Show Summary Details

Page of

PRINTED FROM OXFORD HANDBOOKS ONLINE ( © Oxford University Press, 2018. All Rights Reserved. Under the terms of the licence agreement, an individual user may print out a PDF of a single chapter of a title in Oxford Handbooks Online for personal use (for details see Privacy Policy and Legal Notice).

date: 17 February 2020

Politics, Delegation, and Bureaucracy

Abstract and Keywords

This article studies some of the recent models and associated empirical research on politics, delegation, and bureaucracy. It describes the four core theoretical arguments that constantly emerge from these models and the empirical tests that focus on patterns of delegation strategies. The article also reviews some of the recent theoretical models that are able to provide insights into the circumstances under which the logic of these theoretical arguments breaks down.

Keywords: recent models, empirical research, politics, delegation, bureaucracy, theoretical arguments, delegation strategies

Modern democratic government cannot function without bureaucracy. Given the vast array of policy issues that come before government, the complexity of these issues, and the resources needed to address them, elected politicians have no choice but to delegate at least some responsibility over these issues to bureaucracies. Of course, once politicians delegate, they also face a potential loss of control over the issues that they have delegated. This tension between the necessity of delegation and the potential problems associated with delegation underlines the fundamentally political nature of bureaucracy. Government bureaucracies are surrounded, and affected, by other political actors, including legislators and courts, presidents and governors, prime ministers and cabinet ministers, and interest groups, all of whom attempt to manage this tension. Any analysis of bureaucracies, therefore, needs to view them through the lens of political analysis.

Prior to the 1980s, a dominant theme in the literature on bureaucracy emphasized the “administrative state.” The complexity of policy issues and the rapid expansion of government involvement into new policy areas made it very difficult for politicians to make important policy decisions. By contrast, the increasing professionalism and specialization of bureaucracy equipped bureaucrats with the expertise and experience needed to make these decisions. Consequently, the argument went, bureaucrats run (p. 850) the show, while politicians essentially have no choice but to sit on the sidelines and watch.

In the early 1980s, scholars began strongly to challenge the administrative dominance perspective. In particular, researchers began to examine the design of rules for bureaucratic decision-making and political oversight of bureaucracy (e.g. McCubbins, Noll, and Weingast 1987, 1989; Moe 1989). With the correct institutions governing agency decision-making, some argued, politicians ensure that bureaucratic actors pursue the goals of politicians. Although researchers disagreed about whether this in fact was true, and, more generally, about which elements of structure and process were central to understanding bureaucratic behavior, these arguments have set the agenda for much research on bureaucracy in the last twenty years. A large number of formal models of delegation have explored the instruments that politicians use to delegate to bureaucrats. These models, which have spawned a rich empirical literature, typically focus on two types of delegation instruments: ex ante instruments, which allow politicians to establish the level of discretion bureaucrats have when making policy, and ex post instruments, which allow politicians or other political actors to monitor or audit bureaucrats after agents take action.

This chapter reviews some of these recent models and associated empirical research. We begin by describing the four core theoretical arguments that consistently emerge from these models about the circumstances under which politicians should be expected to grant either more or less discretion to bureaucrats. After reviewing these arguments, we describe empirical tests that focus on patterns of delegation strategies. We argue that considerable support is accumulating for these arguments, and that we can therefore use the models on which they are based to make inferences about the circumstances under which administrative dominance is most likely. We conclude by reviewing recent theoretical models that provide insights into the circumstances under which the logic of these theoretical arguments breaks down. By considering questions of theoretical robustness, we are able to describe areas that are ripe for additional theorizing and empirical research.

1 Core Arguments about Delegation Strategies

Politics, Delegation, and Bureaucracy

Fig. 41.1 Politicians, bureaucrats, and policy uncertainty

In this section we sketch the four most prominent arguments that have emerged from efforts to model delegation. Two standard assumptions underlie these arguments. First, politicians and bureaucrats often want to achieve different objectives. This is not always true, of course; but when it is not, delegation presents few problems, since politicians can trust bureaucrats to act in their interests. Second, there is asymmetric information about how to achieve policy goals, with bureaucrats typically having more expertise than politicians. The challenge for politicians is how to draw on this (p. 851) expertise given that bureaucrats want to put it to ends that differ from those desired by politicians.

