The traditional portrait of the administrative state often features the politically-appointed agency head at its center: the Administrator of the Environmental Protection Agency, for instance, or the Secretary of the Department of Labor. This picture of bureaucratic power, however, is incomplete. For much of that power is, in fact, subdelegated within the agency. The implication is that decision rights are often exercised not by statutory delegates, but rather by lower-level officials and tenure-protected career staff. The purpose of this work is to bring these background actors—like the Securities and Exchange Commission’s Director of Enforcement—squarely to the foreground.

In doing so, this Essay develops a positive theory for how and why agency heads subdelegate their power and analyzes the resulting normative implications. The analysis draws on a rich social science literature to argue that delegations within agencies are best understood as credible commitment devices through which agency heads motivate better-informed but potentially biased subordinates. This decision to commit, however, is itself subject to the internal transaction costs of reviewing subordinate recommendations. These dynamics, in turn, suggest a role for courts to maximize high-quality information within agencies by taking credibility into account in their appointments and removal jurisprudence; deferring to subdelegations promulgated through transparent, deliberative procedures; and providing clearer ex ante guidance as to when internal subdelegations will be judicially enforced. At the same time, political actors (not courts) should police the counter­vailing concern that subdelegation can also be used as a tool of parti­san entrenchment.


    1. Advisors Versus Authorities
    2. Information Production and Transmission
    1. Institutionalizing Commitment
    2. Enhancing Commitment
    3. Applications
    1. Courts
    2. Political Actors



Administrative law’s central anxiety is the exercise of delegated power. This angst, more often than not, is directed at delegations from Congress to an administrative agency—spawning countless legal conun­drums: which institutions should have interpretive primacy, 1 See, e.g., Henry P. Monaghan, Marbury and the Administrative State, 83 Colum. L. Rev. 1, 5 (1983) (“Deference, to be meaningful, imports agency displacement of what might have been the judicial view res nova—in short, administrative displacement of judicial judgment.”); Cass R. Sunstein & Adrian Vermeule, Interpretation and Institutions, 101 Mich. L. Rev. 885, 925 (2003) (“Should courts decide legal questions on their own, or should they give some weight to the views of the relevant agency?”). whether the statutory delegate’s identity matters, 2 See, e.g., Elena Kagan, Presidential Administration, 114 Harv. L. Rev. 2245, 2326–31 (2001) (arguing for an interpretive principle “that when Congress delegates to an executive official, it in some necessary and obvious sense also delegates to the President”); Kevin M. Stack, The President’s Statutory Powers to Administer the Laws, 106 Colum. L. Rev. 263, 267 (2006) (arguing “as a matter of statutory construction the President has directive authority—that is, the power to act directly under the statute or to bind the discretion of lower level officials—only when the statute expressly grants power to the President in name”). and whether the nondelegation doctrine should be revived. 3 See, e.g., Matthew D. Adler, Judicial Restraint in the Administrative State: Beyond the Countermajoritarian Difficulty, 145 U. Pa. L. Rev. 759, 839–44 (1997) (considering and rejecting the revival of a “strong or even a weak” nondelegation doctrine); Eric A. Posner & Adrian Vermeule, Interring the Nondelegation Doctrine, 69 U. Chi. L. Rev. 1721, 1722–23 (2002) (arguing “there just is no constitutional nondelegation rule, nor has there ever been”). These normative debates, in turn, also have positive counterparts. Decades of interdisciplinary scholarship ask why and under what conditions legislatures delegate in the first place. Promi­nent theories, for example, posit that Congress delegates to agencies to exploit their superior expertise, 4 See, e.g., Keith Krehbiel, Information and Legislative Organization (1991) (explaining legislative organization as a means through which legislators gain information and formulate policy amidst uncertainty); Jonathan Bendor & Adam Meirowitz, Spatial Models of Delegation, 98 Am. Pol. Sci. Rev. 293, 294 (2004) (noting one rationale for delegating is to gain superior expertise); J. Bendor, A. Glazer & T. Hammond, Theories of Delegation, 4 Ann. Rev. Pol. Sci. 235, 264 (2001) [hereinafter Bendor et al., Theories of Delegation] (observing “it is often rational for an uninformed principal to delegate authority to an informed agent; it is rather obvious that tapping the agent’s greater expertise can provide gains”). reduce policymaking transaction costs, 5 See David Epstein & Sharyn O’Halloran, Delegating Powers: A Transaction Cost Politics Approach to Policy Making Under Separate Powers 7 (1999) (arguing that legislative decisions to delegate to the executive reflect a tradeoff between the “internal policy production costs of the committee system” and the “external costs of delegation”). or otherwise shift political blame. 6 See Morris P. Fiorina, Congress: Keystone of the Washington Establishment 44–47 (2d ed. 1989) (“The bureaucracy serves as a convenient lightning rod for public frustration and a convenient whipping boy for congressmen. But so long as the bureaucracy accommodates congressmen, the latter will oblige with . . . grants of authority.”).

From these perspectives, politically-appointed agency heads—say, the Administrator of the Environmental Protection Agency (EPA) or the Secretary of Health and Human Services (HHS)—are the main arbiters of delegated authority. In reality, however, much of that power is subdel­egated within the agency. 7 See infra notes 22–34 and accompanying text. Agency heads, that is, take authority granted from Congress or the President and further redelegate it to their subordinates. As a result, tenure-protected career staff and lower-level political officials often make decisions initially granted to their superiors. 8 See infra notes 22–34 and accompanying text. In this manner, characters less familiar to the administrative law litera­ture now come to the foreground: the EPA’s Chief Financial Officer, 9 See Memorandum from Gina McCarthy, Adm’r of EPA, on request to Revise Delegation 1-16 Relating the Charge of Biennial User Fees to Maryann Froehlich, Acting Chief Fin. Officer, EPA (Dec. 2, 2013) (on file with the Columbia Law Review). for instance, or HHS’s Director of the Office for Civil Rights. 10 See, e.g., Office for Civil Rights; Statement of Delegation of Authority, 65 Fed. Reg. 82,381 (Dec. 28, 2000).

Delegation in some form, of course, is a necessity in large organi­zations like bureaucracies. Agency heads have limited time. They have finite resources. They must thus rely on their staff for discrete tasks. These assignments, however, are usually accompanied by structures and processes to coordinate institutional decisionmaking. 11 See Kenneth Culp Davis, Institutional Administrative Decisions, 48 Colum. L. Rev. 173, 173 (1948) (“An institutional decision of an administrative agency is a decision made by an organization and not by an individual or solely by agency heads.”); Jennifer Nou, Intra-Agency Coordination, 129 Harv. L. Rev. 421, 429–30 (2015) (arguing that agency heads “choose intra-agency coordination mechanisms to facilitate the production of information” to enable decisionmaking). In other words, even if staff members or other subordinates make initial recommenda­tions, their decisions usually undergo intra-agency review before becom­ing the “agency’s” decision. As a result, agency heads often remain the final decisionmaker. By contrast, when the EPA Administrator formally subdelegates her authority, she allocates the decision right to someone else within the agency. Such decision rights are usually accompanied by signature authority—literally, the authority to affix one’s signature and sign off on an agency action without higher-level oversight. 12 See David J. Barron & Elena Kagan, Chevron’s Nondelegation Doctrine, 2001 Sup. Ct. Rev. 201, 238 (“Authorship is a familiar concept in agency practice; indeed, agencies today are admirably (if surprisingly) punctilious about this feature of their interpretive rulings and other actions.”).

The resulting stakes can be high. Consider a particularly dramatic controversy over population measures in the 2000 U.S. Census. The 1990 census had relied primarily on mail-in forms and thereby “missed record numbers of people that had been traditionally hard to count, mainly members of ethnic and racial minorities.” 13 Steven A. Holmes, In a First, 2000 Census Is to Use Sampling, N.Y. Times (Feb. 29, 1996), (on file with the Columbia Law Review). In response, President Bill Clinton’s Secretary of Commerce proposed using statistical sampling to adjust the 2000 census count. 14 Id. However, House Republicans sued to block the use of this method, which was expected to benefit Democrats in the congressional-reapportionment process. 15 Barbara Vobejda, Republicans Sue to Prevent Sampling, Wash. Post (Feb. 21, 1998), [http://].
The Supreme Court sided with the House Republicans, holding that the Census Act prohibited the practice. 16 Dep’t of Commerce v. U.S. House of Representatives, 525 U.S. 316, 343 (1999) (“[T]he Census Act prohibits the proposed uses of statistical sampling in calculating the population for purposes of apportionment.”). Nonetheless, the Clinton Administration maintained that pop­ulation counts derived from sampling could still be used for redistricting and determining the allocation of federal funds. 17 See Juliet Eilperin & Barbara Vobejda, Census Battle Decelerates in Congress, Wash. Post (Apr. 15, 1999),
house15.htm [].

On June 20, 2000, the Secretary of Commerce proposed a rule subdelegating the authority to the presidentially-appointed and Senate-confirmed Director of the Census Bureau to decide whether to use statistical sampling. 18 Report of Tabulations of Population to States and Localities Pursuant to 13 U.S.C. 141(c) and Availability of Other Population Information, 65 Fed. Reg. 38,370 (June 20, 2000) (codified at 15 C.F.R. pt. 101); see also 13 U.S.C. § 21(a)(1) (2012) (“The Bureau shall be headed by a Director of the Census, appointed by the President, by and with the advice and consent of the Senate, without regard to political affiliation.”). The Director’s authority would be “final.” 19 Report of Tabulations of Population to States and Localities Pursuant to 13 U.S.C. 141(c) and Availability of Other Population Information, 65 Fed. Reg. at 38,370. Otherwise, “[r]eview of the Director’s decision by the Secretary of Commerce would at a minimum create the appearance that considera­tions other than those relating to statistical science were being taken into account.” 20 Id. Representative Dan Miller, the Republican Chairman of the Subcommittee on the Census, questioned the validity of the proposed rule in a series of letters to various members of the Department of Commerce and the Census Bureau, arguing that “there is a serious difference between delegating . . . to a subordinate, and divesting oneself of final review and reversal power”—the latter being impermissible under the Census Act. 21 Letter from Dan Miller, Chairman, House Subcomm. on the Census, et al., to John H. Thompson, Assoc. Dir. for Decennial Census, Bureau of the Census (Aug. 4, 2000), [].

Agency subdelegation of this nature is a more pervasive phe­nomenon than commonly recognized, sometimes even by agency heads themselves. As one former Securities and Exchange Commission (SEC) member marveled, “[f]rom time to time, you might read in a newspaper about a ‘Commission action,’ and you will have no idea what it is about.” 22 Luis A. Aguilar, Commissioner Aguilar’s (Hopefully) Helpful Tips for New SEC Commissioners, U.S. Sec. & Exch. Comm’n (Nov. 30, 2015),
statement/helpful-tips-for-new-sec-commissioners.html [].
Often it will turn out, however, that “the staff ha[s] taken action pursuant to the more than 376 separate rules where the Commis­sion previously granted delegated authority to the SEC staff.” 23 Id. These SEC subdelegations range in recipient and scope. For example, the Commission’s internal Director of the Division of Trading and Markets is authorized to act on thousands of rule filings from self-regulatory organ­izations, including various exchanges and clearing agencies. 24 17 C.F.R. § 200.30-3(a)(12) (2016). The following examples in notes 22–28 are all pointed out by former SEC Commissioner Luis Aguilar. Aguilar, supra note 22. The SEC has also subdelegated authority to the Director of the Division of Trading and Markets to grant exemptions from important Exchange Act Rules, 25 17 C.F.R. § 200.30-3(a)(7)(ii). as well as from rules regarding short sales. 26 Id. § 200.30-3(a)(15). The Director of the Division of Corporation Finance is authorized to grant or deny issuer waivers. 27 Id. § 200.30-1(a)(10). The Commission has further delegated authority to the Chief Accountant to temporarily suspend rules issued by the Public Company Accounting Oversight Board. 28 Id. § 200.30-11(b)(5).

And the SEC is hardly alone. At one point in time, the EPA recorded over 500 subdelegations by its own count, roughly half within its headquarters and the rest distributed among its regional offices. 29 See U.S. EPA, EPA-202-F-94-002, Delegations of Authority—What Managers Need to Know 1 (1994) [hereinafter EPA, Delegations of Authority]. The Federal Communications Commission (FCC) devotes significant parts of the Code of Federal Regulations (CFR)to hundreds of delegated author­ities. 30 See 47 C.F.R. § 0.201-0.392 (2015). Other agencies like the Consumer Product Safety Commission similarly delegate authority but memorialize it only in internal agency documents. 31 See 16 C.F.R. § 1000.11 (2016) (“Delegations are documented in the Commission’s Directives System.”). Subdelegations are also the subject of recent high-profile litigation. Faced with quorum-killing vacancies, the Election Assistance Commission (EAC) delegated some responsibilities to its Executive Director. 32 See Kobach v. U.S. Election Assistance Comm’n, 772 F.3d 1183, 1190–92 (10th Cir. 2014), cert. denied, 135 S. Ct. 2891 (2015). These tasks included the duty to maintain a federal voter registration form, which the Executive Director recently refused to modify to include state-specific proof-of-citizenship requirements. See U.S. Election Assistance Comm’n, The Roles and Responsibilities of the Commissioners and Executive Director of the U.S. Election Assistance Commission,
Page/348.pdf [] (last visited Oct. 19, 2016).
In the ensuing litigation, the Tenth Circuit upheld the actions. 33 Kobach, 772 F.3d at 1199. Subdelegations are likely to become even more common should legisla­tive gridlock result in further vacancies. 34 See Abbe R. Gluck, Anne Joseph O’Connell & Rosa Po, Unorthodox Lawmaking, Unorthodox Rulemaking, 115 Colum. L. Rev. 1789, 1828–30 (2015) (discussing the vari­ous effects of legislative gridlock); see also Anne Joseph O’Connell, Vacant Offices: Delays in Staffing Top Agency Positions, 82 S. Cal. L. Rev. 913, 966–68 (2009) [hereinafter O’Connell, Vacant Offices] (outlining a timeline of delays in nominations and confirma­tions between 1987 and 2005).

In the real world, subdelegation matters. Yet little is known about this important bureaucratic practice: Under what circumstances do agency heads delegate to their subordinates and why? Which powers are subdelegated, and to whom? What explains variations in these dynamics, if any, across different agencies? At first glance, one might think the positive story just mirrors that of delegation one level up. Perhaps bureau­crats delegate for the same reasons that legislators do. There are good reasons to think, however, that this instinct is misplaced given their diverging goals, cultures, and norms. 35 First, administrative agencies operate according to disparate norms and with a culture distinct from that of Congress. See generally James Q. Wilson, Bureaucracy: What Government Agencies Do and Why They Do It, at xvii (1989) (exploring why “government agencies—bureaucracies—behave as they do”). While agencies are grounded in hierarchy and professionalism, Congress features more horizontal politicking. Id. Agency heads also live in distinct political ecosystems. Agencies are monitored by the President, congres­sional committees, and interest groups possessing different means of control, while legislators are subject to the vagaries of the ballot box. See R. Douglas Arnold, The Logic of Congressional Action 5 (1990).
The respective goals of agency heads and legislators also often differ. While agency heads are tempted by the revolving door, re-election is one of the primary goals of legislators. See id. (arguing that reelection is the “dominant goal” of Congress members); Paul J. Quirk, Industry Influence in Federal Regulatory Agencies 143–45 (1981). Finally, an important insight of the theory of the firm is that transaction costs within institutions (like agencies) differ from those between institutions (like Congress and agencies). As a result, one would expect transaction-cost-reducing measures like delegation to differ in the two contexts.
Thus, while social scientists and legal scholars have long focused on delegation to agencies, the focus here is on delegation within agencies.

Subdelegation, in turn, complicates many longstanding debates in administrative law. Perhaps most obviously, it adds another dimension to separation-of-powers disputes, which emphasize the ways in which the legislature and executive battle to impose their respective preferences. 36 See Epstein & O’Halloran, supra note 5, at 7–10 (advancing a “transaction cost politics” approach to understanding when and why Congress cedes power to the executive branch); Daryl J. Levinson & Richard H. Pildes, Separation of Parties, Not Powers, 119 Harv. L. Rev. 2312, 2341–42 (2006) (“We should expect Congress to be considerably less willing to delegate policymaking discretion to the executive branch when the policy preferences of the two branches diverge.”). The fact that presidential appointees can subdelegate their authority to sympathetic subordinates is a less-recognized means of partisan entrench­ment. Relatedly, defenders of the “unitary executive” have celebrated the myriad ways in which Presidents have centralized and consolidated control over the bureaucracy, especially through their appointments. 37 See, e.g., Steven G. Calabresi & Christopher S. Yoo, The Unitary Executive: Presidential Power from Washington to Bush 10–13 (2008) (describing various “[d]isputes over presidential power to control executive branch subordinates”); David J. Barron, From Takeover to Merger: Reforming Administrative Law in an Age of Agency Politicization, 76 Geo. Wash. L. Rev. 1095, 1096 (2008) (evaluating the trend of Presidents using appoint­ments to control agencies). Subdelegations to career civil servants, however, weaken this mechanism of executive power. When administrations turn over, subdelegations remain in place until and unless they are repealed. 38 See Nina A. Mendelson, Agency Burrowing: Entrenching Policies and Personnel Before a New President Arrives, 78 N.Y.U. L. Rev. 557, 610–16 (2003) [hereinafter Mendelson, Agency Burrowing].

