Administrative agencies live and orchestrate work in that gritty, liminal space between textual law and the decisions that shape citizens’ lives. Because Congress can rarely, if ever, implement or enforce laws on its own, and because much of what the President does requires the work of subordinate officers, no law or policy truly exists outside of the administrative apparatus that implements it. This dependence on administrative agencies holds whether the policy is as misleadingly “simple” as a tax or subsidy, or whether it entails the purportedly more complex operations of technical decisionmaking. Congress does not launch the shuttle or the satellite, nor does it approve the drug or remove it from the market. Nor would any republic worthy of the name trust the President alone with these decisions. Congress does not collect the tax, whether imposed upon income, assets, trades, or, someday, carbon.1×1. Since, as Adam Smith taught us, the specialization of labor increases as the market grows, see Adam Smith, An Inquiry into the Nature and Causes of the Wealth of Nations bk. I, ch. III (Edwin Cannan ed., Modern Library 1937) (1776), even market-based measures such as “cap-and-trade” emissions policies require a far more expansive enforcement mechanism than their many academic fans have imagined. The benefits of the “trade” depend entirely upon the tightness of the “cap,” which is never self-enforcing. Widely touted alternatives to the national administrative state — the employment of contractors, the decentralization of policy to states and localities, and even the use of privatization — only make these policies more administratively complex, more opaque, and often subject to less independent scrutiny.
The rule of law in republican government requires an administrative state sufficiently constrained to protect liberty and sufficiently effective to convert the letter and spirit of law into living reality. Both of these aims demand organization and, sometimes, more administrative complexity. As Professors Daniel Ernst and Joanna Grisinger have demonstrated in recent works, many of the features of the administrative state that render it so cumbersome — internal clearance, multimember commissions, transparency practices and freedom-of-information laws, notice-and-comment rulemaking, and cost-benefit analysis — also stand as its most “liberal” features, and many of them were developed within agencies long before being codified in statute or executive order.2×2. “When Congress adopted the Administrative Procedure Act in 1946, it was less a landmark ‘bill of rights’ for the administrative state than a codification of the agencies’ own best practices for formal adjudication.” Daniel R. Ernst, Tocqueville’s Nightmare: The Administrative State Emerges in America, 1900–1940, at 137 (2014); accord Joanna L. Grisinger, The Unwieldy American State 60–61 (2012). On internal clearance, see Jennifer Nou, Intra-Agency Coordination, 129 Harv. L. Rev. 421, 468–71 (2015).
The friction between organizational reality and administrative law forms the subject of Professor Jennifer Nou’s thoughtful recent article.3×3. Nou, supra note 2. Nou focuses on the internal organization of agencies, which she argues is undertheorized, and hypothesizes that agency heads restructure their agencies with the primary objective of reducing information-processing costs.4×4. Id. at 427–30. Though her article offers valuable lessons, it neglects other dominant motives for agency reorganization — namely, cosmetic changes that seek to placate agency observers, and changes that serve purely ideological goals.
Nou starts by observing that administrative law scholars have paid far less attention to administrative law made within agencies (especially by appointed agency heads) than they have paid to administrative law imposed upon agencies.5×5. Id. at 427–29. Her critique of administrative law scholarship applies as easily to political science and political economy scholarship on administrative structures, in which far more ink has been spilt on the actions of Congress and the President than on agencies themselves. Surely a critical reason for this imbalance lies in the vast authority and leverage that Congress, the President, and the courts have over agency behavior. Another (less satisfying) reason for the tilt concerns the costs of research and our collective willingness to surmount them. The study of agency decisionmaking, especially intra-agency coordination, is probably harder work than the study of what Congress does (through statutes, budgets, or hearings), or what the President does (through executive orders or appointments), or what courts do (through judicial review) to control agency decisionmaking. Put differently, the commonly theorized information asymmetries that prevail between political “principals” and administrative “agents” constrain scholars no less than they constrain politicians and judges.
Nou argues that some features of coordination within agencies will always be left to their appointed chiefs, whether a cabinet-level secretary, a bureau chief, or a center director.6×6. See id. at 432–34. These officials face questions of structural design in the presence of the uncertainty absorption problem — the fact that what is communicated up the chain is not the raw data but more progressively refined inferences from that data. In a partial departure from principal-agent theory, Nou adopts a team production model of administrative structure.7×7. Id. at 439–40. The burden of uncertainty and the imperative of efficient organizational design induce agency leaders to structure their agencies — by means of centralization, specialization, separation, standardization, or procedures such as internal clearance or priority setting — so as to “lower the costs of internal information processing.”8×8. Id. at 451. On the five mechanisms of coordination, consult id. at 451–73.