Many models use a simple, unidimensional spatial framework to implement these two assumptions. A typical model will have a single Politician with ideal point at xP, and a single Bureaucrat with ideal point at xB, as illustrated in Figure 41.1 The level of policy conflict between them is simply the distance between xP and xB. As this distance grows, the Politician becomes increasingly worried that delegating more authority to the Bureaucrat will lead to a policy outcome that the Politician does not like. In the examples that follow, we assume xP = 0 < xB and that both players have quadratic preferences.

The Bureaucrat’s informational advantage is often introduced into delegation models by assuming uncertainty about the outcome that will occur if a particular policy is adopted. A common way to model this asymmetry is to assume that after some policy action, a, is taken, the outcome from this policy is o = a − ɛ. The random variable ɛ is typically drawn from some well-behaved distribution with finite support, for instance, a uniform distribution. If ɛ ∈ [−E, E], then E is a parameter that measures policy uncertainty. As E gets larger, so too does uncertainty about the outcome that will result from adopting a particular policy a. Both players might know the distribution from which ɛ is drawn, but the Bureaucrat has more information. Models typically make the information asymmetry as stark as possible by assuming that the Bureaucrat knows the exact value of ɛ, whereas the Politician knows only the distribution from which ɛ is drawn. If, for example, the Bureaucrat knows ɛ and is allowed to implement any policy choice he wants, he can choose a policy a that will produce an outcome equal to his ideal point, as shown in Figure 41.1.

Scholars use the framework described above to explore the strategies that politicians can use to encourage bureaucrats to adopt policies that serve the interests of politicians. The models tell us how different forms of delegation yield different outcomes. One central thrust of the literature is to use these results to understand the factors that influence how much authority politicians should delegate to bureaucrats. Politicians can use legal instruments, such as legislative statutes or regulations, to determine the range of policies that bureaucrats can adopt or the types of actions they can take in making policy decisions.

Politics, Delegation, and Bureaucracy

Fig. 41.2 The decision to delegate in spatial models

Epstein and O’Halloran (1994) developed a modeling technology that allows one to make explicit the amount of discretion that politicians delegate to bureaucrats. In models of this type, the Politician specifies an interval, such as [ x ¯ , x ¯ ] , from which the (p. 852) Bureaucrat must choose his policy action. The lowest discretion law occurs, for example, when x ¯ = x ¯ —in other words, when the Politician tells the Bureaucrat precisely what to do. As the interval [ x ¯ , x ¯ ] expands, the Bureaucrat has more discretion in the policy space from which to choose his action.

To illustrate the logic of the central results in the literature, we consider a simple example of the type of delegation model originally developed by Epstein and O’Halloran. In this example, the Politician can chose one of two delegation strategies. She can give the Bureaucrat no discretion, instead instructing him to implement some specific policy, z (e.g. x ¯ = x ¯ = z ). Or she adopt some law, y, that gives complete discretion to the Bureaucrat to adopt any policy he wishes (e.g. y is such that x ¯ = and x ¯ = . The example also puts the simplest possible structure on policy uncertainty: ɛ ∈ {0, 1} and each ɛ occurs with probability of one-half.

When will the Politician prefer to delegate substantial authority to the Bureaucrat, and when will she prefer to grant minimal discretion? Suppose the Politician delegates policy-making to a Bureaucrat by adopting y = [—∞, ∞]. As shown in Figure 41.2, the Bureaucrat will take action a = xB + ɛ, ensuring an outcome at xB and yielding the Politician an expected utility E U P ( y = [ , ] ) = x B 2 . The Politician will compare the utility she receives from a policy at the Bureaucrat’s ideal point with the expected utility she receives from her optimal low discretion statute, z. The expected utility of z is E U P ( z ) = 1 2 z 2 1 2 ( z 1 ) 2 , which is maximized when z = 1 2 . As illustrated in Figure 41.2, this yields an outcome at 1 2 if ɛ = 0, and an outcome at 1 2 if ɛ = 1. Consequently, the expected utility to the Politician of adopting her optimal z is 1 4 , and the Politician prefers z = 1 2 to y = [—∞, ∞] whenever x B > 1 2 .

The example illustrates the fundamental tension that bureaucrats’ policy expertise can create for politicians. A politician can allow substantial discretion to bureaucrats, but only at the risk that bureaucrats will use their policy expertise against the politician. But if a politician tries to counter this risk by writing a low-discretion statute, the policy expertise of the bureaucrat is wasted, potentially making both politicians and bureaucrats worse off than they could have been. In Figure 41.2, for example, if z = 1 2 , then whenever ɛ = 1, the outcome is o = 1 2 . Both the Politician and the Bureaucrat would have preferred z = 1, yielding an outcome at xP.

