Learning While Governing: Expertise and Accountability in the Executive Branch

By Sean Gailmard and John W. Patty

Although their leaders and staff are not elected, bureaucratic agencies have the power to make policy decisions that carry the full force of the law. In this groundbreaking book, Sean Gailmard and John W. Patty explore an issue central to political science and public administration: How do Congress and the president ensure that bureaucratic agencies implement their preferred policies?
 
The assumption has long been that bureaucrats bring to their positions expertise, which must then be marshaled to serve the interests of a particular policy. In Learning While Governing, Gailmard and Patty overturn this conventional wisdom, showing instead that much of what bureaucrats need to know to perform effectively is learned on the job. Bureaucratic expertise, they argue, is a function of administrative institutions and interactions with political authorities that collectively create an incentive for bureaucrats to develop expertise. The challenge for elected officials is therefore to provide agencies with the autonomy to do so while making sure they do not stray significantly from the administration’s course. To support this claim, the authors analyze several types of information-management processes. Learning While Governing speaks to an issue with direct bearing on power relations between Congress, the president, and the executive agencies, and it will be a welcome addition to the literature on bureaucratic development.

• Language: English
• Publisher: University of Chicago Press
• Release date: Dec 6, 2012
• ISBN: 9780226924427

Book preview

them.

1   Introduction

Information is the lifeblood of executive branch action. Or at least, it is for effective action. At least three of the four virtues of executive action that Alexander Hamilton cited in Federalist 70—unity, activity, secrecy, and dispatch—presuppose information to act.¹ Information is also the bedrock of bureaucratic legitimacy in the United States. Bureaucratization is indisputably one of the most important developments in American government in the last hundred years, and information is its principal justification. Bureaucrats are elected by no one. They lack the measure of democratic pedigree that elected officials can claim. Most of the vast cadre of professional bureaucrats in Washington and around the country is not even appointed by anyone who is elected to office. Yet bureaucrats have the power to make policy decisions with the full force of law (even trumping, as far as federal bureaucrats are concerned, the decisions of duly elected state legislatures, thanks to judicial decree). What they do have, or are supposed to have, is information.

Correspondingly, the analysis of information has become central to the analysis of executive branch politics. The key premise of the informational literature on bureaucratic politics in particular is that bureaucrats have access not only to reasonably high quality information about policies and programs under their purview but often to better information than other concerned political actors. This in turn gives rise to an information asymmetry between bureaucrats and political overseers that engenders problems for the holding bureaucrats accountable to the aims and vision of elected officials. Similarly, the informational advantages enjoyed by presidents can make it difficult for Congress and other observers of presidential policymaking to determine whether their interests are served by a particular course of action.

There are several variants of this accountability problem. First, actions taken by executive branch officials may be difficult for elected authorities to observe at all, and therefore to scrutinize. This may be either because of claims of secrecy or executive privilege (particularly as pertains to the president), or because the vast scope of the federal bureaucracy relative to Congress raises high barriers simply to figuring out what is going on in all corners of the executive branch. Compounding the difficulty, even when executive actions are perfectly observed (and Congress has evolved practices to help in this respect; see McCubbins and Schwartz 1984; McCubbins et al. 1987), they are often pursued by executive branch policy makers with deeper knowledge of the subtleties of a problem or with information to which Congress and outside observers are not privy. When a policy choice is made in the executive branch—such as escalating a war (e.g., the Vietnam War or the War on Terror), pursuing a particular program of economic recovery (e.g., the Troubled Asset Relief Program and auto industry bailouts), or deferring to industry in response to an environmental catastrophe (e.g., British Petroleum’s interminable response to the Deepwater Horizon explosion in the Gulf of Mexico)—does it represent a generally beneficial response to delicate circumstances, or does it represent the elevation of executives’ own personal ideological (or worse, economic) projects over the public (or at least the legislature’s) interest? Asymmetric information between executive branch actors and Congress (and beyond), combined with conflicting ideological goals or visions of good public policy, makes the assessment of executive branch action problematic and colors our political discourse.

Yet for all the accountability problems it presents, the value of high-quality information in bureaucracies cannot be seriously doubted. We still largely accept the formulation that James M. Landis presented in The Administrative Process in 1938: the informational demands of modern policymaking are excessively taxing on the institutions of government that were established in 1787, and the need for effective public action trumps the joy of constitutional kabuki theater. Therefore, we must reconcile ourselves to institutional forms that leverage the information necessary for policymaking through the power to act on it.

