisions made by agencies pursuant to vague delegations on which everyone can agree, candidates distinguish themselves along policy-irrelevant dimensions, such as personality and entertainment.

Meanwhile, to the extent that there are real policy differences that can show up in the comprehensible top line of legislation, they will need to be quite stark: Criminalize or allow abortion; fight global warming or don't; have a death penalty or don't. The devil will be not in the details but in the ensuing ideological fight. Put differently, if Congress retains real power because the top line is sufficiently divisive that it matters, that's going to be because we have extreme polarization, in which the very ideas of reasonableness, fairness, and safety (all terms used in existing delegations) are themselves contested. That might well be an apt description of our times, but it is hardly a normatively attractive vision of government.

I hope that I have overlooked other possibilities and look forward to what I am sure will be an enlightening conversation.


N.B.  In the foregoing, I say that Stiglitz provides a defense of the administrative state, broadly speaking. I believe that's a fair characterization, but in other chapters Stiglitz also focuses some attention on ways in which the justification falls short. He notes how administration has increasingly shifted away from the due process model and towards less formal means that provide fewer of the sorts of procedural protections that enhance trust. He also points to other reasons why the People--and especially disadvantaged subgroups of the People--distrust government, including agencies. These are important caveats which I'm sure will generate interesting discussion.