Given this tension, what factors lead Politicians to delegate more versus less discretion to bureaucrats? One central theme in the literature concerns the effect of policy uncertainty. Suppose that there was a higher level of uncertainty than in the previous example. It may be the case, for instance, that ɛ ∈ {0, 2}, with each ɛ equally probable. Since the Bureaucrat knows ɛ, the expected utility to the Politician (p. 853) of delegating to the Bureaucrat is unaffected by this increase in the level of policy uncertainty. The Bureaucrat will still adopt a = xB + ɛ, yielding an expected utility to the Politician of E U P ( y = [ , ] ) = x B 2 . But the Politician’s expected utility of setting policy herself is lowered by the increase in policy uncertainty. In this case, E U P ( z ) = 1 2 z 2 1 2 ( z 2 ) 2 is maximized when z = 1, which yields an expected utility of −1. Thus, the Politician would prefer in this case to set policy herself only when xB > 1. A core argument that consistently emerges from models of delegation is that as politicians’ policy uncertainty increases relative to that of the bureaucracy, it becomes more attractive for the politicians to delegate more policy-making authority to the bureaucrats (e.g. Epstein and O’Halloran 1994,1999; Bawn 1995).

The level of policy conflict between the Politician and Bureaucrat is also one of the central variables that influences the type of delegation strategy politicians adopt. The core argument about policy conflict, called the ally principle, is that all else equal, as the policy preferences of politicians and bureaucrats converge, politicians will delegate more discretion to bureaucrats. Return to the example where ɛ ∈ {0, 1}. If x B 1 2 , the Politician can delegate substantial discretion. She need not fear that in so doing she will invite the Bureaucrat to use his expertise against the Politician because the Bureaucrat’s interests are more or less aligned with the Politician’s. By contrast, if x B > 1 2 , the Politician must worry that the Bureaucrat will use his expertise against the Politician. The Politician responds to this concern by limiting discretion (and adopting z = 1 2 ). Bureaucratic discretion will therefore be largest when the bureaucrat is an ally of the politician.

Politicians influence bureaucratic behavior not only by using the ex ante strategies (i.e. actions that are taken before bureaucrats actually implement policies) to limit bureaucratic discretion described above. Following bureaucratic policy actions, politicians—and other political actors, such as courts or auditing agencies—can monitor bureaucratic behavior and attempt to influence their policy-making activities. Because of this second opportunity to influence agency actions, ex ante limits on discretion and ex post monitoring are often viewed as substitutes. Models generally show that politicians are more likely to prefer low discretion statutes when the monitoring environment is sufficiently weak. When the Politician can rely on ex post monitoring, she will have less incentive to pay the cost of writing a low discretion statute, as she’d prefer to rely instead on less costly ex post mechanisms. This substitution effect frequently recurs in the theoretical literature (e.g. Bawn 1997; Epstein and O’Halloran 1994; Gailmard 2002; Huber and Shipan 2002; Huber and McCarty 2004; Bendor and Meirowitz 2004).

The models discussed thus far have assumed a single politician whose preferences are stable during the play of the game. But a central distinguishing feature of delegating to bureaucrats, as opposed to delegating to employees, is that there typically is substantial uncertainty about the preferences of future politicians, since politicians are often replaced by others who have different policy goals. Political uncertainty, then, is uncertainty about the future preferences of politicians, and this variable is central to early theorizing about delegation by Moe (1989), Shepsle (1992), and Horn (1995). More recently, de Figueiredo (2002) explicitly models the effects of political uncertainty on discretion. In his model, politicians have an opportunity to “lock in” (p. 854) policy by limiting bureaucratic autonomy. The model suggests that politicians are most likely to do so when they feel it is unlikely they will control the political process in the future. This allows current politicians to tie the hands of future politicians by constraining bureaucratic autonomy.