With some exceptions we note later, analysis of the politics of executive branch policymaking has largely taken its informational advantage as given. This is true of the literature on bureaucratic statutory discretion, especially as articulated by Epstein and O’Halloran (1999), Gailmard (2002), Volden (2002), and Huber and McCarty (2004), but also to some extent by Huber and Shipan (2002).² This assumed exogeneity of information is also true of a large strand of literature on the presidency following Neustadt (1960), which focuses on the president’s informational advantage as a linchpin of the president’s power to persuade (e.g., Canes-Wrone et al. 2007). When scholars do focus on the development of informational supports for the presidency (or, only somewhat less directly, on the sources of the president’s bargaining powers, e.g., Dickinson 1996), they typically explore these developments as presidentially initiated and determined.

Taking information and information asymmetries as givens can make some sense when analyzing the behavior of mature bureaucracies; they are background conditions that in fact are often satisfied, and the implications of which it is obviously important to explore. Yet policy-relevant information held by executive branch actors does not emerge from the ether. Information must be acquired, expertise must be obtained, and capacity must be developed. In short, information held by executive branch policy makers is endogenous. In particular, it is endogenous to the incentives that executive branch actors have to acquire, share, and use information, and these incentives are in part determined by the organizational structure and political position of executive branch institutions. Taking executive branch information as given elides the paramount developmental problem in executive branch organizations: how can we ensure that they cultivate that information?

In this book, we analyze the microfoundations of information acquisition in the executive branch. We consider the effects of organizational structure on the development of the expertise that has been taken as given in models to date that explicitly consider information asymmetries. The implicit assumption in our analysis is that acquiring expertise and information is more complicated than declaring in a legislative edict that it shall exist. When an executive branch bureaucracy or regulatory agency opens its doors, it cannot ensure that it will be staffed by experts in the policy areas under its domain by putting out a want ad announcing that only experts need apply. Selecting individuals who pass a merit examination does not per se ensure that those individuals will bring with them or develop any particular understanding of the policy or administrative problems their agency will face.³ Requiring individuals to undergo university training in a general subject does not per se ensure that they will vigorously apply themselves in the development of their expertise, or that any expertise they might develop will translate into policy-relevant or administratively relevant expertise.

Instead, we analyze the development of information and expertise in executive branch bureaucracies as the product of incentives that bureaucratic agents face to acquire and share it. We assume that information development and transmission must be consistent with the incentives of individuals and organizations that do the developing and transmitting.⁴ We argue in the following chapters that the organizational structure and political position of executive branch organizations can decisively affect their and their members’ incentives to make costly investments in information and expertise, and therefore can condition whether they live up to their promise in the policy process.

Our theoretical arguments suggest that the logic of accountability and political responsiveness as applied to executive branch bureaucrats is more complicated than simply asking whether the decision making of these unelected officials demonstrates fealty to the citizens’ interests, however those interests might be measured. In particular, accounting for the realities of policymaking under uncertainty immediately raises the question of the degree to which executive branch officials can achieve the electorate’s policy desires. It is useful to consider this dilemma in some detail, as it distinguishes our theoretical treatment, as well as the set of problems it is intended to help us understand, from the traditional focus of the literatures to which we are speaking.

1.1 Accountability and Expertise

Taking bureaucratic expertise as given gives rise to certain tropes of accountability: a pithy synopsis is furnished by the formulation that expert bureaucrats can be considered fully accountable to political principals when they make the same decisions the principals would have made, if they held the information the bureaucrats hold. We call this the standard logic of bureaucratic accountability. This logic emerges straightforwardly from a basic variant of a model in which principals—whether they be voters, members of Congress, or the president—seek to obtain the best possible results from bureaucratic agents who are tasked with choosing some action(s) on the principal’s behalf.