Administrative law scholarship has also increasingly turned from the external separation of powers (between the President, Congress, and courts) to the internal separation of powersthat is, the counterbalances offered by various agency actors. 39 See Neal Kumar Katyal, Internal Separation of Powers: Checking Today’s Most Dangerous Branch from Within, 115 Yale L.J. 2314 (2006) (discussing intra-executive checks and balances in foreign policymaking); Gillian E. Metzger, The Interdependent Relationship Between Internal and External Separation of Powers, 59 Emory L.J. 423, 425 (2009) (noting the importance of considering both “internal constraints” and the “crucial relationship between internal and external checks on the Executive Branch”); Jon D. Michaels, An Enduring, Evolving Separation of Powers, 115 Colum. L. Rev. 515 (2015) [hereinafter Michaels, Enduring, Evolving] (exploring “a secondary, subconstitutional separation of powers that triangulates administrative power among politically appointed agency leaders, an independent civil service, and a vibrant civil society”); Jon D. Michaels, Of Constitutional Custodians and Regulatory Rivals: An Account of the Old and New Separation of Powers, 91 N.Y.U. L. Rev. 227, 235 (2016) (discussing the relationship between administrative separation of powers and external constitutional actors). This literature often presents career staff as nonpartisan keepers of professional norms. 40 See, e.g., Michaels, Enduring, Evolving, supra note 39, at 534 (describing career staff as “subconstitutional, rivalrous counterweights” that “constrain the political leadership atop administrative agencies”). Considering such staff members as recipients of delegated authority with their own prefer­ences and biases, however, calls into question the extent to which they facilitate, rather than buffer, fights between their political principals. In this view, civil servants can amplify, as opposed to mitigate, potential abuses of power. 41 See infra notes 282–285 and accompanying text. Finally, legal scholars have also disputed the extent to which the administrative state is “ossified”—that is, rendered impervious to change by increasing legal requirements. 42 See Jason Webb Yackee & Susan Webb Yackee, Testing the Ossification Thesis: An Empirical Examination of Federal Regulatory Volume and Speed, 1950–1990, 80 Geo. Wash. L. Rev. 1414, 1439 (2012) (“The ossification literature predicts . . . that the accretion of the various procedural constraints imposed by the courts, the White House, and Congress on bureaucratic autonomy has severely impacted the ability of federal agencies to fulfill their § 553 notice and comment rule-writing obligations.”); see also Stephen M. Johnson, Ossification’s Demise? An Empirical Analysis of EPA Rulemaking from 2001–2005, 38 Envtl. L. 767, 782 (2008); Thomas O. McGarity, Some Thoughts on “Deossifying” the Rulemaking Process, 41 Duke L.J. 1385, 1385–86 (1992); Jason Webb Yackee & Susan Webb Yackee, Administrative Procedures and Bureaucratic Performance: Is Federal Rule-Making “Ossified”?, 20 J. Pub. Admin. Res. & Theory 261, 267–68 (2006).
The fear of ossification alone has motivated many to propose reforms like more negotiated rulemaking, reduced judicial review, or the repeal of regulatory analysis requirements that could expedite the rulemaking process. See Thomas O. McGarity, Reinventing Rationality: The Role of Regulatory Analysis in the Federal Bureaucracy 130–31 (1991); Frank B. Cross, Pragmatic Pathologies of Judicial Review of Administrative Rulemaking, 78 N.C. L. Rev. 1013, 1014 (2000); Jody Freeman, Collaborative Governance in the Administrative State, 45 UCLA L. Rev. 1, 8–12 (1997).
Subdelegations to experi­enced subordinates, however, can expedite agency action through the internal reallocation of resources. 43 See, e.g., Delegation of Authority to Director of Division of Enforcement, 74 Fed. Reg. 40,068, 40,068 (Aug. 11, 2009) (codified at 17 C.F.R. pt. 200) (“This action is intended to expedite the investigative process by removing the need for enforcement staff to seek Commission approval prior to performing routine functions.”).

To help shed light on these practices, this Essay advances a positive theory for how and why agency heads subdelegate their power and ana­lyzes the resulting normative implications. 44 A preliminary objection to the pursuit of a positive theory of the bureaucracy is that agencies are too heterogeneous for study as a category. But theories of the firm similarly abstract away from even greater institutional diversity, nevertheless producing useful insights. Many legislative theories reduce legislative forms to their essentials in order to allow for comparative analysis. See Amie Kreppel, Typologies and Classifications, in The Oxford Handbook of Legislative Studies 82 (Shane Martin et al. eds., 2014). Similarly, agencies have recognizable unifying characteristics: They are all policymaking institutions with delegated authority exercised by political appointees and career staff, subject to budget constraints. That said, future work will likely reflect further upon the extent to which agency heterogeneity affects both empirical and theoretical attempts to understand the federal bureaucracy. It is based on a close study of existing and historical subdelegations within a wide range of admin­istrative agencies; in this sense, it is a work of qualitative, grounded theory-building. 45 See Jerry W. Willis, Foundations of Qualitative Research 306–09 (2007) (defining “grounded theory” as a qualitative approach in which researchers examine “successive waves of data” to build a theory that seeks to be parsimonious, internally consistent, and of appropriate scope, explanatory power, and generalizability). See generally Juliet Corbin & Anselm Strauss, Basics of Qualitative Research: Techniques and Procedures for Developing Grounded Theory (2008) (outlining a framework for qualitative-research analysis). The burden is to develop a compelling, parsimonious, and falsifiable explanation for a significant amount of the observed variation in subdelegation practices. 46 In an ideal world, this project would have also incorporated mixed methods, including a large-N quantitative analysis of internal delegations across various agencies. Examining such data over a number of years could help yield more systematic insights on variations over time and under different political constellations. Compiling such a dataset in this context, however, is not feasible: The Code of Federal Regulations is not available in machine-readable XML format, nor do agencies always publish their subdelegations in the CFR, as this paper will discuss. See infra section II.B.3 (discussing agency transparency and the publication of subdelegations). Drawing on a rich social science literature, this Essay argues that delegations within agencies are best understood as credible commit­ment devices through which commis­sioners motivate better-informed but potentially biased subordinates.

The central positive claim is that variations in subdelegation prac­tices reflect differences in agency heads’ willingness to delegate credibly. 47 See, e.g., Helmut Bester & Roland Strausz, Contracting with Imperfect Commitment and the Revelation Principle: The Single Agent Case, 69 Econometrica 1077, 1077–79 (2001); Vijay Krishna & John Morgan, Contracting for Information Under Imperfect Commitment, 39 RAND J. Econ. 905, 905–07(2008). More specifically, the theory predicts that an agency head will commit more credibly to a subdelegation in order to induce more infor­mation production, but only within a bounded range of expected pref­erence divergence and internal transaction costs. As the potential bias of the delegate increases, the agency head is more likely to opt for veto authority. Conversely, as preferences align, the agency head is more willing to subdelegate credibly since she can obtain additional informa­tion without the same risk of bias. At the same time, the transaction costs of reviewing and reversing a subordinate’s recommendation also matter. When the transaction costs of internal review increase, the agency head is more likely to delegate credibly since doing so eliminates those trans­action costs.

Previous legal scholarship on agency subdelegation, by contrast, has largely focused on matters of statutory interpretation: whether Congress explicitly permitted the delegation and, if not, how to understand legis­lative silence or ambiguity. 48 See, e.g., Daniel A. Farber & Anne Joseph O’Connell, Agencies as Adversaries, 105 Calif. L. Rev. (forthcoming 2017) (manuscript at 50), (on file with the Columbia Law Review) (examining statutory issues of subdelegation in the context of adversarial agencies); Nathan D. Grundstein, Subdelegation of Administrative Authority, 13 Geo. Wash. L. Rev. 144, 144–45 (1945) (describing the question “to be probed” as the extent to which the power to subdelegate authority may be implied when Congress does not make an explicit provision authorizing it); Jason Marisam, The Interagency Marketplace, 96 Minn. L. Rev. 886, 892 (2012) (noting “[t]he question of whether Congress has authorized subdelegation is a matter of statutory interpretation”); Note, Subdelegation by Federal Administrative Agencies, 12 Stan. L. Rev. 808, 808–09 (1960) (“[T]he problem of permissible subdelegation is primarily one of statutory interpretation.”). A related issue is the extent to which the President has an inherent constitutional authority to subdelegate or whether the nondelegation doctrine demands congressional authoriza­tion. 49 See Thomas W. Merrill, Rethinking Article I, Section 1: From Nondelegation to Exclusive Delegation, 104 Colum. L. Rev. 2097, 2175 (2004). Still other scholars have proposed that judicial deference to agency interpret­ations should extend only to those signed off on by agency heads, rather than their subordinates. 50 See Barron & Kagan, supra note 12, at 201–05. And perhaps most relevantly here, Professor Elizabeth Magill has examined how agencies “self-regulate” in order to maintain control over subdelegations. 51 See Elizabeth Magill, Agency Self-Regulation, 77 Geo. Wash. L. Rev. 859, 898 (2009) (“An agency may self-regulate in order to control internal delegations so that those within the agency do not have opportunities for arbitrary or corrupt exercises of state power.”). While a broader literature con­siders how agencies delegate to external actors like other agencies 52 See, e.g., Bijal Shah, Interagency Transfers of Adjudication Authority, 34 Yale J. on Reg. (forthcoming 2017) (manuscript at 3) (on file with the Columbia Law Review) (examining “coordinated interagency adjudication in which agencies transfer wholesale their jurisdiction to adjudicate administrative decisions to other agencies”). or private parties, 53 See, e.g., Jon D. Michaels, Privatization’s Pretensions, 77 U. Chi. L. Rev. 717, 718–19 (2010) (arguing governmental organizations use “workarounds” to delegate authority to private parties that can influence administrative outcomes). the main focus here is on internal delegations.

More specifically, Part I frames the problem in terms of the agency head’s choice to delegate final authority to a subordinate or to retain a veto after reviewing the subordinate’s recommendation. The decision involves trade-offs between the relative potential for superior informa­tion, bias, and internal review costs. Part II considers the ways in which agency heads can make their delegations more credible in order to motivate information production. These design choices include greater reputational backing, transparency, longevity, dissolution, and, finally, entrenchment. As such, one of this Essay’s main contributions is to exa­mine how credibility is institutionalized within administrative agencies.

Part III’s normative prescriptions, in turn, follow from the previous positive premises: Generally stated, courts should maximize the use of high-quality information in the administrative state by encouraging credi­ble delegation. Doing so will increase the likelihood that expertise within agencies is used efficiently. Specifically, subdelegation raises constitu­tional concerns given that tenure-protected employees could exercise significant delegated authority. In this context, courts should apply a more functional understanding of political control attuned to the credib­ility and nature of the subdelegation. Moreover, judges should only extend Chevron deference to a subdelegation when the agency head has delegated authority in a way that is transparent and otherwise evinces a commitment to elicit expertise.

Furthermore, courts should also help to encourage credibility when enforcing internal delegations under the Accardi doctrine—the basic principle that an agency must abide by its own rules. 54 See United States ex rel. Accardi v. Shaughnessy, 347 U.S. 260, 266–68 (1954) (holding the agency failed to follow its own regulations by allowing the Attorney General to “dictat[e] the Board [of Immigration Appeal]’s decision”). See generally Thomas W. Merrill, The Accardi Principle, 74 Geo. Wash. L. Rev. 569, 569 (2006) [hereinafter Merrill, The Accardi Principle] (“Agencies must comply with their own regulations. This is sometimes called ‘the Accardi principle’ . . . .”). For recent cases citing Accardi, see, e.g., Burdue v. FAA, 774 F.3d 1076, 1082 (6th Cir. 2014); Brown v. Holder, 763 F.3d 1141, 1148 (9th Cir. 2014). Because the doctrine still creates uncertainties for subordinates deciding whether to invest resources, judges should set clearer bright-line rules to facilitate information-forcing grants of authority. Finally, this Part also grapples with the potential for agency subdelegation to be used as a tool of partisan entrenchment—a means for agency heads to project their preferences beyond their tenures. As such, it argues that political actors, rather than courts, are best situated to prevent the entrenchment of delegations that can undermine democratic accountability.

I. Subdelegation’s Trade-offs

This Part develops a positive theory for when and why agency heads internally delegate power to their subordinates. 55 See sources cited supra note 45. In doing so, it builds on the work of scholars like Sean Gailmard, John Patty, and Matthew Stephenson to highlight the various trade-offs that administrators face in the decision: between more information and bias, and between internal transaction costs and a loss of control. The first section frames the delegation problem as a choice between final and reviewable subdelega­tions—in essence, the choice between treating staff as authorities or advisors. The second section then amplifies these dynamics and the organizational costs of advice giving within the bureaucracy. In doing so, it explores the relationship between information production and flow within agencies and, more specifically, how the transaction costs of con­veying information can influence the agency head’s initial decision to subdelegate.

A. Advisors Versus Authorities

Managing a bureaucracy, like managing a business, requires entrust­ing tasks to others. Agency heads have finite resources and skills. They cannot personally approve—let alone research, analyze, and evaluate—the myriad decisions required to be made on a daily basis. Agency heads must thus rely on their subordinates to perform many functions. 56 See Fleming v. Mohawk Wrecking & Lumber Co., 331 U.S. 111, 122–23 (1947) (noting that the administrator “could hardly be expected, in view of the magnitude of the task, to exercise his personal discretion” on every matter over which Congress granted him authority); Nat’l Nutritional Foods Ass’n v. Food & Drug Admin., 491 F.2d 1141, 1146 (2nd Cir. 1974) (upholding a new commissioner’s ability to “confer[] with his staff” regarding rulemaking proceedings that preceded his tenure rather than personally reviewing all of the materials). These subordinates include an ever-“thickening” array of political appointees 57 See Paul C. Light, Thickening Government: Federal Hierarchy and the Diffusion of Accountability 8 (1995). as well as thousands of civil servants. 58 This characterization is, of course, a simplification. More accurately, federal government personnel consist of three categories: the merit-based competitive service, the Senior Executive Service (SES), and those who are “excepted” from merit-based restrictions. See O’Connell, Vacant Offices, supra note 34, at 925–26. Of these categories, the first two consist almost entirely of career employees who are hired through statutory examination, selection, and placement rules. 5 U.S.C. § 2102(a)(1) (2012). The SES “contains career employees as well as political officials, but political appointees can make up no more than 10 percent of the whole SES (or one-quarter of the SES slots in any one agency).” O’Connell, Vacant Offices, supra note 34, at 925–26. These employees are, in turn, only removable “for cause,” specifically upon a showing of “insubordination.” 5 U.S.C. §§ 4303, 7513(a), 7521(a)–(b); see also Jason Marisam, The President’s Agency Selection Powers, 65 Admin. L. Rev. 821, 863 (2013).
The excepted service, in turn, is composed of those employees exempt from the competitive service by statute, executive action, or the Office of Personnel Management. See Mike Litak, U.S. and Them: Citizenship Issues in Department of Defense Civilian Employment Overseas, Army Law., June 2005, at 1, 3. The excepted service itself contains three subcategories: Schedules A, B, and C. Schedule A includes jobs for which it would be “impractical” to use standard qualification requirements, such as specialized professions like law, medicine, and chaplaincy. Schedule B also refers to jobs for which exams would be inappropriate but for which applicants nevertheless must meet certain qualification standards; it “is typically used for new agencies” as well as student internships. And, finally, Schedule C includes lower-level political appointees that “keep a confidential or policy-determining relationship to their supervisor and agency head,” such as special assistants. Excepted Service Appointing Authorities, U.S. Office of Pers. Mgmt., [] (last visited Oct. 19, 2016).
Political appointees are generally selected by the President or an agency head. 59 Political Appointees, U.S. Office of Gov’t Ethics (Feb. 26, 2016), []. Their tenures are not protected, lasting an average of two years. 60 See Thomas W. Merrill, High-Level, “Tenured” Lawyers, 61 Law & Contemp. Probs., Spring 1998, at 81, 83, 93. They possess a dizzying array of titles such as deputy secretary, deputy undersecretary, principal deputy admin­istrator, or assistant deputy administrator, among others. 61 Light, supra note 57, at 12 tbl.1-3. Civil servants, by contrast, possess tenure and salary protections. 62 Civil Service Reform Act of 1978, Pub. L. No. 95-454, 92 Stat. 1111; see also Ronald N. Johnson & Gary D. Libecap, The Federal Civil Service System and the Problem of Bureaucracy 7 (2007) (observing that civil servants have “strict tenure guarantees, have no expressed ties to the administration or to Congress, and by law are to be politically neutral”). By law, they are hired on the basis of merit, not political ties. 63 Johnson & Libecap, supra note 62, at 7. As a result, they often enter government with professional norms informed by technical or legal training. 64 Id.

Agency heads thus face a choice in how to manage interactions with their subordinates: They can either assign a decision to them while retaining veto authority or else delegate the decision entirely. 65 See Wouter Dessein, Authority and Communication in Organizations, 69 Rev. Econ. Stud. 811, 812 (2002) (“[T]he principal optimally delegates control as long as the divergence in preferences is not too large relative to the principal’s uncertainty about the environment.”); Sean Gailmard & John W. Patty, Giving Advice vs. Making Decisions: Transparency, Information, and Delegation 2–3 (Oct. 7, 2014) [hereinafter Gailmard & Patty, Giving Advice] (unpublished manuscript),
u/4231163/GailmardPatty-TransparencyInformationDelegation-October2014.pdf [http://] (explaining how principals can delegate authority either com­pletely or partially).
Put differently, agency heads can review their underlings’ recommendations or else let them make the final choices. Call the former reviewable subdelegations and the latter final subdelegations. 66 This distinction between reviewable and final subdelegation is artificially simplified for now to gain some initial analytical traction. In reality, as will be discussed below, there is a range of institutional design questions between these two choices, including whether to make such subdelegations temporary or indefinite, transparent or nontransparent, and so on. See infra Part II. To illustrate this (for now) simplified distinction, take the grant of authority at issue in a recent constitutional challenge to how administrative law judges (ALJs) are appointed within the SEC. 67 See Raymond J. Lucia Cos. v. SEC, 832 F.3d 277, 277 (D.C. Cir. 2016) (upholding the constitutionality of SEC ALJs as consistent with the Appointments Clause). The SEC is authorized by Congress to enforce securities laws either by filing civil suit in federal district court or else through administrative adjudication. 68 Id. at 282. The Commis­sion granted its ALJs the authority to conduct administrative hearings. 69 Id. (citing 17 C.F.R. § 200.30-9 (2016)).