2 Empirical Tests of the Four Arguments

After the publication of the early “structure and process” articles in the 1980s, considerable empirical research focused on whether ex ante efforts to restrict bureaucratic behavior actually allow effective congressional control over the bureaucracy in the USA.1 Unfortunately, this empirical literature is highly inconclusive, with some studies finding evidence of such effectiveness (e.g. Bawn 1997; Potoski 1999; Balla and Wright 2001; Potoski and Woods 2001), while others find either mixed support or no support at all (e.g. Hamilton and Schroeder 1994; Balla 1998; Spence 1999; Nixon, Howard, and DeWitt 2002). As we have argued previously (Huber and Shipan 2000, 2002), this lack of conclusiveness is not surprising given the elusiveness of pinning down the quality of control over bureaucracy. In our view, no one has yet been able to solve the very difficult problem of assessing empirical relationships between the preferences of bureaucrats and politicians on the one hand and policy outcomes on the other. Even if accurate measurement were possible, since all theoretical models indicate “control” will never be perfect, one would be left with the difficult task of imposing subjective judgements as to whether particular outcomes indicate sufficient control.

The four theoretical arguments described above, which emerge from a variety of related formal models, have highlighted an alternative pathway for empirical research. The models provide predictions about where final policy outcomes will be located vis-à-vis the preferences of bureaucrats and politicians. But they also provide predictions about the circumstances under which particular delegation strategies—such as more versus less discretion in legislation—should be adopted. Since these strategies are typically easier to observe than are outcomes, the predictions about delegation strategies are easier to test. And if empirical tests support the four arguments from the models, we can have confidence that the models are on the right track, and thus that the model’s implications for understanding delegation outcomes are worth taking seriously. So how have policy uncertainty, the ally principle, substitution effects, and political uncertainty fared in empirical tests?

One strategy for measuring the amount of discretion politicians give to bureaucrats is to look at the design of bureaucratic agencies. Volden’s (2002b) study of welfare (p. 855) boards in state governments provides one of the most impressive sets of empirical tests. Using a variety of dependent variables, including whether the legislature has delegated authority to a welfare board, whether powers are delegated to a policy board or an advisory board, and whether delegation is to independent agencies or those controlled by the executive, he finds robust support for the effects of policy uncertainty: when legislatures face more policy uncertainty in the form of greater demographic changes, they are more likely to delegate authority to agencies and to establish welfare boards. At the same time, he finds mixed support for the ally principle. Like many others, he begins by positing that policy conflict between legislator and bureaucrat will be highest during divided government (i.e. when the governor is of a different party from the legislature), but he finds that in actuality the effects of conflict are more subtle than that. When preferences are aligned, legislatures will delegate to welfare boards and give the government appointment powers. When preferences are not aligned, however, they are still willing to delegate; but they limit these grants of delegation by placing checks on the appointment powers of governors.

Wood and Bohte (2004) also focus on the design of administrative agencies. Drawing on the laws that established 141 US administrative agencies, they examine and code a number of design attributes, including the autonomy of leaders, the authority to engage in rule-making, budgetary autonomy, and the existing of reporting requirements. Their analysis provides substantial support for the ally principle, with features that foster independence more common when there are high levels of conflict (as measured by vetoes and overrides) between the legislative and executive branches.

Other tests have focused not on the design of agencies, but on the nature of legislation itself. Epstein and O’Halloran’s (1999) book-length treatment of delegation and the use of procedural means to control discretion is seminal. They measure discretion in legislation by coding the nature of procedural provisions in all key regulatory laws passed in the post-Second World War US Congress, identifying, for example, whether agencies are subject to time or spending limits and whether agencies are required to either report to or consult with the legislature about policy decisions and actions. Consistent with the ally principle, they find that preference conflict, measured by divided government, leads to legislation that places greater constraints on agencies. And consistent with the idea of policy uncertainty, they find that Congress delegates the most discretion to agencies in situations that are more informationally demanding.2

Huber and Shipan (2002) also focus on discretion levels in legislation. Their study differs from Epstein and O’Halloran in two significant respects. First, their study is broadly comparative, focusing on how variation in the institutional context in which politicians find themselves influences delegation strategies. To this end, they examine differences in delegation processes on the same issue across political systems. They therefore cannot test the uncertainty principle because issue type is held constant. Second, based on an analysis of labor laws across parliamentary democracies and health laws across the US states, they argue that the primary way politicians influence (p. 856) discretion is not by the inclusion of procedural details in legislation, but rather by specifying in more or less detail the substantive policies bureaucrats must implement. They find strong support for the ally principle, with divided government leading to less discretion in the states and coalition or minority government leading to less discretion in parliamentary systems. But they find that the degree to which coalition status or divided government diminishes discretion depends on other variables. Low discretion laws require legislative resources, and thus the ally principle is not robust to situations of low legislative professionalism in the American states, or to high levels of cabinet turnover in parliamentary democracies. And the ally principle depends on there being a clearly identifiable “politician.” In the American states, when only one chamber of the legislature is against the governor, while the other supports him or her, the legislature is less able to use legislative details to limit discretion than in cases where there is unified opposition to the governor across the two chambers.