Depending on the problem at hand, the principal(s) may be thought of as having various tools to influence the agent’s choices, and the assumption in principal–agent modeling is that any principal anticipates the effect of its interactions with the agent on the agent’s decisions, and will use whatever tools are at its disposal to induce the most favorable possible action (from the principal’s own point of view) from the agent. For instance, principals may be able to select the policy preferences and ideological disposition of their agents. In that case the standard logic would imply that the principal selects its ideological clone to be its agent; this implication of the standard logic is dubbed the ally principle by Bendor and Meirowitz (2004). Such an agent is perfectly trustworthy to use its information in the principal’s interest, without labor-intensive monitoring of the agent or the need to double-check its information. If principals cannot entirely determine the agent’s policy preferences, they may be able to use oversight strategies to monitor the agent’s actions and ensure fidelity to the principal’s goals, revise or nullify policy choices the agent makes that do not serve the principal’s ex post interests (as in, e.g., legislative vetoes or Joint Resolutions of Disapproval under the Congressional Review Act of 1996), make threats of costly (for the agent) budget reductions or withering contumely in congressional hearings to elicit information, tailor the discretion agents can use to set policy that they can reveal information without concern over adverse policy consequences, and so forth.

Our problem with the standard logic and its oft-noted implications is not that they are inaccurate as far as they go. The standard logic is, in fact, a fine summary of the accountability problem that political principals face with respect to bureaucratic agents who possess the relevant information and expertise to make informed policy. Furthermore, supposing that it is possible to hire and appropriately control expert agents for the purpose of making policy decisions, traditional notions of political accountability clearly suggest that delegation of authority to such agents, even if unelected, can nonetheless be representative of the voters’ interests. Rather, our concern is with the precondition assumed by this trope: where does bureaucratic information, competence, and capacity come from? More generally, how can bureaucratic agents be induced to learn and to share the information that justifies their existence in a titular democracy? Our approach highlights a logic of bureaucratic accountability under endogenous information that is qualitatively different than versions that suffice when information is exogenously given. The interests of political principals include not just using available information as the principals would use it if they had it, but also acquiring policy-relevant information in the first place. Faithfully reflecting those interests, therefore, includes the development of bureaucratic expertise.

A central tension in our analysis is that inducing an agent to faithfully apply information he or she already possesses may directly conflict with inducing that agent to acquire the information in the first place. If one excises the issue of acquiring information, it is often the case that accountability is maximized by choosing an ideological clone of one’s self as one’s agent (e.g., Bendor and Meirowitz 2004). Given exogenously produced policy expertise, it is frequently the case that the principal can effectively induce the agent to account for all of the principal’s costs and benefits from the various possible policy choices with the benefit of superior information to evaluate their effects on the relevant policy outcomes. This logic has deeply affected the analysis of bureaucratic structure. Our claim in this book is that supposition of exogenously extant expertise is key to the conclusions deduced from this logic. Specifically, inducing an agent to acquire, share, or elicit information (i.e., develop expertise) can require the principal to think about the delegation problem in ways that are very different from how he or she need think about providing incentives for a known expert to utilize that expertise when making policy decisions on the principal’s behalf.

First consider the issue of acquiring expertise, by which we mean making costly investments in information. An agent’s willingness to do so implies that the agent will somehow gain from doing so. For career agents in conventional civil service systems, direct pecuniary incentives (e.g., merit pay) for doing so are infrequently observed. A simplistic reading of this regularity is as an implication that bureaucratic expertise development is unimportant to the principals who control the incentives offered to bureaucrats. However, our arguments in Chapter 2 illustrate that pecuniary incentives are not the only or even necessarily a particularly powerful channel for inducing costly investments in information. Rather, if agents might have intrinsic preferences about public policy outcomes, and acquiring information allows them to improve it, then affording agents some discretion to shape policy choices can be more effective than financial remuneration as an inducement for the acquisition of policy-relevant information and expertise. If agents care about public policy, the discretion to affect it is essentially an incentive to acquire expertise—but providing such discretion is frequently directly at odds with the standard trope of bureaucratic accountability discussed earlier, wherein the personal policy preferences of a perfectly accountable bureaucratic agent shall play no role in the ultimate policy choice. Put a slightly different way, optimally inducing the development of bureaucratic expertise often requires explicit provision of the opportunity for bureaucratic slippage: inducing neutral competence is generally impossible unless the competence can be taken as given.