The relevant statute, however, allowed the Commission a choice in how to treat the ALJ’s initial adjudicatory decision. 70 Id. On the one hand, the SEC could “adopt regulations whereby an ALJ’s initial decision would be deemed a final decision of the Commission.” 71 Id. at 286. This delegation would be a final subdelegation: The ALJ’s decision would be the agency’s de­cision. On the other hand, the SEC could instead choose to extend a reviewable subdelegation—the path it eventually took. 72 Id. More specifically, the SEC established an internal review process in which it gave itself time to review the initial order; should the Commission decline such review, it would then issue an order to that effect. 73 Id. (citing 17 C.F.R. § 201.411(c)). Because the subdelegation was nonfinal in this respect, a panel of the D.C. Circuit upheld the constitu­tionality of the ALJ’s appointment scheme, though the ruling will now be considered en banc. 74 Raymond J. Lucia Cos. v. SEC, 832 F. 3d 277 (D.C. Cir. 2016), vacated and reh’g en banc granted, 832 F. 3d 277 (D.C. Cir. 2017).

At first glance, one might wonder why agency heads would ever choose a final over a reviewable subdelegation—why treat subordinates as authorities rather than mere advisors? Why not instead, as with the SEC’s ALJs, keep the option to reject a decision even while rubberstamping it? After all, delegating is dangerous: There is a persistent risk that the delegate will be defiant or otherwise make a bad decision. 75 Cf. Joshua D. Clinton et al., Separated Powers in the United States: The Ideology of Agencies, Presidents, and Congress, 56 Am. J. Pol. Sci. 341, 345–46 (2012) (finding that “political appointees are ideologically distinctive from career executives”). To minimize that risk, one might think it always better to retain the ability to restrain wayward subordinates. Final subdelegations, by contrast, entail a pledge not to reverse the subordinate’s decision. 76 Dessein, supra note 65, at 811. When authority is delegated in this manner, it is committed to the delegate’s discretion. By subdele­gating a power with finality, the agency head effectively gives up control.

Nevertheless, final subdelegations are pervasive in the real world. The SEC’s Director of Enforcement, as will be discussed, currently exercises the Commission’s subpoena authority. 77 See infra notes 90–101 and accompanying text. The Office of Legal Counsel (OLC) is generally not reviewed by the Attorney General who delegated to the OLC the authority to issue executive branch legal opinions. 78 See infra notes 246–249 and accompanying text (noting how the Office of Legal Counsel is perceived as an independent entity of the Department of Justice’s Attorney General). The Food and Drug Administration (FDA) used to exercise delegated rulemaking authority independently from the Department of HHS. 79 See infra note 245 and accompanying text (noting the historical presumption that the FDA was independent of HHS despite being in the same agency). The puzzle here is to explain why. Of course, agency heads, no less than other political actors, do not seek to maximize their power but rather to optimize it. 80 See Jon Elster, Ulysses Unbound: Studies in Rationality, Precommitment, and Constraints 88–92 (2000) [hereinafter Elster, Ulysses Unbound] (exploring the “proposition that sometimes less is more or, more specifically, that sometimes there are benefits from having fewer opportunities rather than more”); Lisa Schultz Bressman & Robert B. Thompson, The Future of Agency Independence, 63 Vand. L. Rev. 599, 601 (2010) (“[T]he President does not always have a political interest in seeking maximum control of regulatory policy.”); Magill, supra note 51, at 860 (“Individuals, firms, governments, and, yes, of course, bureaucracies—especially bureaucracies—will sometimes have an interest in voluntarily limiting their options.”). Sometimes, in other words, it is beneficial for those in control to cede that control.

On the one hand, doing so can free up resources to pursue higher-priority tasks. By declining decisionmaking authority, agency heads can focus their attention on more important matters. At the same time, subdelegating that authority to others can also be valuable in its own right. When the recipient of the delegation is more expert on a relevant matter, subdelegation can result in better-informed decisions. Indeed, in many ways, high-quality information is the bureaucracy’s raison d’être. 81 See Sean Gailmard & John W. Patty, Learning While Governing 2 (2013) [hereinafter Gailmard & Patty, Learning] (“Yet for all the accountability problems it presents, the value of high-quality information in bureaucracies cannot be seriously doubted.”). Assigning decision rights within the agency can take advantage of know­ledge dispersed within the organization. 82 See Michael C. Jensen & William H. Meckling, Specific and General Knowledge, and Organizational Structure 1 (1990), (on file with the Columbia Law Review) (“[G]etting specific knowledge used in decision-making requires decentralizing many decision rights in . . . firms.”). It can also encourage appointees and civil servants alike to develop more specialized expertise. 83 See generally Gailmard & Patty, Learning, supra note 81, at 25–27 (“We argue that incentives for administrative actors [like bureaucrats] to acquire policy expertise result from the organizational structure and political position they occupy as created by political principals . . . .”); Magill, supra note 51, at 890 (“[S]elf-regulatory measures can be used to establish and delegate to entities that will produce . . . information and then invest in becoming experts—working groups, advisory committees, and the like.”); Matthew C. Stephenson, Bureaucratic Decision Costs and Endogenous Agency Expertise, 23 J.L. Econ. & Org. 469, 471–73 (2007) [hereinafter Stephenson, Bureaucratic Decision Costs] (discussing various models of agency incentives to develop endogenous expertise).

Indeed, social scientists have long recognized that an important function of delegation is to motivate effort and information acquisition. 84 See, e.g., Gailmard & Patty, Learning, supra note 81, at 25–28 (explaining how policy discretion can “induce expertise development”); Philippe Aghion & Jean Tirole, Formal and Real Authority in Organizations, 105 J. Pol. Econ. 1, 3 (1997) (“[T]ransfer of formal authority to an agent credibly increases the agent’s initiative or incentive to acquire information.”); Ryan Bubb & Patrick L. Warren, Optimal Agency Bias and Regulatory Review, 43 J. Legal Stud. 95, 95–101 (2014) (examining how delegation can generate costly information subject to risks of shirking and bias); Matthew C. Stephenson, Information Acquisition and Institutional Design, 124 Harv. L. Rev. 1423, 1422–27 (2011) [hereinafter Stephenson, Information Acquisition] (examining “how different institutional arrangements . . . might affect the production of useful information by government agents”); see also infra note 85 (providing additional sources addressing delegation as a motivator of effort and information gathering). This intuition has driven theories, for example, of why Congress restricts its own ability to amend internal committee proposals 85 See Epstein & O’Halloran, supra note 5, at 14–18 (providing an overview of informational theories of legislative organization); Thomas W. Gilligan & Keith Krehbiel, Asymmetric Information and Legislative Rules with a Heterogeneous Committee, 33 Am. J. Pol. Sci. 459, 459–63 (1989); Thomas W. Gilligan & Keith Krehbiel, Collective Decisionmaking and Standing Committees: An Informational Rationale for Restrictive Amendment Procedures, 3 J.L. Econ. & Org. 287, 287 (1987). and why firm managers delegate to employees. 86 See, e.g., Daron Acemoglu et al., Technology, Information, and the Decentralization of the Firm, 4 Q.J. Econ. 1759, 1759 (2007). Administrative law scholars, of course, are also familiar with the concept that Congress grants agencies discre­tion to capitalize on their expertise. 87 See Margaret H. Lemos, The Consequences of Congress’s Choice of Delegate: Judicial and Agency Interpretations of Title VII, 63 Vand. L. Rev. 363, 372–73 (2010). The unifying insight is that princi­pals are willing to trade off control for information. 88 See Bendor et al., Theories of Delegation, supra note 4, at 242 (describing what is “generally regarded as the major trade-off facing the boss: Is the gain produced by delegating the decision to a more informed party worth the loss produced by having someone with different preferences make the choice”); Jensen & Meckling, supra note 82, at 1–3 (discussing how principals assign decisionmaking rights to agents and how the costs of transferring information between agents influences the organization of markets and firms); Stephenson, Information Acquisition, supra note 84, at 1440 (“When deciding how much discretion to delegate . . . the principal must weigh the potential informational gains of delegation against the costs associated with potential agency bias.”). Congress resists the ability to revise committee bills in order to encourage data gathering. Firm managers empower employees to spur the same goal. 89 See Diego Stea, Kirsten Foss & Nicolai J. Foss, A Neglected Role for Organizational Design: Supporting the Credibility of Delegation in Organizations, 4 J. Org. Design 3, 3 (2015) (noting research “recognize[s] that delegating discretion to employees can foster organizational value creation” by “economizing on scarce managerial attention and allowing for efficient use of local knowledge” and “increas[ing] the autonomous motivation of employees”). So too with agency heads. They are willing to accept the risks of ceding authority to an actor with divergent preferences in exchange for better-informed decisions.

Beyond its intuitive appeal, this account also has real-world res­onance. Consider, for example, the SEC’s experience before and after it subdeledated the Commission’s subpoena power. Beforehand, SEC Chairman Christopher Cox had used a “cumbersome” review process to approve each and every exercise of the power. 90 Zachary A. Goldfarb, In Cox Years at the SEC, Policies Undercut Enforcement Efforts, Wash. Post (June 1, 2009),
article/2009/05/31/AR2009053102254.html [].
During this review, he required the enforcement staff to fly in and physically present the case to commissioners in Washington, D.C., even if the staff members were based in New York or California. 91 Id. He also often “postponed . . . [enforcement] decisions at the last minute, leaving cases unresolved for months,” only to eventually weaken the proposed sanctions. 92 Id. According to several career enforcement lawyers, Chairman Cox’s review “practices had a chilling effect.” 93 Id. In one lawyer’s words:

The presentation of cases is the culmination of the investigative process. When that process is interrupted, delayed or denied, it can’t help but have a negative impact on the people who conduct those investigations . . . . Clearly some people wonder, ‘If they don’t want these kinds of cases, why should I bother doing them even though they’re very important?’ 94 Id. (internal quotation marks omitted) (quoting James T. Coffman, former assistant director of the SEC Division of Enforcement).

As a result, many SEC enforcement officials had stopped investing their time and effort in developing cases to pursue. 95 Id. Chairman Cox’s use of a reviewable subdelegation is thus arguably an example of how such an approach can reduce employees’ incentives and motivation to invest resources.

The internal dynamics of the SEC, however, changed in 2009 with the arrival of Chairwoman Mary Schapiro. An Obama appointee eager to take the “handcuffs” off the enforcement division, Chairwoman Schapiro’s preferences were more closely aligned with those of her enforcement staff. 96 Id. She thus spearheaded an effort to subdelegate the Commission’s authority to issue “formal orders of investigation” to the SEC’s Director of Enforcement. 97 Delegation of Authority to Director of Division of Enforcement, 74 Fed. Reg. 40,068, 40,068 (Aug. 11, 2009) (codified at 17 C.F.R. pt. 200). Such authority could now be exercised without com­missioner approval or cumbersome review. 98 Id. Such orders are practically significant because they empower investigators to issue judicially enforceable subpoenas compelling companies to turn over documents and give sworn testimony instead of forcing them to rely solely on voluntary cooperation. See Russell G. Ryan, Delegation and Accountability at the SEC, Harv. L. Sch. F. on Corp. Governance & Fin. Reg. (Oct. 7, 2009),
delegation-and-accountability-at-the-sec/ [].
The SEC’s Director of Enforcement Robert Khuzamii then announced in a speech that he was further redelegating that authority to the career staff. In his words, “staff will no longer have to obtain advance Commission approval in most cases to issue subpoenas; instead, they will simply need approval from their senior supervisor.” 99 See Robert Khuzamii, Dir., Div. of Enf’t, U.S. Sec. & Exch. Comm’n, Speech by SEC Staff: Remarks Before the New York City Bar: My First 100 Days as Director of Enforcement (Aug. 5, 2009), [] (announcing “inten[t] to delegate [delegated subpoena] authority to senior officers throughout the Division” so that the “staff will no longer have to obtain advance Commission approval in most cases to issue subpoenas; instead, they will simply need approval from their senior supervisor”). Practically speaking, Director Khuzamii continued, “[t]his means that if defense counsel resist the voluntary production of documents or witnesses, or fail to be complete and timely in responses or engage in dilatory tactics, there will very likely . . . be a subpoena on your desk the next morning.” 100 Id.

Pursuant to these subdelegations, the SEC’s tenure-protected career staff could now exercise the agency’s subpoena authority with minimal oversight. As a result, the number of formal orders issued by the staff nearly doubled from 2008 to 2012. 101 See Sarah N. Lynch, SEC Official Says Staff May Have Too Much Power over Subpoenas, Reuters (Nov. 22, 2013, 3:35 pm), []. In this manner, the SEC’s switch from a reviewable to a near-final subdelegation likely helped motivate its subordinates to invest more effort in acquiring information about illegal securities violations. Because subordinates no longer had to seek full Commission review of their decisions, they could devote more information-gathering resources with little fear of reversal.

B. Information Production and Transmission

Nevertheless, the observation remains that agency heads—like the SEC Commissioners under Cox’s chairmanship—still gained some infor­mation on which firms had likely committed securities violations, even when the enforcement staff acted as advisors rather than author­ities. 102 See Goldfarb, supra note 90 (describing how Chairman Cox and commissioners reviewed cases presented by SEC enforcement officials before reducing recommended penalties). In other words, even when SEC staff members were subject to Commission review, they presented evidence about potential enforcement cases, even if the research was more limited or less accurate than it might otherwise have been under subdelegated authority. Agency heads, however, face at least two different kinds of challenges regarding such research, the first regarding its quality and the other pertaining to its review. Both chal­lenges are distinct kinds of transaction costs associated with information transmission, which in turn yield important implica­tions for information production within bureaucracies. 103 See Richard M. Cyert & James G. March, A Behavioral Theory of the Firm 79 (1992) (observing “[w]here different parts of the organization have responsibility for different pieces of information . . . , [one would expect] some attempts to manipulate information as a device for manipulating the decision”); Lee J. Alston & William Gillespie, Resource Coordination and Transaction Costs: A Framework for Analyzing the Firm/Market Boundary, 11 J. Econ. Behavior & Org. 191, 192 (1989) (developing a framework for classifying transaction costs and noting that “the types of transaction costs encountered in organizing production vary across the factors of production and over the stages of the production process” (emphasis omitted)).

The first problem for agency heads is that it is not always in the staff members’ interest to reveal their private information or gain enough information to make a precise assessment. Doing so may lead an agency head to make a decision with which the staff member disagrees, espe­cially when their preferences diverge. 104 In addition to policy bias, another concern for principals is the incentive for their agents to shirk—to exert a sub-optimal amount of effort gathering costly information. See Bubb & Warren, supra note 84, at 96. Shirking is of particular concern when the agent cannot benefit fully from her decision. Id. Such a prospect is likely to lead agency heads to choose more biased agents. Id. (arguing that principals often choose biased delegates since “policy bias can be harnessed to mitigate the problem of shirking”). The analysis here, however, assumes that agency subordinates fully benefit from the consequences of the agency’s decision under subdelegated authority. Say, for example, a civil servant at the EPA reads two public comments or reports regarding the effects of climate change with opposing viewpoints. Cognitive heuristics and bias can lead her to report conclusions in line with her prior beliefs without presenting contradicting evidence or highlighting relevant uncertainties to the EPA Administrator. As such, when agency heads review a subor­dinate’s decision without being able to independently evaluate and verify the basis upon which the recommendation was made, the subordi­nate has an incentive to misrepresent the underlying information. 105 See Stephenson, Information Acquisition, supra note 84, at 1457. This condition of “oversight without transparency” contrasts with other contexts in which the reviewer has “full transparency” with regard to the subordinate’s underlying evidence or when the reviewer can simply observe the subordinate’s research efforts. Id. While these latter cases may apply to certain reviewable subdelegations in agencies, the analysis here assumes (as is usually the case) that the underlying recommendations are presented in some form to the agency head, whether verbally in a meeting or by written memos or briefings. For work examining “under what circumstances a principal, operationalized as an overseer with the power to invalidate agent policy choices, would prefer to observe the policy proposed by an agent whose expertise, ability, and policy preferences are all known,” see John W. Patty & Ian R. Turner, Ex Post Review and Expert Policymaking: When Does Oversight Reduce Accountability? 3 (Jan. 24, 2017) (unpublished manuscript),
abstract=2781750 (on file with the Columbia Law Review).

At the same time, agency heads are not naïve. They will act or not act upon the advice of staff members given what they know (or think they know) about their staff members’ preferences—what is often referred to as the “ally principle.” 106 See Stephenson, Information Acquisition, supra note 84, at 1440. Take again the continuing example of SEC Chairman Cox’s tenure. A Republican, Chairman Cox was known to be concerned with excessive regulatory enforcement, much to the chagrin of his proenforcement career lawyers. 107 See Goldfarb, supra note 90. Such lawyers thus had an incentive to overstate the strength of a case in the hopes of convincing Chairman Cox to pursue it. Agency heads like Chairman Cox thus had reason to dismiss such information as “cheap talk”: recommendations that are not credible to the skeptical receiver. 108 See Joseph Farrell & Matthew Rabin, Cheap Talk, 10 J. Econ. Persp. 103, 116 (1996) (“Cheap talk consists of costless, nonbinding, nonverifiable messages that may affect the listener’s beliefs.”). More generally, when agency heads suspect that recommending staff members or political underlings are biased in the sense of having divergent preferences, they will discount the recommendations accordingly.