Nou’s article offers some valuable lessons. She argues more clearly than previous scholars that agency design choices are almost always incomplete, both by choice and by necessity. It may be infeasible to write a contract for agency structure that incorporates every future contingency. Even if feasible, building in this level of detail may be too costly (“incomplete contracts”), or political decisionmakers may prefer to rely upon trust, reputation, and informal agreements (“relational contracts”). A considerable literature in political science suggests that Congress and the President delegate far more often to existing agencies with which their committees and offices already have a relationship than they delegate to new agencies created at the same time that policy is delegated.9×9. Two representative publications — many more have followed — include David Epstein & Sharyn O’Halloran, Delegating Powers (1999); and Daniel P. Carpenter, The Forging of Bureaucratic Autonomy (2001). An important recent example comes in Congress’s delegation of authority over tobacco products to the FDA in the Family Smoking Prevention and Tobacco Control Act of 2009 (FSPTCA), Pub. L. No. 111-31, 123 Stat. 1776 (codified at 21 U.S.C. §§ 387–387u (2012)). This delegation was not only a jurisdictional reliance upon previous arrangements, but also a procedural reliance, insofar as the FSPTCA’s scheme for regulation of tobacco products borrowed heavily from the existing “drug and device” model. In the drafting process, Senator Edward Kennedy’s staff had intimate knowledge of the FDA and deep familiarity with the drug and device model, and many terms in the FSPTCA were left vague precisely because Congress expected the FDA to fill out their meaning as it had for drugs and devices, see id. § 2.
A second key insight comes in the application of team production models to administrative procedure, with assignments made by agency leaders and work informed by standard operating procedures.10×10. Nou, supra note 2, at 467–68. Indeed, Nou might have applied the team production concept more explicitly to rulemaking, insofar as agencies have perceived their rule-drafting groups explicitly as “teams” since at least the 1970s.11×11. In the 1970s, the FDA had a multidimensional rule-writing team in its Bureau of Drugs, which drafted critical regulations covering subjects ranging from clinical trial procedures to the “bioequivalence” of generic drugs and the pioneer molecules they were based upon. See Daniel Carpenter, Reputation and Power 357–62 (2010); Daniel Carpenter & Dominique A. Tobbell, Bioequivalence: The Regulatory Career of a Pharmaceutical Concept, 85 Bull. Hist. Med. 93, 126–27 (2011). Just as with team production, moreover, there are inter-individual dependencies in rule drafting. Any team member’s marginal product is impossible to separate from that of others, such that if the economist fails to do her work, or the chemist fails to do his, the rule does not get drafted (or later gets rejected), and the lawyer’s marginal product drops to zero no matter how many hours she puts in.
Finally, Nou’s typology of intra-agency coordination mechanisms offers a helpful place to start for lawyers and scholars studying this question in the future. These conceptual guides would be as useful in a public management, public policy, or political science class as they would be in an administrative law course.
Exactly how coordination works in agencies and exactly for which reasons it is undertaken are complex matters. While there are hints of this complexity in Nou’s article, her treatment and interpretation of cases and evidence contribute more to the debit than to the credit side of the ledger. Her account interprets structural changes as undertaken to enhance the sifting function of the administrative structure, which is its ability to separate signal from noise (a “sieve”). Yet scholars in other disciplines who have empirically examined intra-agency coordination repeatedly come to the conclusion that structural change across and within agencies often serves ceremonial ends (a “show”) or follows from the agency head’s partisan or ideological policy preferences (a “shove”) instead of a mere preference for one type of information over another.12×12. See, e.g., Donald P. Moynihan, The Dynamics of Performance Management 13–14, 13 tbl.1.1 (2008) (arguing that performance management reforms in the 1990s and early 2000s were instituted for symbolic goals, but that these goals sometimes served efficiency gains as well); Jonathan Bendor, Serge Taylor & Roland Van Gaalen, Stacking the Deck: Bureaucratic Missions and Policy Design, 81 Am. Pol. Sci. Rev. 873, 873 (1987) (theorizing that agency leaders reshape policy structures to favor their preferred outcomes, and that political officials can respond with their own reshaping); James G. March & Johan P. Olsen, Organizing Political Life: What Administrative Reorganization Tells Us About Government, 77 Am. Pol. Sci. Rev. 281, 289–91 (1983) (arguing that most comprehensive reorganization efforts are ceremonial). Even on Nou’s terms of interpretation, EPA Administrator Anne Gorsuch’s 1981 dispersing of enforcement functions among several divisions13×13. Nou, supra note 2, at 458–59. seems to be a shove, and political scientists have quantified the nature of other reforms that decimated environmental enforcement in the 1980s.14×14. See B. Dan Wood & Richard W. Waterman, The Dynamics of Political Control of the Bureaucracy, 85 Am. Pol. Sci. Rev. 801, 818–24 (1991). One might also worry that agency heads could use reorganization to quiet or punish whistleblowers or dissident decisionmakers. Similarly, organizational changes under Chairwoman Jane Alexander at the NEA15×15. Nou, supra note 2, at 431–32. seem much more like legitimation than information processing. Nou occasionally points to cases where these forces were at work, but repeatedly falls back to the organizational economics model of change as serving efficiency purposes, both as explanation and as normative justification for intra-agency coordination.