Acknowledging the need to induce the development of bureaucratic expertise raises accountability issues along dimensions other than the provision of meaningful policy discretion to the agent. For example, whereas the traditional trope of bureaucratic accountability suggests that a principal can never gain from binding his or her hands with respect to the agent’s retention, the reward of future employment should in general be tied to the agent’s performance, accounting for endogenous expertise can reverse this conclusion. In particular, ensuring relatively stable tenure of employment (i.e., binding the principal’s hands with respect to retention), as archetype civil service systems do, can magnify the agent’s incentive to acquire expertise precisely because this provides a longer time horizon over which the agent’s investments in expertise can be leveraged by the agent as he or she makes policy decisions in the future. True, political principals would have nothing to worry about if, consistent with the ally principle, they selected their ideological clones and gave them plenary discretion to choose public policy. But real principals do not have such fine-grained and instantaneous control to select (and change) executive agents to suit their ideological goals—and given the value of relatively long tenure for ensuring the value of expertise investments they may not want it. Given that constraint and the need for bureaucratic expertise, it can be useful for political principals to provide bureaucratic agents with both discretion to determine the course of public policy and relatively stable tenure—even though those agents do not share the principals’ goals. Although such an approach sacrifices control over policy, its upside is better information in the bureaucracy.

Sharing and eliciting information both imply some sort of communication between a holder and a seeker of information. Strategically, the completeness and value of this communication depends on the congruence of preferences on the two sides of the communication about what to do with the information revealed or communicated. Given this, the key question for political principals is, with whom is the agent communicating? The structure of the policy process often implies that executive branch agents are not communicating only or primarily with the political principals to whom the agent is ultimately answerable. Instead, the agent may be communicating with other actors in the executive branch or with private interests that hold policy-relevant information. In either case, agents are only useful if they can effectively exchange information with those they communicate with. This requires a concordance of ideologies and preferences not between the agent and its principals, as the ally principle implies, but between the agent and its confidantes. If an agent is funneling information to executive branch policy makers, it must provide information that is credible and believable to those policy makers—which in turn requires that the agent’s ideology is closely linked with those it advises. If an agent elicits information from, say, private interests, it must have enough in common with them ideologically to be a credible repository of information. Additionally, in this case it must have enough discretion about how to use the information in making policy that the informed private interests do not simply assume that any proposal the agent makes will be overridden by a political principal once the relevant information is revealed.

In sum, in all these cases where the availability of information is key—where it is acquired, shared, or elicited—the standard tropes of accountability are turned on their head. Rather than an agent that is a perfect ideological replicant of its principal, an effective agent will be ideologically differentiated from its principal. And rather than an agent whose policy decisions are easily overseen and revised by its principal, an effective agent will have a zone of relatively durable discretion to determine policy choices. The effectiveness of an agent at bringing greater information to bear on public policy is endogenous to the organizational and political features that map into its ideological predispositions and its responsiveness to its principals.

1.2 Other Views on Bureaucratic Expertise

It is important, before embarking on the development of our arguments, to briefly consider more specifically where we are attempting to contribute to the related literatures. In particular, it is useful to note the giants’ shoulders upon which we are attempting to climb as well as the theoretical cupboard space that we are attempting to make a little less bare. We believe that the arguments presented here represent a complement to existing theoretical treatments of delegation and bureaucratic politics more generally. Though the study of information has deeply affected the political science literature on bureaucratic behavior, organization, and accountability, very little of that literature has considered the issue of why or how executive actors develop their informational capacity, or the effect of organizational structure on that development.⁵ There are, however, some prominent exceptions to this general tendency.

THEORETICAL LITERATURE. It is highly doubtful that we could have developed our arguments without the work of Tom Gilligan and Keith Krehbiel.⁶ Although these models were substantively focused on the institutional structure of legislatures, their theoretical contributions are clearly and closely related to our analysis. Gilligan and Krehbiel’s central theme is the effect of organizational structures in Congress on the incentives of legislators and committees to acquire policy-relevant information and bring it to bear when enacting legislation. The allocation of decision authority is the primary reason why our analysis leads to different results than the model presented in Gilligan and Krehbiel (1987), which is consistent with the ally principle. In Gilligan and Krehbiel (1987), the principal designing the committees is also the ultimate policy maker; it clearly cannot get more information from any expert agent besides its own ideological clone. We consider situations in which the agent’s designing principal (1) may set it up to communicate with other policy makers or (2) may set it up as a decision maker to receive information from an informed party. In both of those situations, as we will argue, the ally principle clearly can fail.