In other contexts, these game-theoretic dynamics have inspired a rich literature on strategic communication. 109 The seminal work in this literature is Vincent P. Crawford & Joel Sobel, Strategic Information Transmission, 50 Econometrica 1431 (1982). One insight from this work is that when preferences sufficiently diverge, information cannot be conveyed reliably within an organization. 110 See id. at 1450. Put more colloquially, when bosses sufficiently distrust their subordinates, they ignore their subordi­nates’ advice. Conversely, as preferences become more aligned, the amount of information that can be credibly conveyed increases. 111 See id. Within a limited range of preference divergence, informative communi­cation can thus take place; it is in the interests of the sender, that is, to reveal genuine information and the receiver will act accordingly. Under these circumstances, principals—like agency heads—would prefer to dele­gate to their agents rather than to engage in costly strategic communica­tion. 112 See Dessein, supra note 65, at 811; Stephenson, Information Acquisition, supra note 84, at 1459. When preferences are sufficiently aligned, in other words, agency heads would prefer final rather than reviewable subdelegation.

The underlying logic is straightforward: Final subdelegation rep­resents a commitment to let a subordinate benefit from any expertise gained. Such delegation therefore creates the strongest incentive for the subordinate to develop private information. By contrast, if the subordi­nate is subject to reversal, she will be less inclined to acquire information when it is costly do so. The agency head thus benefits the most from final subdelegation as long as the bias is within a range in which informative communication could have otherwise occurred. Final delegation allows the agency head to avoid the costs of strategic communication while maximizing the benefits of the subordinate’s private information. In this sense, the agency head’s decision to subdelegate can be understood as a trade-off between informed-but-biased decisions (under a final subdele­ga­tion) and noisy-but-unbiased decisions (under a reviewable subdelegation).

The EPA’s enumerated delegation practices help to illustrate. The EPA currently documents all of its subdelegations in a Delegations Manual, which, in its own words, serves as the “legal record of the authority of an Agency employee or representative to act on behalf of the Administrator.” 113 U.S. EPA, Introduction, para. 1, in U.S. EPA, EPA Delegations Manual [hereinafter EPA Delegations Manual Introduction] (on file with the Columbia Law Review). As the manual further explains:
“The laws, Executive Orders and regulations which give EPA its authority typically, but not always, indicate that “the Administrator shall” exercise certain authorities. Official delegations of authority represent the basic direction to senior Headquarters and Regional management officials to exercise these delegated authorities. It is EPA’s policy that, in order for other Agency management officials to act on behalf of the Administrator, the authority granted by Congress or the Executive Branch must be delegated officially . . . through the Agency’s delegation process.”
The manual specifies a number of criteria EPA admin­istrators should use in considering whether or not to delegate authority. 114 Id. para. 3(a). Revealingly, one general principle is that “authority and accountability should rest as close as possible to where the covered action takes place” within the organization. 115 Id. If the action affects regional or local actions, for example, the manual states that responsibility should be delegated to a Regional Administrator. 116 Id. Indeed, doing so would encourage the Regional Administrator to develop more location-specific expertise. Simi­larly, issues that affect a single substantive program within an agency—for example, air pollution or hazardous waste—also merit subdelegation in the EPA’s view. This policy again makes sense from an informational per­spective: Agency heads gain better decisions from motivated subject-matter experts.

Final subdelegations, however, would only make sense for the EPA Administrator within a limited range of potential preference divergence. Otherwise, the Administrator is better off with a reviewable subdelega­tion so as to maintain control. Put differently, when the potential costs of bias outweigh the costs of review, agency heads will be more hesitant to delegate. This dynamic is reflected in the EPA’s attention to the scope of the delegation—privileging delegations with limited discretion and thus a bounded risk for bias, while discouraging those in which the risk is higher. 117 Id. For instance, the EPA prioritizes “recurring” subdelegations. 118 See id. (noting standard setting and rulemaking should remain with the Administrator but processes that “may be more manageable at lower organizational levels” may be delegated with sufficient guidance). Recurring actions usually do not require much discretion, so their dele­gation is a low-risk matter. In contrast are subdelegations in which the risks of preference divergence are so great that the Administrator would do well not to delegate in the first place. Indeed, the manual states that “[i]f action under the authority is likely to set legal or programmatic precedents, it should remain with the Administrator or Assistant Administrator(s).” 119 Id. These actions include “rulemaking and standard-setting which set long-term commitments for the Agency.” 120 Id. However, even these responsibilities are delegable if “sufficient guidance exists to assure program consistency.” 121 Id. Put differently, when discretion is ade­quately constrained, subdelegation is more desirable since the risks of bias are limited.

Taking a step back, note that this motivational account is most plausible only when the agency head’s preferences are transparent to the subordinate ex ante, as was the case within the SEC’s enforcement divi­sion. 122 See Goldfarb, supra note 90. This insight is related to, though distinct from, conditions that Professors Gailmard and Patty refer to as “top-down transparency” situations in which subordinates have access to the principal’s information prior to making their own recommendation. See Gailmard & Patty, Giving Advice, supra note 65, at 5–6 (defining “top-down transparency” as conditions in which “the agent has access to the principal’s information prior to making his or her own policy and/or message choices”). The extent to which the subordinate knows in advance how the agency head is likely to act upon being given information is itself an important factor in deciding how much time and effort to exert when gathering that information. When the subordinate knows, for instance, that the agency head has sufficiently divergent preferences such that the subordinate’s recommendation is likely to be treated as cheap talk, then the subordinate lacks the incentive to gather that information in the first place. 123 See Stephenson, Information Acquisition, supra note 84, at 1459 (noting that “if the principal and agent have such different preferences that the principal will disregard any report by the agent, then the agent has no incentive whatsoever to acquire information, since this information will always be ignored”). To illustrate, since SEC investigators knew that Chairman Cox had an antienforcement agenda and would likely discount their views, they were understandably less motivated to develop the underlying cases. By contrast, if Chairman Cox’s preferences were nontransparent or more closely aligned, then the marginal incentive to engage in information gathering could actually be higher, since the staff might work harder in the hopes of persuading Chairman Cox to proceed.

In addition to the information loss from strategic communication, the agency head must also consider the organizational transaction costs associated with reviewing the work of their subordinates. 124 Such costs may be understood as a species of “decision costs.” See Stephenson, Bureaucratic Decision Costs, supra note 83 at 20. Understanding them as transaction costs, however, helps to draw the connection to insights form the theory of the firm and a broader organization literature. Most, if not all, administrative agencies operate according to vertical hierarchies re­quiring structures and processes to facilitate the ratification or rejection of decisions made below. Many agencies, for example, have multiple layers of adjudicators with ALJs or other hearing officers at the bottom, whose judgments are then reviewable by an internal appellate body and sometimes the agency heads as well. 125 For example, EPA’s Environmental Appeals Board “hears permit and civil penalty appeals in accordance with regulations delegating this authority from the EPA Administrator.” About the Environmental Appeals Board (EAB), Envtl. Prot. Agency, http:/
aboutepa/about-environmental-appeals-board-eab [] (last visited Jan. 16, 2017). The Social Security Administration has an internal Appeals Council that “may grant, deny, or dismiss a request for review” and then, if review is granted, “either decide the case or return . . . it to an administrative law judge for a new decision.” Brief History and Current Information About the Appeals Council, Soc. Sec. Admin., http:// [] (last visited Jan. 16, 2017). The Department of Agriculture’s National Appeals Division “reports directly to the Secretary of Agriculture” and works “to provide fair and timely hearings and appeals to USDA program participants.” National Appeals Div., U.S. Dep’t of Agric., http:// [] (last visited Jan. 16, 2017).
These review processes are costly, both for litigants and the agency. Indeed, pages of the CFR and Federal Register are devoted to detailed and often complex rules regarding agen­cies’ internal review and appeals processes. 126 For the Social Security Administration’s hearing and appeals process, for example, see Procedures of the Departmental Grant Appeals Board, 45 C.F.R. §§ 16.1–.23 (1981).

In reality, agency heads can revise these procedures as a means of controlling wayward subordinates. 127 See Nou, supra note 11, at 440 (“Principal-agent problems often arise because of information asymmetries, and, in this respect, efforts to reduce internal information-acquisition costs can also help mitigate potential agency problems.”). For simplicity’s sake, however, such transaction costs will be treated exogenously here. Assume, in other words, that these review costs are influenced by factors independent of the main variables considered thus far: agency-head and subordinate pre­ferences. 128 Alternatively, transaction costs could also be treated endogenously, which would complicate the theory in ways not addressed here. SEC Chairman Cox, recall, required enforcement staff members to fly to Washington, D.C., and orally present cases for full commission review—a process which imposed high transaction costs. See supra note 91 and accompanying text. One rationale for doing so may have been to decrease the ultimate number of enforcement cases pursued in line with Chairman Cox’s preferences. See supra notes 93–95. This assumption, in turn, has some real-world bite. Internal review processes are often fixed by statute, executive order, or sheer institutional path dependence. 129 See Nou, supra note 11, at 473–78 (providing examples of path-dependent coordination mechanisms as well as “mandatory design requirements” imposed by the President and Congress). For example, many agency heads inherit internal clearance procedures from their predecessors. 130 Id. at 468–71 (discussing internal clearance procedures). These proce­dures require multiple internal officers to review and sign off on a staff recommendation before the decision is approved by the agency head. 131 Id. Each stage of this process imposes substantial transaction costs that the agency head must bear when choosing to review a lower-level official’s recommendation.

As such, the agency head’s subdelegation decision now becomes a function of two different kinds of costs: the information losses due to cheap talk as well as the transaction costs of review. The former category will vary depending on the expected preference divergence of the sub­ordinate, while the latter will depend on exogenous factors that can increase or decrease how expensive it becomes for an agency to engage in intra-agency review. The agency head will thus choose a reviewable rather than a final subdelegation when she expects the review to result in a decision that she prefers, net the transaction costs of reviewing that decision. Conversely, the agency head would prefer to subdelegate the decision entirely as long as the expected preference divergence is sufficiently limited; doing so would eliminate the costs of internal review.




In sum, agency subdelegation practices can be explained as mech­anisms that allow agency heads to capitalize on expertise developed by lower-level officials within an agency. To generate such expertise, agency heads can either treat their subordinates as authorities or advisors. They can extend either reviewable or final grants of authority. This decision, in turn, will be a function of two factors: the expected preference diver­gence of the delegate and the internal costs of review. The greater the preference divergence, the more hesitant the agency head will be to ex­tend final authority. At the same time, delegating this authority becomes more attractive as the exogenous transaction costs of intra-agency review increase. Because subdelegation eliminates those costs, it becomes a more efficient means of agency decisionmaking.

II. Credibility as a Continuum

While the previous Part presented the agency head’s decision as a simplified choice between a reviewable and a final subdelegation, in practice, the range of internal delegation choices is more accurately described along a continuum—a continuum of credibility. At one end are reviewable subdelegations, which are noncredible, and on the other are perfectly credible final subdelegations. Reviewable subdelegations are noncredible in the sense that the recipient knows that the agency head can overrule her decision. Final subdelegations, by contrast, are more credible in that the recipient has reason to believe that the agency head will not do so. Between these two analytic poles are a host of institutional design choices that agency heads can use to make a subdelegation more or less credible. These internal choices exist independently of the ability of external actors like courts to serve a credibility-enhancing function (an issue that will be discussed later). 132 See infra section III.A.

Indeed, the informational benefits of subdelegation only arise when those delegations are dependable. To confidently invest time and re­sources in their decisions, agents must be able to rely on a subdelegation and know that their efforts are not all for naught. The ability of agency heads to spur information production through subdelegation thus re­quires a credible commitment. By contrast, when the subdelegation is not reliable the agent will be less willing to invest time and resources due to the fear that the resulting decision will be overruled. 133 See Stea et al., supra note 89, at 4 (“Employees who believe that managers’ explicit or implicit promises of delegated discretion are not credible will fear that, after having mobilized a high degree of motivation in carrying out their tasks, they may face . . . opportunistic reneging on the part of managers . . . leading to smaller contributions of effort . . . .”). Of course, subdelegations are “loaned, not owned.” 134 George Baker, Robert Gibbons & Kevin J. Murphy, Informal Authority in Organizations, 15 J.L. Econ. & Org. 56, 56 (1999). Those who labor under them always face the risk that their power will be revoked. In a more dynamic, multi-stage game between agency heads and their subdelegates, this threat likely looms large. Those revocations, however, entail procedural or political costs that are often prohibitive in practice. 135 See, e.g., infra notes 186–190 and accompanying text (describing the costly internal process to change subdelegations at EPA).

This Part now explores how the agency head chooses to subdelegate rather than whether to do so. The first section, more specifically, explains why credibility matters more within, as opposed to between, different institutions. The second section then provides a taxonomy of commitment-enhancing design choices. These options include increased reputational backing, transparency, longevity, dissolution, and, finally, entrenchment.

A. Institutionalizing Commitment

Public law and legal theory scholars have long puzzled over how and why political actors engage in precommitment. 136 See Jon Elster, Don’t Burn Your Bridge Before You Come to It: Some Ambiguities and Complexities of Precommitment, 81 Tex. L. Rev. 1751, 1754 (2003) (“When precommitting himself, a person acts at one point in time in order to ensure that at some later time he will perform an act that he could but would not have performed without that prior act.”); see also Jon Elster, Ulysses and the Sirens: Studies in Rationality and Irrationality 88–103 (1979) (analyzing political precommitment by examining the phe­nomenon in various democracies); Elster, Ulysses Unbound, supra note 80, at 88–174 (analyzing the justifications for constitutions as political precommitments); John Ferejohn & Lawrence Sager, Commitment and Constitutionalism, 81 Tex. L. Rev. 1929, 1929–31 (2003) (discussing different justifications for constitutional precommitments); Jon D. Michaels, The (Willingly) Fettered Executive: Presidential Spinoffs in National Security Domains and Beyond, 97 Va. L. Rev. 801, 804–27 (2011) (analyzing case studies in how “the Executive appears to expend considerable energy to disempower itself”); Eric A. Posner & Adrian Vermeule, The Credible Executive, 74 U. Chi. L. Rev. 865, 894–913 (2007) [hereinafter Posner & Vermeule, Credible Executive] (describing “executive self-binding” mechanisms that can increase executive credibility). See generally Daryl J. Levinson, Parchment and Politics: The Positive Puzzle of Constitutional Commitment, 124 Harv. L. Rev. 657 (2011) [hereinafter Levinson, Parchment and Politics] (exploring the constitutional law implications of political precommitment and constraints). The question has understandably occupied constitutionalists in particular—those grappling with a mere “parchment” document that self-interested individuals create to bind themselves. 137 Levinson, Parchment and Politics, supra note 136, at 662. In one familiar view, the American Constitution facilitates credible commitments by establishing certain rights like property protection and institutions like judicial review, which ultimately benefit otherwise short-sighted political actors. See Ferejohn & Sager, supra note 136, at 1929 (“Governments can be more effective in pursuing their own goals by committing themselves to respecting certain rights (including property rights, of course, but other rights as well) and institutions (such as an independent judiciary).”). Another prominent perspective draws on James Madison’s notion that ambitions can “counteract” ambitions: The Constitution created three autonomous, rivalrous actors to check and balance each other in ways that preserve and ultimately expand their power. See Levinson, Parchment and Politics, supra note 136, at 669–70 (“Madison’s strategy of constitutional design was to create a set of structural arrangements that would selectively empower political decisionmakers whose interests and incentives would tend to be in alignment with constitutional rights and rules.”). Put differently, constitutions create self-entrenching powers and structures reinforced by the incentives they generate for political actors. Id. at 663 (describing the dynamic as that of “incentive compatibility”). The commitment problem arises because political actors must restrain their ambitions in order to benefit from the stability and protections constitutions can provide. Credible commitment by agency heads to their own subordinates, however, arguably implicates a different form of authority and thus a distinct set of considerations. While constitutional authority might be understood as democratic in nature 138 Max Weber refers to such authority as “rational-legal” authority.” See Max Weber, The Three Types of Legitimate Rule, 4 Berkeley Publications Soc’y & Institutions 1, 2–3 (Hans Gerth trans., 1958). —that is, concerned with the legitimate exercise of collective coercion over third parties—intra-agency authority is managerial in form. 139 See Christopher McMahon, Managerial Authority, 100 Ethics 33, 35 (1989) (“It seems clear that the conceptual apparatus of authority applies to relations between managers and employees. Employees routinely subordinate their own judgment to that of their employers.”).