Franchino also finds support for the uncertainty and policy conflict arguments in his studies of delegation in the European Union. He finds that the Council of Ministers delegates more authority to a supranational bureaucracy, the European Commission, in specialized and technical policy areas, as measured by the existence of detailed rules within the laws themselves, the creation of “action programmes,” and the use of specialized committees to write the law. He also finds that as conflict between member states increases in the Council, discretion for implementation typically declines (e.g. Franchino 2004, n.d.; see also 2001).

An impressive degree of support, then, is building for the ally principle and policy uncertainty. Scholars have also begun to test for, and find evidence of, substitution effects. For over two decades now, researchers have examined the extent to which ex post mechanisms cause bureaucrats to anticipate potential reprisals from legislatures and to modify their actions accordingly (e.g. Weingast and Moran 1983; Wood and Waterman 1991; Olson 1999; Shipan 2004). More recently, scholars have begun to examine the link between ex ante and ex post strategies (e.g. Aberbach 1990). If little oversight occurs, is this because bureaucrats anticipate what politicians desire, and thus need not be monitored directly, or because bureaucrats have substantial autonomy to do what they want? Bawn (1997) finds evidence of this substitution effect at the level of individual member of Congress: those members who sit on relevant oversight committees are least likely to seek ex ante limits on discretion.3 Looking across systems, Huber and Shipan (2002) find evidence of the substitution effect. In the American states, they find that the presence of a legislative veto over agency actions allows state legislatures to write less detailed statutes. And across parliamentary democracies, they show that countries write less detailed labor laws when the legal system is structured to protect the politicians’ interests as ex post and when corporatist bargaining arrangements give politicians an ex post check on labor policy.

Finally, several studies have begun to examine the effects of political uncertainty (i.e. uncertainty about the identity of future politicians). The Wood and Bohte (2004) study discussed earlier, for example, finds strong evidence that Congress is much (p. 857) more likely to limit agency discretion when political uncertainty is high (e.g. conflict exists within the enacting coalition and levels of electoral turnover are high). At the state level, Volden (2002b) finds that legislatures are far more likely to establish policy boards, which are insulated from political winds, if the dominant party in the legislature is in decline.4 In the cross-national context, Gilardi (2002) examines a range of policy areas and finds that agencies are likely to be most independent when political uncertainty is high. And Huber (2000) reports that politicians in parliamentary systems adopt global budgets for health care departments, thereby limiting discretion, when cabinet turnover is high.5

3 Implications of the Formal and Empirical Work on the Four Arguments

As the preceding section illustrates, empirical support is beginning to accumulate for the four arguments about delegation described above. Bureaucrats are granted more discretion when there is policy conflict between politicians and bureaucrats, when policy uncertainty is high, when ex post opportunities to influence bureaucratic behavior are limited, and when politicians have more certainty that they will retain power into the future. But these tests focus on patterns of delegation strategies, rather than on the substantive outcomes of political-bureaucratic interactions. If we take the results of the empirical tests to suggest that the models are on the right track, what are the more general lessons about the outcomes of delegation processes?

At the most basic level, one implication of these empirical tests is that some nontrivial level of political control over bureaucrats must exist. It is interesting to observe that these tests to date do not have clear measures of bureaucratic preferences. Instead, the typical approach is to assume that there exists a privileged political actor who exercises non-trivial control over bureaucrats. In studies of Congress, this political actor is the president, and empirical measures of preference conflict between the “Politician” (Congress) and the “Bureaucrat” (president who controls agencies) are usually captured by the presence or absence of divided government. In studies of the US states, divided government is also used, and in parliamentary democracies, scholars have focused on division across parties during coalition and minority government. With these proxies for preference conflict, we find substantial support for the ally principle. If in fact the president (or governor or cabinet minister) is not exercising influence over bureaucracy, it is very difficult to imagine why this support (p. 858) would exist in the data. To our knowledge, there is no competing theory that assumes the absence of political control and that yields the same prediction that legislators should give less discretion to agencies during periods of divided (or coalition) government.