The differences are important because Nou presents her theory as predictive, in part. “Holding all else constant,” she writes, “increases in exogenous levels of uncertainty will prompt intra-agency reforms that promote the more efficient transmission of privileged information to the agency head.”16×16. Id. at 430. A theoretical perspective that neglects the behavioral forces driving many organizational changes risks foundering on the shoals of what social scientists call observational equivalence.17×17. Observational equivalence exists when two theories predict the same outcome in every circumstance. Thus, to the observer, the theories are functionally equivalent with no way of saying which is right and which is wrong. Yet the differences between the sieve, shove, and show models of structural reform are also important because judicial interpretation of agency coordination policies as serving informational ends should, in many cases, proceed quite differently from judicial interpretation of policies that serve ideological or factional purposes, or that are undertaken to gain legitimacy or trust from certain constituencies. If an agency leader who is deeply skeptical of the relevant statutory regime and its purposes proposes an intra-agency coordination scheme that squelches administrative behavior that in turn would have substantially advanced statutory intent — EPA Administrator Gorsuch’s changes18×18. Nou, supra note 2, at 458–59. might well compose one example — then any justification that invokes informational efficiencies should be greeted with skepticism, perhaps even “hard look” hostility. So too, while reputational incentives can often lead an agency to serve consensual public purposes, if the audience whose approval is sought in intra-agency coordination is that of a particularly powerful special interest (a coalition of systemically important institutions in finance, for example), then those who oversee the agency should ask whether the restructuring serves to further regulatory capture.19×19. See James Kwak, Cultural Capture and the Financial Crisis, in Preventing Regulatory Capture 71 (Daniel Carpenter & David A. Moss eds., 2014). On the observational equivalence of efficiency and capture explanations, consult Daniel Carpenter, Detecting and Measuring Capture, in Preventing Regulatory Capture, supra, at 57.
To this argument, Nou might respond that judges are not well equipped to assess organizational changes outside their competence. Yet if judges commonly call upon a range of experts in other areas of decisionmaking, they could easily call upon accumulated expertise in public administration and management in cases regarding intra-agency coordination. Beyond this, the review of agency organization requires far less technical expertise than review of the substance of regulations, which courts do routinely under the Administrative Procedure Act section 706(2)(A).20×20. Pub. L. No. 89-554, § 706(2)(A), 80 Stat. 378, 393 (codified at 5 U.S.C. § 706(2)(A) (2012)). Additionally, courts are already expected to ferret out pretextual motives in a variety of contexts, such as review for discriminatory intent under the Equal Protection Clause. And even if judges were to exercise “hard look” sparingly in questions of intra-agency coordination — if they were to question, remand for reconsideration, or invalidate only a tiny portion of the most inappropriate and statute-undermining organizational changes — they would nonetheless send strong signals about the long-run feasibility of the most strong-armed changes to agencies.21×21. Relatedly, Professors B. Dan Wood and Richard Waterman find a large and statistically differentiable increase in EPA hazardous waste inspections after Anne Gorsuch was held in contempt by Congress (she resigned three months after the citation). Wood & Waterman, supra note 14, at 820. While a congressional contempt citation is different from “hard look” review, they both concern internal agency governance (Gorsuch was cited for contempt for refusing to turn over papers relating to internal decisionmaking), see id. at 819.
In the hyper-partisan world of the twenty-first century, where special interests observe every word and deed of the administrative state, we should first turn to political and ideological goals (the “shove”) to interpret moves of intra-agency coordination. This is true despite the fact that, in particular and lasting constellations of U.S. political economy, bureaucratic autonomy has been possible. It is also all the more true if and when we acknowledge, as Nou does, that appointed agency leaders take the principal actions in intra-agency coordination. If anything, Nou seems to have misidentified the proper site of intra-agency coordination that leads to informational efficiency. Intra-agency coordination proposed and accomplished by higher-level civil servants may better serve ends of informational efficiency than that undertaken by appointed agency leaders. The case of Administrator Anne Gorsuch, in which Nou seems to acknowledge implicitly that informational efficiency was not the goal of organizational restructuring, suffices to buttress this point.22×22. There is a different critical point to be made here, namely that whereas Nou locates decisionmaking power in appointed leaders, previous analysts, including Professor Herbert Kaufman and myself, have tended to examine career-based bureau chiefs. See, e.g., Herbert Kaufman, The Administrative Behavior of Federal Bureau Chiefs (1981); Carpenter, supra note 9, ch. 1. This difference amounts more to an analytic focus than to a factual dispute, yet it again points to the necessity of empiricism. Appointments and structural changes that are announced above may often have risen through the ranks, emerging from career officials lower in status. So too, moves taken by careerist officials may often have originated from higher-ranking appointed officials. Only careful, often historical, research will allow the law professor or the judge to discern these origins.