Managerial authority is, for lack of a more precise word, authoritar­ian. 140 See Edward B. Rock & Michael L. Wachter, Islands of Conscious Power: Law, Norms, and the Self-Governing Corporation, 149 U. Pa. L. Rev. 1619, 1697 (2001) (“The choice between transacting in markets versus bringing transactions within firms is essentially the choice between using contracts as a governing device versus using the authority structure of the firm’s hierarchy.”); McMahon, supra note 139, at 35. It operates not from democratic premises but rather by “fiat.” 141 Cf. Stephen M. Bainbridge, The Business Judgment Rule as Abstention Doctrine, 57 Vand. L. Rev. 83, 104 (2004) (“[W]hile markets allocate resources via the price mechanism, corporations do so via fiat—i.e., authoritative direction.”); Oliver E. Williamson, The Vertical Integration of Production: Market Failure Considerations, 5 J. Reprints for Antitrust L. & Econ. 321, 325 (1973) (observing that firms can resolve conflict through “fiat”). Thus, credibility is even more tenuous in this context given the background norms and a culture of obedience within firms and agencies. Employees, for instance, are expected to “subordinate their own judgment to that of their employers” even if they might otherwise disagree with it. 142 See McMahon, supra note 139, at 35 (“It seems clear that the conceptual apparatus of authority applies to relations between managers and employees. Employees routinely subordinate their own judgment to that of their employers.”); see also Herbert A. Simon, Models of Man 184 (1957) (“W enters into an employment contract with B when the former agrees to accept the authority of the latter and the latter agrees to pay the former a stated wage (w).”). Ronald Coase contrasts the authority exercised within firms with that of market-based free exchange and negotiation; the authority within firms is directive, while the latter is bargain based. See R.H. Coase, The Nature of the Firm, 4 Economica 386, 392 (1937) (“[T]he operation of a market costs something and by forming an organisation and allowing some authority (an ‘entrepreneur’) to direct the resources, certain marketing costs are saved.”). This distinction between political and managerial authority is arguably implicit in the Administrative Procedure Act (APA). The APA subjects binding legislative rules to notice and comment but exempts rules of “agency organization, procedure, or practice.” 143 5 U.S.C. § 553(b)(3)(A) (2012). Rulemaking provisions more generally do not apply in “matter[s] relating to agency management or personnel.” 144 Id. at § 553(a)(2). Because managerial matters do not coerce third parties, the APA does not subject them to the same externally legitimizing proce­dure. 145 Id. at §§ 553(a)–(b). The authority of agency heads over their employees is assumed to be unilateral.

Of course, the concept of managerial authority needs more refine­ment in the administrative agency context. Unlike in private firms, many agency employees are not employees at will but rather tenure protected. 146 See supra note 62 and accompanying text. Their salaries, as discussed, are also politically shielded and compressed within a uniform federal pay structure. 147 See Johnson & Libecap, supra note 62, at 5; Katyal, supra note 39, at 2331–32 (discussing “civil-service personnel policies”). Many agency employees are thus not subject to traditional employment contracts. Even more so than the incomplete contracts in traditional firms, 148 See Sanford J. Grossman & Oliver D. Hart, The Costs and Benefits of Ownership: A Theory of Vertical and Lateral Integration, 94 J. Pol. Econ. 691, 716 (1986) (providing a theory of firm ownership and vertical integration in terms of “when it is too costly for one party to specify a long list of the particular rights it desires over another party’s assets”). agency heads usually cannot contract either ex ante or ex post for optimal perform­ance. As a result, the credibility problem becomes even more acute for agency heads who otherwise enjoy managerial authority.

B. Enhancing Commitment

This section thus turns to the ways in which agency heads have enhanced the credibility of their delegations vis-à-vis their subordinates. More specifically, it considers subdelegation’s institutional dimensions, which individually or in combination can serve to reassure a subordinate that her decision is less likely to be internally overruled. These design choices include specification, reputational backing, transparency, longev­ity, dissolution, and, finally, entrenchment.

1. Specification. — Subdelegations, like contracts, specify their terms and conditions. In particular, they often contain clear reservations of the agency head’s ability to overrule a delegated decision. The EPA Administrator, for instance, explicitly “retains the right to exercise or withdraw, at any time, an authority which has been delegated.” 149 EPA Delegations Manual Introduction, supra note 113, para. 2(a). By contrast, the absence of such provisions can be read (as some judges have done) as an implicit commitment to grant a subordinate the right to make a final decision. 150 See, e.g., Service v. Dulles, 354 U.S. 363, 383–87 (1957). Most of the SEC’s subdelegations, for example, declare that the SEC “hereby delegates, until the Commission orders otherwise, the [specified] functions” to various internal agency actors. 151 See, e.g., 17 C.F.R. § 200.30-1 (2016). In this manner, agency heads can explicitly or implicitly delegate their authority credibly in the text of the subdelegation.

2. Reputation. — Reputations are important, especially to repeat players. Repeat players deal continuously with the same people and can thus develop a persona (artificial or real) that can work to or against their advantage. Developing a reputation for reneging on promises, for example, can diminish the value of all future assurances by rendering them untrustworthy. 152 See Avinash K. Dixit & Barry J. Nalebuff, Thinking Strategically 142 (1991) (dis­cussing the implications that can result from broken oral promises). Thus, even if reneging provides short-term benefits, those benefits can be outweighed by the long-term costs of being unable to engage in believable future promises. Agency heads who deal repeat­edly with the same career staff thus have an incentive to develop a reputa­tion for delegating credibly. Doing so can maximize the information-generating value of existing and future subdelegations.

Put in more game-theoretic terms, assume an infinitely repeated game involving an agency head and subordinate, both of whom are risk neutral. 153 See Baker et al., supra note 134, at 56–57. During each period, the subordinate investigates a project and recommends that her boss accept or reject it. More concretely, say the subordinate is charged with investigating the facts necessary to bring an enforcement case in court. Assume that the agency head is perfectly informed but lacks the resources to develop and initiate projects. She may thus assign the investigatory task to a zealous internal lawyer. To motivate this lawyer, the agency head promises the lawyer autonomy in making enforcement decisions on behalf of the agency. If this assurance is believable, the lawyer will invest more effort in searching for suitable enforcement suits; after all, each additional hour spent on the case will translate into results the lawyer wants. 154 Id. (discussing instances in which the assurance of autonomy motivates subordinate actors).

Consequently, the payoffs resulting from this extra motivation can outweigh the expected costs of enforcement decisions that the agency head would prefer not to have made. 155 Id. Since the agency head has the information to assess an enforcement case before it is brought in court, however, she will be tempted to renege on the promise by rejecting a case not in her interest. As a result, the dominant strategy may be not to delegate, even if doing so would be more efficient. 156 Id. However, if the agency head values her reputation for delegating authority more than the benefits from reneging on her promise to ratify all proposals, then delegation becomes the dominant strategy; in this sense, reputational considerations render final delegation “feasible.” 157 Id.

But how do agency heads cultivate such reputations? One way is simply through sustained practice. The longer an agency head does not overrule her subordinate’s decisions, the more she will be perceived as allowing her staff to make final policy decisions. One illustration is the historical dynamic between the Secretary of HHS and the Commissioner of the FDA. For decades, the HHS Secretary had delegated her authority under the Food, Drug, and Cosmetics Act to the FDA Commissioner. 158 Commissioner of Food and Drugs, 38 Fed. Reg. 6668, 6668–69 (Mar. 12, 1973) (codified at 21 C.F.R. pt. 2). As a result, “[a]n unbroken practice of deference to the FDA seemed to have developed at the HHS level,” so much so that the “practice had hardened into a convention.” 159 See Adrian Vermeule, Conventions of Agency Independence, 113 Colum. L. Rev. 1163, 1208 (2013). The HHS Secretary thus had a rep­utation for not overturning the FDA Commissioner, which, as we will later see, was sullied in the context of a decision involving emergency contraceptives.

Agency heads can also develop reputations as lazy or otherwise uninterested in nitty-gritty policy details. They may focus their attention on raising public awareness or improving the agency’s standing by traveling and giving public speeches rather than engaging in day-to-day decisionmaking. Such administrators often gain reputations as delegators by default. 160 See Robert A. Katzmann, Regulatory Bureaucracy: The Federal Trade Commission and Antitrust Policy 104 (1980) (describing some commis­sioners of the Federal Trade Commission who prefer to “rely upon the recommendation of an attorney adviser or a trusted colleague with whom they generally agree” in “less-than-significant matters”). Agency heads can also cultivate reputations for nonstrategic reasons. In one Hobbesian view, the “weak bonds of words can be strengthened in two ways: a fear of the consequence of breaking one’s word; or a glory, or pride, in not breaking it.” 161 See Dixit & Nalebuff, supra note 151, at 148 (citing Thomas Hobbes, Leviathan 71 (J.M. Dent & Sons 1973) (1651)). In other words, agency heads may continue to abide by a subdelegation from a sense of pride in doing so—with the unintended consequence that doing so earns an impression as a hands-off manager confident in her professional staff. The EPA’s own delegations guidance states: “Show confidence in the redelegation decisions you make. Once decisions are redelegated, it is important to demonstrate trust, integrity, and consistency toward redele­gatees and the actions they take under the redelegations.” 162 See EPA, Delegations of Authority, supra note 29, at 4. Regardless of the underlying motivation, staff may rightly come to expect that their recommendations will remain undisturbed. 163 See, e.g., Ryan, supra note 98 (“It’s no secret that the SEC commissioners already approve nearly all staff requests for formal orders, appropriately deferring to the staff’s judgment and familiarity with the facts warranting scrutiny.”).

3. Transparency. — Another strategy agency heads can use to make their subdelegations more credible is to make them more transparent. Instead of being confined to internal staff documents, training manuals, or meeting minutes, 164 See, e.g., Frankl v. HTH Corp., 650 F.3d 1334, 1340 (9th Cir. 2011) (describing how the National Labor Relationships Board assigned petition-granting authority to the General Counsel during an internal agency meeting). internal delegations can be publicly memori­alized. Such transparency enables monitors—political officials, lobbyists, and interest groups—to become aware of the real power brokers within an agency and invest their resources accordingly. 165 See Posner & Vermeule, Credible Executive, supra note 136, at 903 (discussing “transparency as a way to reduce the costs to outsiders of monitoring . . . actions”). The resulting relation­ships, in turn, render it more likely that news of an internal revocation or reversal of authority has occurred. These revelations can then invite costly litigation or otherwise result in political sanctions.

Contrast the SEC, which routinely publishes its subdelegations in the Federal Register and Code of Federal Regulations, 166 See 17 C.F.R. §§ 200.30-1–.30-18 (2016) (identifying delegations of authority to various SEC officials). with the EPA. As discussed, the EPA records the bulk of its internal delegations in a Delegations Manual hosted on an internal server. 167 See 40 C.F.R. § 1.5(b) (2016) (“EPA’s Directives System contains definitive statements of EPA’s organization, policies, procedures, assignments of responsibility, and delegations of authority.”). The manual is unavailable to the public except through a time-consuming Freedom of Information Act (FOIA) request or else a physical trip to the EPA’s head­quarters in Washington, D.C., or one of the agency’s regional offices. 168 Id. (“Copies [of EPA’s Directives System] are available for public inspection and copying at the Management and Organization Division, 1200 Pennsylvania Ave., NW., Washington, DC 20460. Information can be obtained from the Office of Public Affairs at all Regional Offices.”). Barring such travel and time-consuming photocopying efforts, the documents were available to the author only through a FOIA request. By restricting public access in this manner, the EPA Administrator makes it more difficult for outsiders to know who is actually making decisions within the agency. Because these monitors are less likely to help enforce subdelegations, EPA staff members are more likely to believe that their decisions will be subject to internal review and potentially overruled. As a result, the subdelegations are less credible for generating resource-intensive information.

4. Longevity. — Internal delegations of authority can also vary in terms of their temporal scope. Subdelegations with and without time limits, however, have more complicated implications for credibility than it may first seem. As an initial matter, when an agency head delegates indefinitely, staff members are likely to view the delegation as more reliable relative to delegations with an expiration date. Thus, for instance, when the FCC subdelegated authority to its General Counsel to issue “written determinations on matters regarding the interception of telephone conversations,” the General Counsel had no prima facie rea­son to believe that her determinations would be second-guessed. 169 47 C.F.R. § 0.251(f) (2015).

By contrast, when the EPA Administrator issued a “temporary” delegation to a regional administrator to approve or disapprove a rulemaking petition for Wyoming aquifer exemptions, 170 See Memorandum from Lisa P. Jackson on Approval of Temporary Delegation of Authority to Propose a Rule Approving or Disapproving Certain Aquifer Exemptions in Wyoming to James B. Martin, Adm’r of Region 8, 1 (June 25, 2012) (on file with the Columbia Law Review). the regional administrator knew that her authority was time-limited to the specific petition at issue. Such an expiration date is akin to an advance notice of the power’s revocation. If a subordinate understands that her authority will expire at a given point in time, she is less likely to make long-term investments in information acquisition. At the same time, however, the subordinate may also have an even greater incentive to labor under a temporary delegation if she believes her efforts may result in the ex­tension of that authority. In other words, she may be more motivated to invest the time and resources necessary to obtain the optimally informed results if she thinks the delegation may become indefinite, especially when that information has increasing marginal returns.

To illustrate, consider once again the manner in which the SEC first delegated its authority to issue formal orders of investigation to the SEC’s Director of Enforcement. In the SEC’s own words: “The Commission is adopting this delegation for a one-year period, and at the end of the period will evaluate whether to extend the delegation.” 171 Delegation of Authority to Director of Division of Enforcement, 74 Fed. Reg. 40,068 (Aug. 11, 2009) (to be codified at 17 C.F.R. pt. 200). The subdele­gation, in other words, was to sunset after a one-year trial period. The stated purpose of this sunset provision was to “permit the Commission to evaluate the Division’s use of the delegation and to consider whether extension of the delegation was appropriate.” 172 Delegation of Authority to the Director of Its Division of Enforcement, 75 Fed. Reg. 49820 (Aug. 16, 2010) (to be codified at 17 C.F.R. pt. 200). Explicit revisitation language like this can incentivize information acquisition, even for tem­porary delegations. Indeed, after the enforcement staff almost doubled the rate of enforcement actions, the SEC’s delegation was extended indefinitely. 173 See id. The SEC gave the following rationale:
“The Commission has determined that it is appropriate to extend the Division’s authority to issue formal orders of investigation. In making this determination, the Commission considered the increased efficiency in the Division’s conduct of its investigations permitted by the delegation, and the Division’s continued effective communication and coordination in addressing pertinent legal and policy issues with other Commission Divisions and Offices when formal order authority is invoked.”

5. Dissolution. — Yet another way that delegations can be made more credible is to dissolve, even if temporarily, the body making the delega­tion in the first place. Doing so succeeds as a commitment device because it renders an action irreversible in a sense: The actor that granted the authority is no longer around to reverse it (or will not reconstitute itself for an indefinite period of time). 174 See Dixit & Nalebuff, supra note 152, at 151 (explaining that cutting off communication serves as a credible commitment device). Last wills and testaments, for instance, are credible because they cannot be revoked after death. 175 Id. Similarly, when commissions and multimember boards delegate author­ity in antici­pation of quorum-busting vacancies, these delegations are credible because the agencies will lack a sufficient number of members to revoke them. When Congress is gridlocked, the prospect of quick confirmations of new commissioners becomes even more remote. Recall, for example, the Election Assistance Commission’s delegation to its Executive Director to manage a federal voter registration form in antici­pation of commissioner vacancies. 176 Kobach v. U.S. Election Assistance Comm’n, 772 F.3d 1183, 1190 (10th Cir. 2014). The subdelegation was credible to the Executive Director because the Commission lacked a quorum to act shortly thereafter. Rely­ing on this subdelegation, the Director rejected Kansas’s and Arizona’s requests to include proof-of-citizenship requirements—a decision that was later upheld in court. 177 Id. at 1190–91.

5. Entrenchment. — Finally, agency heads can also make their subdel­egations more credible by visibly entrenching them. Entrenchment is an ambiguous concept, to be sure, 178 See Daryl Levinson & Benjamin I. Sachs, Political Entrenchment and Public Law, 125 Yale L.J. 400, 408 (2015) (“Political ‘entrenchment’ is discussed more often than it is defined, and it is not clear that any single definition captures all uses of the term.”); Eric A. Posner & Adrian Vermeule, Legislative Entrenchment: A Reappraisal, 111 Yale L.J. 1665, 1666 (2002) (“‘Entrenchment’ is a promiscuous word in the academic literature.”). but as defined here, it refers to raising the costs of reversing the delegation beyond those necessary to enact, amend, or repeal it in the first place. 179 See Mendelson, Agency Burrowing, supra note 38, at 589 (characterizing “[a]dministrative policy entrenchment” as “a decision [that] is likely to be reversible at least as a procedural matter, [but] it is probable that the change will be costly” (footnote omitted)); see also Levinson & Sachs, supra note 178, at 408 (“At the most general level, ‘entrenchment’ means that political change has been made more difficult than it otherwise would (or should) be.”). Professors Daryl Levinson and Benjamin Sachs’s definition can be understood to be concerned with the increase in political costs necessary to change something relative to the costs of the status quo—what they more generally refer to as “functional” entrenchment. Id. at 407; see also Christopher Serkin, Public Entrenchment Through Private Law: Binding Local Governments, 78 U. Chi. L. Rev. 879, 888 (2011) (“[A]n action is entrenching to the extent that it limits the policy choices available to future governments.”). In other words, agency heads can entrench a delegation by increasing the effort or expenditure necessary to repeal or otherwise change it. In this sense, entrenchment lies along a continuum from incremental increases in reversal costs to “absolute” entrenchment, when the reversal costs are essentially infinite and “the right of repeal is denied for all time, under any conditions, and by whatever procedure.” 180 Julian N. Eule, Temporal Limits on the Legislative Mandate: Entrenchment and Retroactivity, 12 Am. B. Found. Res. J. 379, 384 (1987); see also Anupam Chander, Note, Sovereignty, Referenda, and the Entrenchment of a United Kingdom Bill of Rights, 101 Yale L.J. 457, 462 (1991) (observing that “[a]bsolute entrenchment renders the entrenched rule permanent and immutable” (emphasis omitted)). At the same time, the degree of entrenchment must be ex ante visible to delegates—that is, perceivable before they decide how much to invest in expertise. Otherwise, the strategy does not effectively serve as a commitment mechanism.

As a baseline, consider that agency heads can subdelegate their authority at low cost. For example, they have historically subdelegated through highly informal means, and courts have upheld them when doing so. Take a Ninth Circuit case involving the National Labor Relations Board (NLRB). 181 Frankl v. HTH Corp., 650 F.3d 1334 (9th Cir. 2011). The National Labor Relations Act authorizes the Board to petition a federal district court for appropriate relief upon issuance of a complaint charging that any person has engaged or is engaged in an “unfair labor practice.” 182 Id. at 1340. In 2007, the NLRB subdele­gated this authority to its General Counsel—through a vote recorded only in a meeting’s minutes. 183 Id. This means of delegation had very low transaction costs and did not influence the court’s eventual conclusion that the delegation was legal. 184 Id. at 1354.