Second, these results about divided and coalition government also remind us that “bureaucratic politics” is often simply one more arena where standard distributive battles among politicians are waged. Legislators who craft new legislation or design new agencies are not simply worried about how they might motivate recalcitrant bureaucrats. They are also worried about how they can use legislation and agency design to control rival politicians. Delegation, then, is often as much about using bureaucracy to control the actions of other politicians as it is about figuring out strategies to counter administrative dominance. This perspective on bureaucracy is not something one finds in the models themselves, but rather in the results from empirical tests of insights from the models.

Third, these models and empirical tests caution us about making inferences about administrative dominance by simply observing the strategies of politicians. If politicians exercise little ex post oversight, is this because they cannot influence bureaucrats? If legislators write vague statutes that give bureaucrats wide scope to fill in the policy details, does this mean that bureaucrats are likely to use their discretion to work against the interests of politicians? The fact that the data support the four arguments suggests a clear “no” to both questions. Politicians are likely to give substantial discretion to bureaucrats precisely in those situations where it is in the interest of politicians to do so, such as when the bureaucrat is an ally, or when the legislative majority can rely on effective ex post mechanisms—such as friendly judges—to help ensure desirable policy outcomes even in the presence of considerable bureaucratic discretion.

The fact remains, however, that with existing models and empirical techniques, we do not know how much control over bureaucrats exists. Indeed, the models underline that such control is quite difficult to achieve. In the stark model presented above, for instance, when policy uncertainty is low ( x ¯ { 0 , 1 } ) , the Politician delegates authority to the Bureaucrat only if policy conflict is relatively low ( x B 1 2 ) . As policy uncertainty increases, the Bureaucrat’s policy expertise becomes more valuable to the Politician, and thus the Politician delegates authority even if policy conflict is larger (xB ≤ 1). The model therefore underlines the simple fact that the ability of bureaucrats to usurp the policy-making role of politicians increases as their relative expertise increases.

Since empirical tests support the models with these policy uncertainty assumptions, one clear implication is that administrative dominance is most likely when technical complexity is high. This, of course, has been central to the study of delegation since at least the work of Max Weber. But the more general models go beyond this simple insight to help identify more precisely the conditions under which the ability of politicians to use delegation strategies to control bureaucrats is most limited. Consider models that combine ex ante and ex post strategies and assume that the politicians must pay a cost to adopt legislation that limits discretion. The models (p. 859) show that the ability of politicians to influence bureaucratic behavior is substantially limited only when several specific conditions exist:

  • (a) Politicians and bureaucrats disagree about which outcome to pursue (such as should exist during divided or coalition government);

  • (b) Politicians must lack the capacity to pay the cost of writing detailed statutes that limit discretion (such as in systems with unprofessionalized legislatures);

  • (c) Politicians must not be able to rely on ex post mechanisms (such as auditing or courts) to ensure desirable bureaucratic actions.

Only when these three conditions are simultaneously met can we infer that statutes granting large discretion to bureaucrats are essentially abdicating policy-making authority to them.

4 Theoretical Robustness: How Fragile are the Arguments?

The four arguments—policy uncertainty, the ally principle, substitution effects, and political uncertainty—are robust in two senses. First, they have emerged from a number of different types of models, albeit models that make similar assumptions about policy uncertainty. Second, empirical research is beginning to build support for them. Recent theoretical research, however, has begun to reveal circumstances under which the theoretical predictions falter. What are these circumstances, and what are the implications of these models for future studies of bureaucracy?

As noted above, the models of delegation to bureaucrats that generate the four arguments typically assume that bureaucrats have more expertise than politicians. But what if politicians try to redress this imbalance by investing in information themselves? Gailmard (2002) examines this question, and his model has implications for theoretical arguments about preference divergence and delegation. For reasons similar to those discussed in conjunction with the stark model above, his model illustrates that as the preferences of a Bureaucrat diverge from a Politician’s, the value of policy expertise to the Politician increases. But this fact, Gailmard observes, should create incentives for the Politician to invest more in information when these preferences diverge. Since the Politician’s policy expertise increases with preference divergence, the Politician has less need to grant the Bureaucrat discretion. Thus, with endogenous specialization by the Politician, discretion can decrease with policy conflict because policy uncertainty can also decrease.