Nou’s sometimes casual, sometimes deliberate reliance upon the logic of organizational efficiency leads to some misguided readings and interpretations. Nou takes at face value those laws and changes that quite plausibly have ideological or ceremonial facets, if not intentions. Nou makes much hay of the fact that agency leaders are called before Congress to defend their agency’s behavior,23×23. Nou, supra note 2, at 478–79. but Professor Morris Fiorina decades ago noted that these interactions are stimulated by congressional blame avoidance.24×24. See Morris P. Fiorina, Congress: Keystone of the Washington Establishment 67–68 (2d ed. 1989). So too, the origins of the Information Quality Act of 200025×25. Pub. L. No. 106-554, § 515, 114 Stat. 2763A-153, 2763A-153 to 2763A-154, reprinted in 44 U.S.C. § 3516 note (2012) (Policy and Procedural Guidelines) (cited in Nou, supra note 2, at 448 n.158). lay not in a deliberative consideration of agency structure and informational efficiency, but in an appropriations rider. The most aggressive parties to first use the act included think tanks and private associations aligned with business and deregulatory causes. Two such early parties were the Competitive Enterprise Institute and the Salt Institute.26×26. On the odd procedural history of this Act and its users, consult Chris Mooney, Interrogations, Bos. Globe (Aug. 28, 2005), http://www.boston.com/news/globe/ideas/articles/2005/08/28/interrogations. This is not to say that these special interests did not self-consciously detect issues with data quality in the lawsuits they brought. It is to say that the very questions of informational efficiency and data quality were themselves partisan issues and that some groups plausibly saw data quality restrictions as means for defanging the regulatory state.
What does this mean for Nou’s intriguing suggestion that Chevron-like deference is due to agency heads’ intra-agency coordination choices? Here I largely applaud Nou for her move — no one has apparently considered the potential applicability of Chevron to intra-agency coordination decisions before — but think that she has somewhat oversimplified the problem. I think she has a case for Chevron deference, in the sense that judges ought not, as a matter of course, to interfere in these decisions. Yet such deference would have to rest on either (a) the inability of judges to interpose their discretion in these situations or (b) the empirical accuracy of the informational efficiency rationale for internal restructuring. Her argument that agency internal governance would largely benefit from greater transparency is compelling. When she notes that prodigious litigation might result from transparency policies,27×27. Nou, supra note 2, at 481–82. I think it is precisely here that Chevron-like deference is most called for, so that agencies are not tacitly or explicitly penalized for openness.
I would call instead for a qualified Chevron approach. Courts ought to be on the lookout for cases where either (a) clear statutory regime–undermining organizational changes are implemented for the detectable intent of changing policy without changing statute, and without a correspondingly plausible offset in informational efficiency, or (b) agencies are changing structure for legitimation, but in a way that serves one special interest audience at the expense of another. The NEA reforms under Chairwoman Alexander discussed by Nou would, I surmise, easily pass such a qualified Chevron approach.28×28. See id. at 431–32. The alterations imposed by EPA Administrator Anne Gorsuch might well not. While protections for whistleblowers comprise a separate area of law, organizational shifts that intend to squelch or punish agency dissidents should also receive scrutiny.
Agencies live among audiences — constituencies, overseers, and actors that observe their actions only imperfectly and often from afar. The opaqueness of these observations means that agency actions are always subject to plural interpretations. Chairwoman Jane Alexander’s restructuring of the NEA in the context of the controversial Robert Mapplethorpe art exhibit surely stands as a manifold gesture to the agency’s critics. So too, in the interpretation of much empirical research, most agency reorganizations have imposed displays of conformity to social and ideological expectations.29×29. See March & Olsen, supra note 12, at 287. Interpretations of intra-agency coordination as “show” or as “shove” also have their weaknesses. They can lead to overly skeptical readings of organizational change. Yet a rather large body of literature has drawn out the test implications of these theories, abrading them against those of other accounts. Absent that kind of careful empirical adjudication, law professors will not be alone in misreading intra-agency coordination; judges will as well.
* Freed Professor of Government, Faculty of Arts and Sciences, and Director of Social Sciences, Radcliffe Institute for Advanced Study, Harvard University.