Relative to this baseline, agency heads can first entrench their delegations by increasing the procedural burden associated with amend­ing or repealing them. 185 See Magill, supra note 51, at 894 (“[A]n agency could use a self-regulatory measure to rearrange a decisionmaking process in a way that will entrench existing policy.”). Agency heads could, for example, impose an expensive internal approval process to help embed already-existing sub­delegations (i.e., those previously imposed without this costly review). To illustrate, the EPA started to require a time-consuming clearance pro­cedure before its internal delegations could be changed. 186 See EPA Delegations Manual Introduction, supra note 113, para. 3. Under this scheme, the office requesting the revised subdelegation must first consult with the EPA’s Human Capital Planning and Policy Division to ensure no conflicts. 187 Id. para 3(a). Then it must prepare a transmittal memo justifying the new delegation and specifying its scope and limitations. 188 Id. para. 3(c). The proposal must then go through an intra-agency comment process in which other interested offices review the change. 189 Id. para 3. During this process, other offices can also raise concerns and negotiate any amendments to the subdelega­tion. 190 Id. This time-consuming process can entrench preexisting delegations by rendering them more costly to revise, particularly given extant resource constraints.

Beyond raising procedural costs, agency heads can also functionally entrench a delegation by empowering actors to mobilize supporters and other interest groups to fend off subsequent attempts at repeal. 191 See Levinson & Sachs, supra note 178, at 429–30 (describing methods of “functional” entrenchment involving “strengthening political allies or weakening political opponents,” “changing the composition of the political community,” and “empowering a different governmental institution and consequently a different set of political actors and groups”); Magill, supra note 51, at 894 (noting agency heads “could empower an internal agency unit with predictable views to be in charge of the agency choice,” thus rendering it “more difficult for political opponents to oppose the effort or to dislodge it once it is in place”); Posner & Vermeule, Credible Executive, supra note 136, at 896 (referring to “informal” means of self-binding in terms of increased “political costs”). The credibility of such delegations increases because the subdelegate knows she will now have allies that can help prevent an agency head’s attempts to overrule her judgment. A revealing example is the 2011 decision by HHS Secretary Kathleen Sebelius to overrule FDA Commissioner Margaret Hamburg’s decision regarding the emergency contraceptive Plan B. 192 See Gardiner Harris, Plan to Widen Availability of Morning-After Pill Is Rejected, N.Y. Times (Dec. 7, 2011), (on file with the Columbia Law Review). Commissioner Hamburg had labored under subdelegated authority 193 Delegations from the Secretary of Health and Human Services to the Commissioner of Food and Drugs, 21 C.F.R. § 5.10(a)(1) (2004); accord 2 FDA, SMG 1410.10, FDA Staff Manual Guides—Delegations of Authority para 1(A)(1) (2016),
downloads/AboutFDA/ReportsManualsForms/StaffManualGuides/UCM273771.pdf (on file with the Columbia Law Review).
to determine that Plan B was safe and effective for over-the-counter use. For decades, as discussed, the HHS Secretary had explicitly subdelegated the authority to make such determinations under the Food, Drug, and Cosmetic Act to the FDA Commissioner. 194 See supra notes 157–158 and accompanying text (discussing the historical dynamic between the HHS Secretary and the FDA Commissioner). As a result, the FDA Commissioner had a number of allies and interest groups ready to support her conclu­sions. 195 See Cole Petrochko, Groups Ask HHS to Rethink Plan B Ruling, Medpage Today (Dec. 21, 2012), [
A566-BWPN] (describing various groups’ opposition to the HHS Plan B decision).
Indeed, the ensuing public outcry and eventual litigation that resulted in the overturning of Secretary Sebelius’s decision attest to the functionally entrenched nature of the subdelegation. 196 See Tummino v. Hamburg, 936 F. Supp. 2d 162, 170 (E.D.N.Y. 2013) (describing the unprecedented nature of the Secretary’s action in “overru[ling] the FDA in an area which Congress entrusted primarily to the FDA . . . and which fell within the scope of the authority that the Secretary expressly delegated to the Commissioner”).

C. Applications

This Part has thus far developed some of the conceptual resources necessary to advance a positive theory for why and how agency heads subdelegate. The main argument has been that the extent to which an agency head credibly commits to a delegation is a function of two main variables: the expected preference divergence with the subdelegate and the relative transaction costs of internal advice giving. Thus, one would expect to see institutional variations in credibility as a result of changes along either dimension, holding all else constant. While the previous discussion used a variety of examples to illustrate the plausibility of these claims, this section now seeks to apply the theory in further contexts and explore some resulting hypotheses.

Perhaps most obviously, partisan changes in presidential administra­tions should result in changes to the credibility of previously delegated authorities. When Democrat-appointed agency heads in agencies with left-leaning careerists—such as the EPA, HHS, and Department of Labor—are replaced by Republican appointees, 197 See Joshua D. Clinton & David E. Lewis, Expert Opinion, Agency Characteristics, and Agency Preferences, 16 Pol. Analysis 3, 4 (2008) (displaying agencies on a left–right spectrum using a method for measuring agency preferences based on expert surveys and a “multirater item response model to jointly analyze the responses and objective information about agency characteristics”). for example, one would expect either less credible subdelegations or their revocation altogether. Alter­natively, one might expect new agency heads to redelegate their authority to other agents within the agency with closer preferences.

To illustrate, return again to the Clinton Secretary of Commerce’s proposed rule to grant power to the Census Director to use statistical sampling when determining population measures. Recall that such sam­pling was known to likely help Democrats given that it would result in greater counts of minorities. Despite an adverse court ruling and opposi­tion from congressional Republicans, the Clinton Administration finalized the rule on October 6, 2000—shortly before the presidential election that year. 198 Report of Tabulations of Population to States and Localities Pursuant to 13 U.S.C. 141(c), 15 C.F.R. § 101.1 (2001). The rule was published in the Federal Register and stated that “[t]he Director of the Census shall make the final determination” regarding population measures. 199 Id. § 101.1(a)(1). It further specified that “[t]he determination of the Director of the Census shall not be subject to review, reconsideration, or reversal by the Secretary of Commerce.” 200 Id. § 101.1(a)(4). In this manner, a Democratic agency head subdelegated final authority in a clearly specified and transparent manner to a subordinate he knew shared his preferences. Doing so helped to ensure that the Director invested effort in the determination.

After the 2000 presidential contest resulted in George W. Bush’s election, however, a new fleet of Republican political appointees swept into power. In particular, President Bush chose Donald Evans to be his Secretary of Commerce, a loyalist who had served as chairman of Bush’s presidential campaign. 201 See Don Evans: Former Secretary of Commerce, White House, [] (last visited Oct. 20, 2016). On February 23, 2001, Secretary Evans issued a new final rule revoking the subdelegated authority. 202 Report of Tabulations of Population to States and Localities Pursuant to 13 U.S.C. 141(c), 15 C.F.R. § 101.1 (2002). His preferences were now misaligned with those of the acting Census Director, a career civil servant. 203 Steven A. Holmes, Census Power Is Returned to Secretary of Commerce, N.Y. Times (Feb. 17, 2001), (on file with the Columbia Law Review) (“Under the Clinton regulation that Mr. Evans rescinded, the head of the Census Bureau—an acting director who is a career civil servant—would have had to make a decision whether to adjust the population count.”). Bush’s appointment to the position would not be confirmed until more than a year later, meaning that a preference divergence between Secretary Evans and the acting Census Director would persist for some time. 204 See Tom Mesenbourg, Remembering Louis Kincannon U.S. Census Bureau: Census Blogs (Dec. 17, 2012),
12/remembering-louis-kincannon.html [] (“President George W. Bush nominated Louis for director of the Census Bureau in November 2001, and the Senate confirmed him unanimously on March 13, 2002.”).
Perhaps not surprisingly, Secretary Evans finalized the revocation less than two weeks before the Census Director was expected to make his decision regarding the use of statistical sampling. 205 Holmes, supra note 203. The Bush Office of Legal Counsel also advised Secretary Evans that revoking the subdelegation would not require submitting to APA notice-and-comment procedures, even though the Clinton Admini­stration had earlier used such procedures to adopt the rule. 206 Applicability of APA Notice and Comment Procedures to Revocation of Delegation of Authority, 25 Op. O.L.C. 99, 99 (2001). This episode illustrates how new political appointees can revoke delega­tions to subordinates no longer expected to share their preferences.

Turning now to the second dimension regarding internal trans­action costs, exogenous variations in those costs would be expected to influence the agency head since subdelegation can eliminate those transaction costs. The agency head will no longer have to spend limited resources reviewing staff recommendations and can instead farm out the decisionmaking altogether. This perspective helps to illuminate the ways in which intra-agency inefficiency can explain final subdelegation as a practice. The more these transaction costs increase, holding other factors constant, the more likely an agency head is to subdelegate with finality to eliminate those costs. These dynamics become even more pronounced in light of agencies’ fixed budget constraints, which exert pressure to use organizational resources more cost-effectively.

Indeed, this general insight has been used to explain, with plausible empirical support, why Congress delegates to executive agencies rather than subdelegating internally to its committees to pass legislation. 207 See Epstein & O’Halloran, supra note 5, at 7–9 (arguing that “when deciding where policy will be made, Congress trades off the internal policy production costs of the committee system against the external costs of delegation”). In Professors David Epstein and Sharon O’Halloran’s theory, the median legislator delegates to the executive branch to avoid the transaction costs associated with Congress’s internal committee system, as long as her preferences are sufficiently aligned with those of the President. Such committees contribute to inefficient delays, holdups, and logrolling. So within some range of expected preference divergence, the median legis­lator would prefer to delegate a policy to an administrative agency rather than attempt to pass a bill to accomplish her desired end. 208 See id. at 46–47 ( “[P]olitical governance structures should minimize the political transaction costs associated with the implementation of a given policy.”).

This comparative-transaction-cost theory is similarly illuminating in the agency-subdelegation context. Here, the internal transaction costs associated with a reviewable subdelegation include both the costs that arise as a result of expected preference divergence with the staff member (principal–agent dynamics) as well as those that arise even when prefer­ences are aligned (team-theory issues). 209 See David A. Weisbach & Jacob Nussim, The Integration of Tax and Spending Programs, 113 Yale L.J. 955, 984 (2004) (characterizing team theory as an approach that “assume[s] away any divergence of preferences among individual agents”). While the principal–agent dynamics have already been explored, it is also important to consider the implications when agency heads and staff members are on the same team—that is, when they have the same preferences. In these situations, transaction costs arise as a result of the need to coordinate bureaucratic advice giving.

Perfect communication is difficult enough between two individuals, let alone within large organizations like administrative agencies. In these settings, it is rare for low-level career staff to meet directly with agency heads. Rather, staff recommendations are usually filtered through mul­tiple levels of an internal hierarchy. Supporting information, in turn, is distilled into bullet points and briefings instead of conveyed in its full and lengthy form as reports or academic papers. Consequently, as dis­cussed, reviewable subdelegations require costly internal structures and processes to coordinate the necessary review. 210 See Nou, supra note 11, at 435–40 (arguing that “internal managerial forms” are developed to minimize information-processing costs). It is thus understandable why agency heads would seek to avoid these information-processing costs altogether by delegating a decision when the expected biases are suffi­ciently small. The agency head can essentially get the same bang for less buck.

As a result, the higher the internal transaction costs, the more one would expect to see subdelegations increase on net. This dynamic may, for instance, help to explain why the bulk of agency subdelegations are technical in nature. Of course, these issues are also the areas in which the information asymmetry between agency heads and more expert subordi­nates is the greatest. But the internal transaction costs associated with reviewing such decisions are also especially high. They are high due to what Professors Jacob Gersen and Adrian Vermuele refer to as “tacit expertise,” information possessed by experts that is “costly to transmit to nonexperts[] and which is always distorted in the transmission.” 211 See Jacob Gersen & Adrian Vermeule, Thin Rationality Review, 114 Mich. L. Rev. 1355, 1357 (2016). Because technical information is difficult to compile and communicate to more generalist audiences like political appointees, one would expect agency heads to delegate more in these areas to reduce internal transaction costs.

III. Implications

This Part now takes a step back to evaluate the normative implica­tions of this information-forcing theory of agency subdelega­tion. 212 To be sure, the theory would benefit from further empirical testing to confirm, reject, or revisit it, should the relevant data become available in a more useable form. See supra note 46 (describing the difficulties of collecting data on agency internal delegations). The normative implications here are explored should the associated hypotheses indeed be borne out by such analyses. The previous analysis argued that internal delegation, at its core, presents a trade-off for the agency head between better information, on the one hand, and political control, on the other. Where this balance is struck will depend on the administrator’s preferences, those of her subordinates, and the transaction costs of internal review. What is in the agency head’s interest, however, is not always best for the administrative state. Indeed, much of administrative law is designed to cabin the discre­tion of bur­eaucratic actors to conform to a legal regime that itself struggles to balance accountability with agency expertise. This Part will thus examine how courts and political actors can help to optimize the acquisition of high-quality information in the administrative state.

A. Courts

The first challenge from a normative standpoint is to ask how courts can encourage research incentives without sacrificing the requisite amount of democratic accountability. 213 Judges, no less than other government officials (or human beings), also have their own personal preferences and biases. See Lee Epstein, William M. Landes & Richard A. Posner, The Behavior of Federal Judges 5 (2013) (modeling judges as “motivated and constrained, as other workers are, by costs and benefits both pecuniary and nonpecuniary”). Thus, a full model would incorporate this insight to consider what incentives judges would have to adopt these doctrinal innovations in the first place. The more legalistic analysis here, by contrast, assumes that judges have an interest in vindicating the normative goals reflected in existing doctrines and case law. What is “requisite,” of course has no easy answer, but the first section will explore the relevant constitu­tional constraints with respect to executive power. It will ultimately argue for a more functional understanding of political control attuned to the nature of the subdelegation. The second section will then turn to issues of statutory interpretation and, specifically, how courts should read statutes to determine whether subdelegation is congressionally author­ized. The final section will then turn to the possibility of courts further helping to encourage credibility when externally enforcing subdelega­tions under the Accardi doctrine—the basic principle that an agency must abide by its own rules. 214 See sources cited supra note 54.

1. Choosing Subdelegates. — Subdelegation raises constitutional worries since agency heads may entrust significant duties to subordinates with attenuated relationships to the President. The core concern is that the President, in whom the Constitution vests the “executive Power” 215 U.S. Const. art. II, § 1, cl. 1. and who must “take Care” to “faithfully execute” the laws, 216 Id. art. II, § 3. will lose con­trol of an unelected bureaucracy. To mitigate this possibility, the Appointments Clause and other constitutional provisions ensure that the President is able to hire loyalists in key positions and fire insubordi­nates. 217 Id. art. II, § 2, cl. 2. At the same time, Congress may sometimes grant the ability to appoint certain officials to other actors besides the President as well as place restrictions on their removal. 218 Id. The potential for litigation arises when agency heads redelegate their powers in ways that seem to contravene these constitu­tional mandates.

Specifically, the Appointments Clause requires the President to nominate, and the Senate to confirm, all principal officers. 219 Id. (“[The President] shall nominate . . . and with the Advice and Consent of the Senate, shall appoint . . . all other Officers of the United States . . . but the Congress may by Law vest the Appointment of . . . inferior Officers . . . in the President alone, in the Courts of Law, or in the Heads of Departments.”). Granting the President this authority allows her to employ the individuals most likely to carry out her agenda dutifully. Recognizing that placing the burden of appointing every government official on the President is impractical, however, the Clause also allows Congress to modify this requirement for inferior officers by assigning appointment power in “the President alone,” “Courts of Law,” or in “Heads of Departments.” 220 Id. All other nonofficers—employees—in the executive branch can be selected by other government officials or through processes like those currently governing the competitive civil service. 221 For a discussion of the competitive civil service selection process, see supra note 58.

The Supreme Court has yet to settle on an “exclusive criterion” 222 Edmond v. United States, 520 U.S. 651, 661 (1997); Morrison v. Olson, 487 U.S. 654, 671 (1988) (“The line between ‘inferior’ and ‘principal’ officers is one that is far from clear, and the Framers provided little guidance into where it should be drawn.”). for drawing the line between principal and inferior officers, but one starting point is to ask whether the officer in question is subordinate to “some higher ranking officer.” 223 Edmond, 520 U.S. at 662–63. In Justice Scalia’s words: “Whether one is an ‘inferior’ officer depends on whether he has a superior.” 224 Id. at 662. Indeed, in Edmond v. United States, the judges on the Coast Guard Court of Criminal Appeals were inferior officers because their work was “directed and supervised at some level by others who were appointed” by the President and Senate-confirmed. 225 Id. at 663, 666. The judges could not render a final decision unless it survived review by the Court of Appeals for the Armed Forces, subject to procedures set forth by the Judge Advocate General. 226 Id. at 664–65.

While Edmond could be read categorically, more functional readings are also possible, especially when reconciling it with previous cases in which the Court used a more flexible, multifactor approach. In Morrison v. Olson, for example, the Court held that an independent counsel—appointed by a special panel of federal judges and removable by the Attorney General—was an inferior officer. 227 487 U.S. at 671. The Court largely based its conclusion on four factors: the circumscribed scope of the counsel’s duties, the restricted jurisdiction of the counsel’s office, the limited duration of appointment, and the fact that the counsel was removable for cause. 228 Id. at 671–72. As a general matter, then, courts scrutinize both the nature of the exercised duties and the mechanisms of internal control.