Another early contribution to the theoretical literature upon which our work builds is that of Bawn (1995). In arguments similar to some we present, Bawn focuses attention on the trade-off between accountability and expertise in administrative agencies. Our work hopefully complements Bawn’s insofar as our theoretical frameworks provide some the microfoundations of this trade-off. Aghion and Tirole (1997), leading a large subsequent contract theory literature, develop a principal–agent model in which delegation of authority to an agent induces the agent to acquire information.⁷ Bendor and Meirowitz (2004), upon naming the ally principle, developed several protomodels in which it would fail due to endogenous information. Ting (2011) models a bureaucratic agency’s investment in capacity to implement policy; the central point is that capacity can give the agent some agenda-setting power with respect to a legislative principal, and thus investment in it may actually be higher when the legislature and agency are ideologically divergent. Recent work by Matthew Stephenson analyzes the effect on incentives for information acquisition stemming from a variety of well-known legislative and judicial devices for controlling the decisions of administrative agents (e.g., postures of judicial scrutiny of or deference to agency actions).⁸ An important difference between our work and that of Stephenson is the type of institutional details that are analyzed. Stephenson’s work provides an understanding of how administrative-legal institutions that map into, for instance, policy enactment costs facing agents promote both agency activity and the quality of policymaking, but does not (and is not intended to) provide insight into why the agent does not share the principal’s preferences or why the agent has been given discretionary control of the policy lever in question.

EMPIRICAL LITERATURE. Feldman (1989) presents a penetrating study of information production by bureaucratic analysts in the U.S. Department of Energy. Though Feldman finds that analysts’ information is often not used in policymaking, analysts produce it because it is the only way that they can hope to have an influence on policy making, and because they value the work for intrinsic, professional, and social reasons.⁹ Feldman’s ethnographic analysis taps into similar sorts of microfoundations of organizational incentives for information production that we analyze theoretically and historically in this book.¹⁰ Second, Carpenter (2001) argues that bureaucratic information and the capacity to act on it form one pillar of bureaucratic autonomy, that is, the ability of bureaucrats to pursue an agenda that is not chosen by titular principals and may be at odds with their goals. The drive for autonomy, Carpenter argues, therefore provides one important incentive for the development of capacity in bureaucracies. Third, Lewis (2008) contends that the thickening of layers of presidential appointees in the bureaucracy dampens the application of programmatic expertise by career bureaucrats and thereby undermines the effectiveness of executive branch agencies.¹¹ Lewis’s analysis therefore draws a link between the political-organizational structure of bureaucracies and the availability of information possessed by bureaucrats.

Despite the quality and insight of this work, endogenous information—particularly any attention to microfoundations of information provision—is still relatively peripheral in the analysis of bureaucratic organization and accountability. In addition, the theoretical issues that endogenous information raises have not been squared with the development of executive branch institutions in the United States, nor have the forms of institutions that we have inherited been systematically analyzed from this perspective. This book is our attempt to push the bureaucracy and executive branch literatures in that direction.

1.3 A Typology of Endogenous Information Problems

The core of the book consists of three sections. Each section examines a particular type of endogenous information problem. The first chapter of each section is devoted to a theoretical exposition of the problem and the second to an analysis of historical cases or episodes in light of the models. The sections and endogenous information problems pertain to situations in which (1) a bureaucratic agent has to acquire information from scratch by investing time in learning it, (2) a bureaucratic agent who is informed of some policy-relevant information has to decide to what extent it should be shared with other executive-branch policy makers, and (3) an agent has to elicit information from private, regulated interests outside the government that may be particularly privy to it. Our working shorthand for this typology is that the endogenous information problem may be variously (1) acquiring information, (2) sharing information, or (3) eliciting information.

We can think of this typology as resulting from a two-dimensional classification of endogenous information problems: (1) whether or not some actor is exogenously informed of the relevant information and (2) whether the actor who depends on that information to make a decision is formally part of the executive branch policy-making process. When the answers to these two questions are no and yes, respectively, we wind up with problems of type 1; when the answers are yes and yes, respectively, we wind up with problems of type 2; when the answers are yes and no, respectively, we wind up with problems of type 3. It is, of course, entirely possible that the answers are no and no, but such situations are apparently outside the scope of an analysis of public policy-making institutions.