While subdelegations run to inferior officers, they often implicate nonofficers, or employees, as well. The Supreme Court has characterized this group as those who do not exercise “significant authority” 229 Cf. Buckley v. Valeo, 424 U.S. 1, 126 (1976). and usually perform “ministerial” tasks. 230 Cf. Freytag v. Comm’r, 501 U.S. 868, 881 (1991). In Freytag v. Commissioner, for example, the Court found special trial judges of the United States Tax Court to be officers, not employees. 231 Id. The trial judges were appointed by the Chief Judge of the Tax Court. 232 Id. Because these judges “take testimony, conduct trials, rule on the admissibility of evidence, and have the power to enforce compliance with discovery orders,” and thus exercise “significant discretion,” the Court held that they had to be appointed in accordance with the Appointments Clause. 233 Id. at 881–82. At the same time, the Court has also suggested that the analysis turns on the extent to which such discretion is controlled by or subordinated to that of an officer. 234 Buckley, 424 U.S. at 126 n.162 (distinguishing employees who are “lesser functionaries subordinate to officers of the United States” from officers who “are not subject to the control or direction of any other executive, judicial, or legislative authority”).

Similar considerations also come into play when analyzing the agency head’s ability to delegate to internal actors protected by congres­sional for-cause removal restrictions. The Supreme Court con­tinues to allow such restrictions based on evolving tests regarding the functional effects of those constraints. While earlier cases tended to focus on whether the official was acting in an executive, as opposed to “quasi-judicial” or “quasi-legislative,” matter, 235 See Humphrey’s Ex’r v. United States, 295 U.S. 602, 624 (1935). more recent cases suggest that the relevant analysis turns on the extent to which the removal restrictions impede the President’s ability to perform his duties. 236 As the Morrison Court put it, the “real question” is “whether the removal restrictions are of such a nature that they impede the President’s ability to perform his constitutional duty.” See Morrison v. Olson, 487 U.S. 654, 691 (1988); see also Free Enter. Fund v. Pub. Co. Accounting Oversight Bd., 561 U.S. 477, 498 (2010) (striking down a dual for-cause provision on the grounds that it hindered “the President’s ability to ensure that the laws are faithfully executed—as well as the public’s ability to pass judgment on his efforts”). Though Free Enterprise Fund analyzes the effects of the dual for-cause provision on the President’s functional ability to execute the laws, the analysis as a whole is arguably more formalistic in that it places great emphasis on the difference between single versus double layers of insulation. Id. This is an indeter­minate standard, to be sure, but in Morrison, the Court upheld for-cause removal restrictions on an independent counsel exercising purely execu­tive prosecutorial functions. 237 Morrison, 487 U.S. at 691–93; see also Kent H. Barnett, Avoiding Independent Agency Armageddon, 87 Notre Dame L. Rev. 1349, 1356–58 (2012). As in the appointments context, the Court noted that the Attorney General still exercised control over the indepen­dent counsel by retaining “‘good cause’” removal power, nominating candidates for the position, and declining to appoint anyone if no inves­tigation seemed warranted. 238 Morrison, 487 U.S. at 696 (quoting 28 U.S.C. § 596(a)(1) (1982 & Supp. V 1987)).

In this manner, courts have interpreted the Constitution to con­strain the extent to which government actors can exercise authority without sufficient political accountability. As a result, agency heads are restrained in terms of the duties they can subdelegate. Before doing so, agency heads must consider the ways in which such lower-level officials are appointed, as well as the relevant removal restrictions, and calibrate the amount of delegated discretion accordingly. Both inquiries, as demonstrated above, can turn on the amount of internal control exer­cised by an accountable superior, as well as the nature of the duties exercised. The more internal control, the broader the delegated duties can be and the more likely such tasks can be performed by, for example, civil servants or other subordinates with tenure protections.

In order to enhance agency heads’ ability to encourage more infor­mation investment, however, courts should also consider the credibility of the subdelegation to assess the amount of intra-agency control exer­cised over the final decision. The less credible the delegation, the more functional control the agency head should be understood to exer­cise. Conversely, the more credible the delegation, the less control the agency head should be understood to exert. Thus, for example, the less trans­parent and more temporary (i.e., less credible) the delegation, the more courts should assume that agency heads have more power to review and reverse the subdelegation. Doing so would help judicial doctrine more closely align with institutional realities.

Some D.C. Circuit decisions regarding the Appointments Clause help to illustrate this approach. In these cases, the D.C. Circuit has used the finality of an official’s decision as the main factor in determining whether a superior exercised sufficient control. 239 In Landry v. FDIC, for example, the D.C. Circuit held that ALJs who could only recommend agency action were employees, and not constitutional officers, because they could not make final decisions. 204 F.3d 1125, 1132–34 (D.C. Cir. 2000). Based on this holding, as previously mentioned, the D.C. Circuit in Raymond J. Lucia Cos. v. SEC also used the lack of finality in ALJ decisions to hold that ALJs were employees and could thus exercise the SEC’s delegated authority. 832 F. 3d 277, 284–89 (D.C. Cir. 2016). The D.C. Circuit is currently considering Lucia en banc. Raymond J. Lucia Cos. v. SEC, 832 F. 3d 277 (D.C. Cir. 2016), vacated and reh’g en banc granted, 832 F. 3d 277 (D.C. Cir. 2017). Unlike presidentially appointed and Senate-confirmed Article III judges, ALJs are hired by agencies as “necessary” for the agency to conduct formal adjudications. 5 U.S.C. § 3105 (2006); see also Kent Barnett, Resolving the ALJ Quandary, 66 Vand. L. Rev. 797, 804 (2013). Agencies select ALJs from a list prepared by the Office of Personnel Management (OPM) according to various criteria, including experience as an attorney, veteran status, and so on. Id. at 804–05. For a contrary view, see Bandimere v. SEC, No. 15-9586, 2016 WL 7439007 (10th Cir. Dec. 27, 2016). A subdelegation that explicitly specifies final decisionmaking authority, of course, is more credible than those that do not. Thus, by using finality to assess whether an agency head sufficiently supervises a subordinate for the purposes of characterizing that subordinate as an employee or inferior officer, the D.C. Circuit’s doctrine reflects practical intra-agency dynamics.

One consequence of this refinement, however, would be that the recipient of a final subdelegation would be more likely to be deemed an inferior (or principal) officer rather than an employee (or inferior offi­cer), thereby narrowing the class of constitutionally appointed subdele­gates. Similarly, because the agency head exercises less control, any removal restrictions on the recipient of the subdelegation would be more suspect. To mitigate this implication and allow the agency head the most discretion to choose the best-informed subdelegate, courts should also scrutinize the nature of the duty involved to determine whether it capital­izes on the delegate’s expertise. If so, then judges should be more willing to allow authority to be exercised by tenure-protected officials or ap­pointed by a wider array of actors. To illustrate, agency heads some­times internally delegate rulemaking authority, which is quasi-legislative and thus constitutes a substantial exercise of power given the ability to bind general classes of individuals. When the subdelegation is expertise based, such as the FCC’s grant of rulemaking authority to its Chief of the Office of Engineering and Technology regarding spectrum standards, 240 47 C.F.R. § 0.241(a)(1)(ii) (2015) (“The Chief of the Office of Engineering and Technology is delegated authority, by notice-and-comment rulemaking if required by statute or otherwise in the public interest, to issue an order amending rules . . . that reference industry standards to specify revised versions of the standards.”). then courts should be more willing to uphold the subdelegation scheme.

2. Statutory Interpretation. — Turning from constitutional to statutory constraints, courts currently treat internal and external agency delega­tions asymmetrically. While delegations to an internal actor are presump­tively valid absent express statutory proscription, those to actors outside of the agency are not. 241 La. Forestry Ass’n v. Sec’y of U.S. Dep’t of Labor, 745 F.3d 653, 671 (3d Cir. 2014) (recognizing the “prohibition against subdelegation to an outside entity in the absence of express congressional authorization”); U.S. Telecom Ass’n v. FCC, 359 F.3d 554, 565 (D.C. Cir. 2004) (“[T]he cases recognize an important distinction between subdelegation to a subordinate and subdelegation to an outside party. The presumption that subdelegations are valid absent a showing of contrary congressional intent applies only to the former.”); Shook v. D.C. Fin. Responsibility & Mgmt. Assistance Auth., 132 F.3d 775, 783–84 (D.C. Cir. 1998) (“[I]t would be unusual, if not unprecedented, for Congress to authorize the Board of Education to delegate its own governing authority, its policymaking function, to another outside multi-member body.”); Assiniboine & Sioux Tribes of Fort Peck Indian Reservation v. Bd. of Oil & Gas Conservation, 792 F.2d 782, 796 (9th Cir. 1986) (expressing “reluctan[ce] to read broad authority to subdelegate [to an external actor] . . . absent clear proof of legislative intent”). In other words, when statutes are otherwise silent, judges read such silence to permit internal subdelegation but to prohibit delegation to an external entity—whether another agency, private party, or a state. 242 See F. Andrew Hessick & Carissa Byrne Hessick, The Non-Redelegation Doctrine, 55 Wm. & Mary L. Rev. 163, 191–93 (2013); Marisam, supra note 48, at 891 (terming the prohibition against external delegations “the anti-redelegation doctrine”). The relevant cases usually justify this approach with one of two rationales. First, accountability for internal delegations purportedly remains with the agency head, while external delegations “blur” the lines of responsibility. 243 See U.S. Telecom Ass’n, 359 F.3d at 565–66. Second, delegations to external entities increase the likelihood that the resulting discretion will be exercised by actors with different interests; as one court put it, they “aggravate[] the risk of policy drift inherent in any principal-agent relationship.” 244 Id.

While these justifications may be true as a relative matter (an open empirical question), it is worth pausing before accepting them at face value. Simply because an agency is formally within another, for example, does not ensure that the lines of accountability will remain intact. The Plan B fiasco, for example, erupted precisely because many presumed that the FDA was independent of HHS despite being within the same agency. 245 See Vermeule, supra note 159, at 1207–09. Few had historically held HHS accountable for the FDA’s determinations as a result. The same is true of the OLC, which is housed within the Department of Justice (DOJ) and provides “authoritative legal advice” to the President and executive agencies. 246 Office of Legal Counsel, U.S. Dep’t Justice, [http://] (last visited Oct. 20, 2016).
Though laboring under subdelegated authority from the DOJ’s Attorney General, 247 28 C.F.R. § 0.25 (2016) (granting authority to “[r]ender[] opinions to the Attorney General and to the heads of the various organizational units of the Department on questions of law arising in the administration of the Department”). OLC is often perceived as an independent entity responsible for its own actions. 248 See Vermeule, supra note 159, at 1209. Efforts of the DOJ or President to pressure or overrule the OLC would likely result in a public outcry. 249 Id. at 1210 (“[T]he convention of OLC independence is enforced by anticipation of political consequences for breaching the convention.”). As these examples reveal, it is not always the case that internal delegations preserve accountability at the top or ensure against policy drift.

Doctrinal innovations, however, could help the reality more closely match the judicial rhetoric—that is, could facilitate more hierarchical accountability while also optimizing investment in expertise. Specifically, courts could calibrate Chevron deference to the crediblity of the delega­tion, thereby encouraging agency heads to delegate in a manner that both allows for more public input and motivates costly information acquisi­tion. 250 Chevron, of course, provides that judges must defer to an agency’s reasonable construction of a statutory ambiguity when the statute itself evinces a legislative intent to delegate that interpretive authority. Chevron U.S.A. Inc. v. Nat. Res. Def. Council, Inc., 467 U.S. 837, 844 (1984). Its two-step test is a familiar one: First, the judge must ask “whether Congress has directly spoken to the precise question at issue.” Id. at 842. If Congress’s intent is “clear,” then that intention governs, but if the statute is ambiguous or silent, then in step two, courts ask whether the agency’s interpretation is “permissible” and, if so, defer accordingly. Id. at 842–43. Indeed, courts currently extend Chevron deference to interpretive decisions to subdelegate statutory authority internally. 251 See, e.g., Halverson v. Slater, 129 F.3d 180, 181 (D.C. Cir. 1997) (holding that “[p]ursuant to the first step of the familiar Chevron analysis, . . . [the act] limits delegation” to a third party based on the text of the statute). However, agen­cies do not appear to be calibrating their analysis on the form of the delegation. Recall, for example, the Ninth Circuit decision upholding an NLRB delegation to its General Gounsel recorded only in internal meeting minutes. 252 Frankl v. HTH Corp., 650 F.3d 1334, 1340 (9th Cir. 2011). The court engaged in an interpretation of the statute under Chevron’s familiar two steps but did not consider the highly informal nature of the mechanism through which authority was granted. 253 Id. at 1347–54.

By failing to do so, the Ninth Circuit did not apply the Mead doctrine, which holds that Chevron deference is due when Congress has delegated authority “to make rules carrying the force of law,” and the agency has acted pursuant to that authority when interpreting the stat­ute. 254 See United States v. Mead, 533 U.S. 218, 226–27 (2001). More specifically, Mead conditions deference on the extent to which Congress provides for a “relatively formal administrative procedure” that fosters “fairness and deliberation,” such as notice-and-comment rulemaking or formal adjudication. 255 Id. at 230. Such qualities of open deliberation would serve to enhance the credibility of subdelegations. Internal delega­tions issued after public comment, for example, would alert interest groups that could help facilitate the entrenchment of the delegation.

More transparent grants of authority also encourage internal agency actors to rely on the delegation and apply their expertise accordingly. Thus, courts should extend Chevron deference only when subdelegations are promulgated in this manner. By contrast, courts should scrutinize under a less deferential Skidmore standard those subdelegations that are merely contained in nonpublic agency staff manuals, meeting minutes, or other informal documents. Under Skidmore, courts retain primary interpretive authority but consider whether the agency has exhibited the “power to persuade,” usually by virtue of its expertise and experience administering the stat­ute. 256 Skidmore v. Swift & Co., 323 U.S. 134, 140 (1944). As such, agency heads would have an incentive to engage in more credible subdelegations since doing so would promise more Chevron-based deferential judicial review. 257 See id. at 138. Skidmore considered an amicus brief filed by the Department of Labor’s Administrator of the Wage and Hour Division, who had issued an “interpretative bulletin” containing a standard for calculating working time. Id. The Court held that “[t]he weight of such a judgment in a particular case will depend upon the thoroughness evident in its consideration, the validity of its reasoning, its consistency with earlier and later pronouncements, and all those factors which give it power to persuade, if lacking power to control.” Id. at 140. In Barnhart v. Walton, the Court considered the Social Security Administration’s interpretation of the Social Security Act, first through a series of informal means and then through notice-and-comment rulemaking. 535 U.S. 212 (2002). In holding that Chevron applied, the Court explained that deference was due depending on “the interpretive method used and the nature of the question at issue.” Id. at 222. As applied to the case at hand, the inquiry could include a number of factors:
“In this case, the interstitial nature of the legal question, the related expertise of the Agency, the importance of the question to administration of the statute, the complexity of that administration, and the careful consideration the Agency has given the question over a long period of time all indicate that Chevron provides the appropriate legal lens through which to view the legality of the Agency interpretation here at issue.”

3. Judicial Enforcement. — Courts can also play a stronger role than they currently do to enforce subdelegations by creating more bright-line ex ante rules for agency heads seeking to credibly commit. The Accardi doctrine generally requires agencies to follow their own rules, including procedural rules regarding internal delegation. 258 See sources cited supra note 54. In the paradigm cases relevant here, a lower-level agency official acts pursuant to a subdele­gated power. When the agency head overrules or otherwise reverses the subdelegate’s decision, an adversely affected litigant then brings suit citing the delegation. The litigant argues that the agency head has violated Accardi by failing to follow its own procedural rule. 259 See, e.g., Service v. Dulles, 354 U.S. 363, 372 (1957); Chevron Oil Co. v. Andrus, 588 F.2d 1383, 1385–86 (5th Cir. 1979); Skokomish Indian Tribe v. Gen. Servs. Admin., 587 F.2d 428, 431 (9th Cir. 1978). In these circumstances, courts will look at the form and language of the internal delegation to determine whether or not that rule should be enforced. 260 See Merrill, The Accardi Principle, supra note 54, at 589–90 (describing as a recurring issue in Accardi cases “the need to determine the meaning of an agency regulation”); id. at 596–603 (discussing the basis of Accardi’s application to legislative rules).

Both of these dimensions, however, currently serve as sources of ex ante uncertainty for the subordinate acting pursuant to delegated authority. First, take the form of the subdelegation: Accardi has particular bite with legislative rules. 261 United States ex rel. Accardi v. Shaughnessy, 347 U.S. 260, 260 (1954); see also Magill, supra note 51, at 878; Merrill, The Accardi Principle, supra note 54, at 603 (“[T]he Accardi principle applies only to legislative regulations because only legislative regulations create binding legal duties on agencies and agency personnel.”). Legislative rules are those rules that are legally bind­ing on the agency, courts, and the public. They are generally required to go through notice and comment. 262 See Robert A. Anthony, “Interpretive” Rules, “Legislative” Rules and “Spurious” Rules: Lifting the Smog, 8 Admin. L.J. Am. U. 1, 3 (1994) (referencing APA notice-and-comment procedures); David L. Franklin, Legislative Rules, Nonlegislative Rules, and the Perils of the Short Cut, 120 Yale L.J. 276, 277 (2010) (same). Nonlegislative rules, by contrast, merely clarify rather than create new obligations; they are exempt from notice and comment. 263 Nonlegislative rules are often also referred to as “guidance documents.” See, e.g., Nina A. Mendelson, Regulatory Beneficiaries and Informal Agency Policymaking, 92 Cornell L. Rev. 397, 399 (2007) (“Guidance documents can closely resemble legislative rules, leading some to call them ‘nonlegislative rules.’”). Nonlegislative rules are the bread and butter of everyday agency life: internal memoranda, guidance documents, inter­pretive rules, press releases, and staff training manuals, among others. If an agency head uses a legislative rule to subdelegate a power, a court is likely to enforce it, especially when the litigant can show prejudice. 264 Magill, supra note 51, at 879 (“[A]n agency is bound to follow its rules that affect the rights of individuals where substantial prejudice results from the violation of those rules.”). Thus, as Professor Magill has observed, this aspect of the “doctrine means that the agency can opt into court enforcement of the rule in the future and therefore make its self-limitation more credible in the first instance.” 265 See id.; see also Elizabeth Magill & Adrian Vermeule, Allocating Power Within Agencies, 120 Yale L.J. 1032, 1064–65 (2011) (observing that the Accardi doctrine helps “top-level agency officials . . . to control delegations of power within the agency”). In theory at least, agency heads could convincingly commit to abide by an internal delegation through a legislative rule.

In practice, however, the analytical distinction between legislative and nonlegislative rules is notoriously “tenuous” and “hazy.” 266 See Am. Hosp. Ass’n v. Bowen, 834 F.2d 1037, 1045–46 (D.C. Cir. 1987) (describing the “spectrum” between clearly interpretive and clearly substantive as a “hazy continuum”); Chisholm v. FCC, 538 F.2d 349, 393 (D.C. Cir. 1976) (“The distinction between an interpretative rule . . . and a legislative rule . . . is often tenuous.”); Franklin, supra note 262, at 287–88 (explaining the difficulty of distinguishing legislative and nonlegislative rules based on APA definitions). A brief survey of the various tests used by lower courts reflects their muddled and oft-conflicting nature. Some courts, for example, rely on the agency’s own characterization, while others apply a more objective test examining the rule’s practical impact; still others ask whether the challenged rule creates new legal obligations instead of simply clarifying previous ones. 267 Franklin, supra note 262, at 287–88. A related inquiry asks whether the agency intended the rule to be binding on third parties and, thus, a legislative rule, as opposed to only internally binding on the agency’s own employees. 268 See Vietnam Veterans of Am. v. Sec’y of Navy, 843 F.2d 528, 538 (D.C. Cir. 1988) (stating that courts have determined what types of “substantive agency statements” qualify for a “‘policy statement’ exemption” from notice and comment “on the basis of the agency’s intent to be bound”); Final Bulletin for Agency Good Guidance Practices, 72 Fed. Reg. 3432, 3437 (Jan. 25, 2007) (“[W]hile a guidance document cannot legally bind, agencies can appropriately bind their employees to abide by agency policy as a matter of their supervisory powers over such employees without undertaking pre-adoption notice and comment rulemaking.”). The D.C. Circuit has held that objective indicia of intent include whether there is an adequate legislative basis for the agency action without the rule, the rule is published in the CFR, the agency has invoked its general legislative authority, or the rule effectively amends a previous legislative rule. 269 Am. Mining Cong. v. Mine Safety & Health Admin., 995 F.2d 1106, 1112 (D.C. Cir. 1993). But see Health Ins. Ass’n of Am. Inc. v. Shalala, 23 F.3d 412, 423 (D.C. Cir. 1994) (stating that CFR publication is no more than “a snippet of evidence of agency intent”). If the answer to any of these questions is affirmative, then judges are more likely to deem the rule legislative. 270 Am. Mining Cong., 995 F. 2d at 1112.

Beyond these factors, other courts look at the extent to which the agency has actually treated the rule as binding in enforcement pro­ceedings or other litigation. This inquiry is complicated by the fact that most agencies use boilerplate language renouncing the binding effect of a rule. 271 Id. at 1111. This doctrinal “smog” thus complicates an agency head’s sub­dele­gation efforts. 272 See id. at 1108–09. That said, the cases do suggest that agency heads can take specific actions to increase the likelihood of judicial enforcement, thereby increasing the subdelegation’s internal credibility. For example, the agency head could subject the subdelega­tion to notice and comment (even if not required by the APA), 273 5 U.S.C. § 553(c) (2012) (stating notice and comment are not required for matters related to agency personnel or management). explicitly state that the subdele­gation has binding effect, or publish it in the CFR. Depending on which test is applied, many courts will be more likely to enforce the resulting subdelegation against the agency head under the Accardi doctrine.

But even when subdelegation takes the form of a legislative rule, courts must then engage in regulatory interpretation to determine whether the rule indeed divested the agency head of her authority—yet another source of doctrinal uncertainty. 274 See Merrill, The Accardi Principle, supra note 54, at 589–90 (discussing the “need to determine the meaning of an agency regulation” as a “recurr[ing]” issue in D.C. Circuit Accardi cases). Courts currently disagree on the appropriate method for interpreting regulations, with some focusing on text, others on intent, and still others invoking purpose. 275 See Kevin M. Stack, Interpreting Regulations, 111 Mich. L. Rev. 355, 359–60 (2012) [hereinafter Stack, Interpreting Regulations] (describing how courts “sometimes rely exclusively on the regulation’s text and canons of construction, but in other instances . . . invoke aspects of the regulation’s procedural history, the court’s construction of the authorizing statute’s purposes or congressional intent, or the agency’s own justification for the regulation, among other tools”). Judges will also sometimes defer to the agency head’s interpretation, while at other times pronouncing the rule’s meaning to be “plain” or evincing a pur­pose contrary to that claimed by the agency. 276 See Merrill, The Accardi Principle, supra note 54, at 589–90; see also Bowles v. Seminole Rock Sand Co., 325 U.S. 410, 414 (1945) (giving “controlling weight” to an agency interpreta­tion so long as it is not “plainly erroneous or inconsistent with the regulation”). Subdelegations, of course, are more likely to be credible when they are textually clear. Clarity, however, is usually in the eye of the beholder: Judges currently have no benchmark to determine just how much clarity is required. 277 Cf. Stack, Interpreting Regulations, supra note 275, at 357–60 (describing the inability of courts to develop a “consistent approach” to interpreting regulations); Brett M. Kavanaugh, Fixing Statutory Interpretation, 129 Harv. L. Rev. 2118, 2137 (2016) (reviewing Robert A. Katzmann, Judging Statutes (2014)) (“If the statute is 60-40 in one direction, is that enough to call it clear? How about 80-20? Who knows?”). Regulatory interpreta­tion can thus also be a source of ex ante uncertainty for the delegate.

Finally, even when an agency head enshrines a subdelegation in a clear legislative rule, courts will only enforce it if the agency head’s failure to abide by the rule affects the rights of individuals in a manner that results in substantial prejudice. 278 See United States v. Caceres, 440 U.S. 741, 754–55 (1979) (rejecting the use of the exclusionary rule in a case of surveillance that violated agency rules because “none of respondent’s constitutional rights has been violated here, either by the actual recording or by the agency violation of its own regulations”); Am. Farm Lines v. Black Ball Freight Serv., 397 U.S. 532, 538 (1970) (holding that the Interstate Commerce Commission’s failure to strictly comply with its own regulation by granting operating authority to a freight carrier that failed to reveal certain information did not prejudice the ability of other carriers to “make precise and informed objections to [the competitor’s] application”); see also Magill, supra note 51, at 879–80. In the Ninth Circuit’s Barraza-Leon case, for example, the petitioner claimed that the immigration judge had violated an Immigration and Naturalization Service regulation requiring that deportable aliens be notified of discretionary-relief eligibility. 279 United States v. Barraza-Leon, 575 F.2d 218, 221–22 (9th Cir. 1978). The court held that even if there was an error, the error was harmless and did not result in prejudice given that the petitioner would likely not have qualified for discretionary relief in the first place. 280 Even then, realistically speaking, most recipients of subdelegated authority are not lawyers. Id. They are often scientists, economists, or policy analysts with other subject-matter expertise. Of course, lawyers within agencies could inform and educate career staff about these legalities, but whether they do so or not is an open question. The author has been unable to find any guidance documents or internal agency memoranda on the issue. The EPA Delegations Manual also does not contain such information. In situations like these, when agency heads revoke or violate a subdelegation without harm to a third party, courts will not enforce the original subdelegation.

The upshot is that these doctrinal uncertainties and exceptions make it more difficult for agency heads to commit credibly to a subdele­gation. As a result, subordinates cannot rely on courts to predictably enforce their delegated authority. Instead, they must make edu­cated guesses about whether a court will characterize the delegation as a legislative rule, interpret the rule in a way that reflects the agency head’s intent to divest authority, and find that it results in some harm to a third party. 281 See supra notes 221–243 and accompanying text. As such, courts should adopt more bright-line rules in this con­text to make clearer the circumstances under which they will enforce the delegation against the agency head. What those rules are may not be as important as picking them. For example, Accardi could apply only to those subdelegations published as rules that undergo notice and com­ment and clearly express in the regulatory text an intent to delegate final authority. These modifications would make the Accardi analysis much simpler and more straightforward, thus allowing agency heads more of a safe harbor when seeking to credibly delegate authority. These innova­tions would augment the agency head’s ability to spur more internal investment in expertise.

B. Political Actors

This section now turns from the courts to the role political actors like the President and Congress could play to mitigate the democratic accountability concerns that internal agency subdelegation presents. In particular, it focuses on the normatively troubling prospect that agency heads could entrench their delegations and thereby prevent future parties and Presidents from exerting intra-agency control. Given the heated con­texts in which the practice often occurs, the prospect is especially worrisome in an age of increased political polarization. 282 See Cynthia R. Farina, Congressional Polarization: Terminal Constitutional Dysfunction?, 115 Colum. L. Rev. 1689, 1691 (2015) (addressing agency legitimacy prob­lems in an era of political polarization); Gillian E. Metzger, Agencies, Polarization, and the States, 115 Colum. L. Rev. 1739, 1762–65 (2015) (discussing the impact of political polar­ization on administrative agencies). Take, for instance, recent complaints by Republican minority commissioners at the FCC regarding internal delegations to the Chief of the Wireless Telecommunications Bureau. 283 See Press Release, FCC, Joint Statement of Commissioners Ajit Pai and Michael O’Rielly on the Abandonment of Consensus-Based Decision-Making at the FCC (Dec. 18, 2014), (on file with the Columbia Law Review) (arguing that the FCC Chair both improperly delegated a report on the state of competition in the wireless industry to the Wireless Telecommunications Bureau without their assent and that the Bureau improperly “grant[ed] a T-Mobile petition that asked the Commission—not the Bureau—to regulate cellular data roaming rates”). In their view, these subdelegations “cut the Commissioners out of the decision-making process entirely.” 284 Id.

More sharply, one of the commissioners, Michael O’Rielly, publicly observed that:

[I]t is extremely problematic for the Commission to have a 3-to-2 vote that includes broad delegation to the staff to address a subject area further. As a minority commissioner, it is bad enough to have your ideas and concepts rejected as a whole and have little to no input on an item (unless you completely ignore your principles). It is worse to see the extension of those decisions expanded upon for years to come by a bureau under delegated authority. 285 Michael O’Rielly, Delegated Authority: Serious Objections and Solutions, FCC: FCC Blog (Feb. 2, 2015), [].

Put differently, Commissioner O’Rielly objected to what he perceived as a conflict between the majority and minority parties on the Commission and the resulting erosion in norms of collegiality and consensus. To him, the majority Democratic party, led by the Obama-appointed chair, instead sought to entrench its power through the unelected agency staff.

It may be tempting to ask courts to police this prospect through the Accardi doctrine, perhaps refusing to enforce subdelegations when liti­gants can present evidence of partisan entrenchment. But a well-developed literature in the election law context suggests that political actors may better police against entrenchment concerns than judges can. 286 See Heather K. Gerken, Lost in the Political Thicket: The Court, Election Law, and the Doctrinal Interregnum, 153 U. Pa. L. Rev. 503, 517–18 (2004) (“Lacking a sound framework for adjudicating political process claims, the Supreme Court’s election law opinions often lack analytic coherence and thus provide little guidance to lower courts or other political actors.”); Samuel Issacharoff, Gerrymandering and Political Cartels, 116 Harv. L. Rev. 593, 630–45 (2002) (discussing the Supreme Court’s “constitutional entanglement” with the redistricting process and “intrusi[ons] into state political arrangements”); Samuel Issacharoff & Richard H. Pildes, Politics as Markets: Partisan Lockups of the Democratic Process, 50 Stan. L. Rev. 643, 644–45 (1998) (criticizing the Supreme Court’s failure to “articulate a . . . highly functional account of what features of democratic politics should be the focus of constitutional analysis”). For starters, entrenchment does not admit to an obvious judicially man­ageable standard. 287 See Nathaniel Persily, The Place of Competition in American Election Law, in The Marketplace of Democracy: Electoral Competition and American Politics 171, 172–74 (Michael P. McDonald & John Samples eds., 2006) (noting “[d]ifferent election laws will have different competition-related effects, and maximizing competitiveness along one dimension might diminish it on another”). The concept of raising the costs of repeal or amend­ment, for example, begs of normative baseline questions, particularly in the context of functional entrenchment concerns: How many allies cre­ated as a result of subdelegation is too many? How much greater must the marginal costs of repeal have to be to constitute too much? Such questions also require difficult inquiries regarding mixed motives: Did the agency head intend to entrench power or merely motivate internal expertise? 288 See Michael J. Klarman, Majoritarian Judicial Review: The Entrenchment Problem, 85 Geo. L.J. 491, 542 (1997) (“Even if one agrees that entrenchment problems have significant antimajoritarian implications and that they are not sufficiently self-correcting, one still might reject an anti-entrenchment theory of judicial review if the task it prescribes for courts is unmanageable.”). Perhaps for these reasons, along with more familiar separation-of-power concerns, courts have historically been deferential to agency heads’ judgments of how to manage their internal resources and affairs, especially when such decisions are not fixed by statute. 289 See Gillian E. Metzger, The Constitutional Duty to Supervise, 124 Yale L.J. 1836, 1872 (2015) (noting Justice Scalia’s assertion that individuals “‘cannot seek wholesale improvement of this program by court decree, rather than in the offices of the Department or the halls of Congress, where programmatic improvements are normally made’” (quoting Lujan v. Nat’l Wildlife Fed’n, 497 U.S. 871, 891 (1990))).

Accordingly, political actors have the best set of tools and the strong­est incentives to prevent the entrenchment of delegations to internal agency actors with preferences that are no longer aligned. For starters, Congress and the President should thus seek to render subdelega­tions more transparent, thus allowing interest groups and other monitors to aid with fire-alarm oversight. Congress could, for example, amend the Freedom of Information Act to require the publication of internal subdelegations. 290 5 U.S.C. § 552 (2012). The President could also issue an executive order re­quiring the same. In addition, or alternatively, the Office of Management and Budget or the Office of Information and Regulatory Affairs could formulate disclosure guidance for agencies pursuant to recent open gov­ernment initiatives. 291 See Memorandum on Transparency and Open Government to Heads of Exec. Dep’ts & Agencies, 74 Fed. Reg. 4685, 4685 (Jan. 21, 2009) (ordering executive departments and agencies to “put information about their operations and decisions online and readily available to the public”); Memorandum from Peter R. Orszag, Dir., Office of Mgmt. & Budget, on Open Government Directive to Heads of Exec. Dep’ts & Agencies 2–6 (Dec. 8, 2009),
open-government-directive [] (requiring executive depart­ments and agencies to take four steps toward “the goal of creating a more open government”). For examples of such guidance documents, see Memorandum from Cass R. Sunstein, Adm’r, Office of Info. & Regulatory Affairs, on Clarifying Regulatory Requirements: Executive Summaries to Heads of Exec. Dep’ts & Agencies 1 (Jan. 4, 2012), http:// [http://] (requiring exec­utive summaries “for lengthy or complex rules”); Memorandum from Clay Johnson III, Deputy Dir. for Mgmt., Office of Mgmt. & Budget, on Policies for Federal Agency Public Websites to Heads of Exec. Dep’ts & Agencies (Dec. 17, 2004), [] (detailing policies for use of federal agency public websites).

Because publication is often not sufficient to garner presidential or legislative attention, Congress might also consider requiring agencies to submit internal delegations after a presidential transition to the relevant congressional committees or the Government Accountability Office. Doing so would help to enable legislative monitors to alert aligned agency officials or else to police subdelegations otherwise con­trary to cur­rent congressional preferences through hearings, appropria­tions decisions, or statutory amendments. Congress or the President could go even further by passing legislation or drafting an executive order that would automatically sunset all internal agency subdelegations—thereby requiring incoming agency heads to affirmatively review and ratify them. Such measures would help to ensure that subdelegations to lower-level officials do not persist due to simple path dependence or more conscious efforts to engage in partisan entrenchment.


This Essay has examined the positive dynamics and normative implications of delegation, so to speak, one-level down: the grants of authority that agency heads routinely assign within administrative agen­cies. It has argued that such internal subdelegations are best understood as credible commitment devices through which commission­ers motivate better-informed but potentially biased subordinates. The main positive claims are that an agency head will enhance the credibility of her delegation in order to induce more information production, but only within a bounded range of expected preference divergence. Moreover, when intra-agency review costs increase, agency heads will be more likely to delegate final authority in order to eliminate those transaction costs. This study’s main effort has been to better understand subdelegation as a bureaucratic practice and the ways in which credibility can be institu­tionalized within administrative agencies.

Future research questions, of course, persist: For example, are there stable features of an agency’s mission or jurisdiction that help to explain agency variation beyond relative degrees of preference divergence and transaction costs? When and why do agency heads choose internal versus external delegations? How do the dynamics of multimember commis­sions differ from those of executive agencies? Only when agencies make subdelegations more transparent can such questions be systematically addressed. These and other inquiries suggest that a fruitful research agenda remains in the continuing project to develop administrative law, not only from above but also from within the agency.