For each class of endogenous information problem we consider, we analyze historical episodes of organizational development in the executive branch from the perspective of the attendant theoretical model. The purpose of the models is not so much to make predictions in a narrow sense about what event should have occurred as it is to provide an interpretive lens through which to understand the cases.

For the first class of endogenous information problem—inducing an agent to acquire information—we analyze the development and consolidation of civil service in the federal government. The central argument is that policy discretion for bureaucratic agents combined with stable careers subject to merit protection provides strong incentives for individuals particularly motivated by substantive public policy issues to (1) select themselves into public sector employment and (2) dedicate themselves to developing expertise in the policy issues and programs within their purview. Merit system protection obviously dulls the incentives of bureaucrats in the sense that they face limited discipline in the event of poor performance, and these points are sufficiently well understood and sufficiently obvious that we do not need to spend a great deal of time making them. Our point in this section is that the very dullness of discipline (or termination) threats in civil service provides a long time horizon over which personal dedication to policy and program details can bear fruit, and this in turn provides motivation for policy-motivated agents to make those investments. When agents are strongly policy motivated in the first place—and do not need strong extrinsic incentives to obtain utility from exerting themselves—civil service’s incentives supporting dedication and personal investment create value for a political principal that can outweigh the costs of weak extrinsic incentives for good performance.¹²

This theoretical argument is developed in the first chapter of Part I. We then review the evolution of civil service in the federal government, with particular attention to the period from the passage of the Pendleton Act in 1883 to the New Deal. Over this time period, civil service employment was transformed from a regime of clerkship on the job and rotation of offices based on party affiliation to a regime of stable employment irrespective of partisan changes, with greater and greater responsibility for and impact on policy at the top levels of the civil service hierarchy. We argue that the developments in the civil service that took place over this half century operated in a reinforcing pattern.

Part II of the book addresses endogenous information problems in which an agent of some political principals holds some policy-relevant information and must decide whether to share it with other policy makers. If these latter policy makers who receive the agent’s communication are the same as the agent’s principals, the principal obviously wants a complete clone of itself. This problem is more interesting when considered in light of expansive and growing unilateral power for presidents and their agents in the executive branch. When these actors are the recipients of an informed agent’s communication, that agent’s principals are better off when the agent’s ideology and worldview are linked not to the principals but to the recipients of the agent’s communication. Only when the informed sender of this information and its recipient share common policy preferences and ideological commitments can it be guaranteed that the sender has an incentive to fully and completely inform the receiver without dissembling.

We apply this theoretical framework in Part II to understand the growth of the institutional presidency, particularly the incentives for Congress to proactively facilitate this growth. The president’s informational advantage with respect to Congress is sometimes taken as exogenous and an underlying cause of the growth of presidential policy influence relative to Congress. But it is not exogenous to the presidential-congressional relationship—it is determined by that relationship. The informational support that underpins the president’s de facto role as policy maker in chief is provided by the institutional presidency, and the institutional presidency is created and supported by Congress. This seems at one level paradoxical, from a principal–agent standpoint at least: Congress has actively participated in creating an agent that is more responsive to the president than to Congress itself. Yet it is easily resolved by considering the purpose that this particular sort of agent serves for Congress: to make sure that the president’s discretionary powers over public policy are used pursuant to relevant information. As Edward Corwin put it, the president has the unique ability, in the daily course of events on the job, to endanger the lives of every American. Congress may not favor the president’s policy agenda at any given time; but that notwithstanding, Congress has every incentive to ensure that the president does not knowingly and in a foreseeable way use his authority to undermine the safety or interests of the United States.

In other words, the implication of the theoretical analysis in Part II is that, taking the president’s ability to act as given, Congress has strong incentives to provide the president with incentives to use that ability wisely. Now, Congress may wish the president had less authority to act, but it is not entirely and always up to Congress. As long as the president has a remotely credible constitutional claim to authority that he wishes to exercise, he will make that claim and exercise that authority. It does not matter if these claims—such as notions of a stewardship presidency, strong forms of unitary executive arguments, invocation of the president’s oath of office as a grant of substantive power, claims of judicial sanction to act as the sole organ of the United States in foreign affairs, and the like—are incontrovertible. Constitutional scholars and political scientists may contest the legitimacy of these claims, but in the game of politics what matters is that such claims provide cover to take a desired