Bendor and Meirowitz (2004) take a different approach to fleshing out the origins of information asymmetries. Like Gailmard, they focus on endogenous investment in expertise, but unlike Gailmard, they focus on incentives for bureaucrats (p. 860) to become informed. In one variant that they analyze, a Bureaucrat cannot precommit to becoming informed about ɛ, but rather must pay a cost to do so after the discretion level is announced. In this model, the ally principal can fail. If the Politician does not delegate, the Bureaucrat’s expected utility is that associated with the Politician’s uninformed policy choice. Since this is a better outcome for an allied Bureaucrat than for one with divergent preferences, the Bureaucrat with divergent preferences has more incentive to pay the cost of specialization. The Politician, recognizing this incentive to specialize, is more likely to delegate to a non-ally bureaucrat under certain levels of policy uncertainty. Both the Gailmard and the Bendor– Meirowitz model suggest, then, that the ally principle may depend a great deal on the assumption of some exogenous source of information asymmetry. If one tries to make this asymmetry endogenous, then the ally principle can be turned on its head, with more discretion going to bureaucrats who are further from the politician’s ideal point.

The standard models from which the ally principle emerges also assume that bureaucrats are very good at what they do (i.e. that they have high “capacity”)—bureaucrats not only know ɛ, they can execute whatever policy action they wish. But even if individuals at the top of a bureaucratic hierarchy know which policy consequences will result from adopting a particular policy, they may be unable to execute this policy. Huber and McCarty (2004) explore delegation when bureaucratic capacity is low. Their model distinguishes between policy uncertainty, as described above, and bureaucratic capacity, which is the Bureaucrat’s ability to execute his intended action. A Bureaucrat with low capacity has less ability to take the action he intends, even if he knows ɛ. Low capacity therefore makes it more difficult for the Bureaucrat to comply with a statute—the Bureaucrat may try to comply, but his lack of capacity may push the action he intends out of compliance. Consequently, a low-capacity Bureaucrat with preferences that diverge from the Politician’s is less likely to make policy concessions toward the Politician’s ideal point because the impact of such efforts on policy compliance is relatively low. In this situation, the Politician must give more discretion to the “enemy” Bureaucrat. This extra discretion induces the “enemy” Bureaucrat to forgo attempting to implement his most preferred policy, and instead to make policy concessions toward the Politician’s ideal point. Consequently, the ally Bureaucrat, who pays a much lower policy cost of his unreliable attempts at policy compliance, receives less discretion than the “enemy” Bureaucrat.

The ally principle stems from models where a single Politician delegates to a single Bureaucrat. In many political systems, however, the distribution of agenda power for, say, setting discretion levels is divided among several politicians. If the multiple politicians are making collective decisions by majority rule in a single chamber, then the existence of multiple politicians may not be much of an issue. One could simply assume that the “politician” who seeks an ally in the bureaucracy is the median legislator, or some other pivotal politician. But often institutional separation of powers will create multiple pivotal politicians. In such cases, how does one think about an ally principle?

(p. 861) Studies of this question focus primarily on presidential systems. These studies assume legislatures propose and presidents veto, and typically explore the extent to which discretion levels—which are proposed by the legislature, not the president—are influenced by the extent to which the bureaucracy is an ally of the legislature. Volden (2002a), for example, develops a model where an executive can veto legislation that sets discretion levels. His model suggests that when the executive and legislature have divergent preferences, the ally principle will hold for some configurations of status quos and ideal points, but will fail for others. If the status quo is extreme relative to the preferences of the legislature and executive agent, for instance, then as the ideal points of the legislature and agent diverge, with the agent moving closer to the status quo, more discretion must be given to the agent in order to induce the agent to change the status quo. Similar results are found in Bendor and Meirowitz (2004) and Huber and Shipan (2002).

McCarty (2004) examines the multi-principal issue through the lens of the appointment processes. The politicians who grant discretion, McCarty notes, are often different from those who control appointments, with non-trivial consequences for strategic behavior. In his model, a President, who has the power to appoint a Bureaucrat, is constrained by the Legislature, which grants policy-making discretion to this Bureaucrat by providing budgetary resources to change the policy. If the President appoints a Bureaucrat who is too close to the President’s ideal point, then the Legislature, whose preferences diverge from the President, will write laws that limit discretion. Given the Bureaucrat’s expertise, both the President and the Legislature can be made better off if the President appoints a Bureaucrat who is between the President’s and the Legislature’s ideal point rather than at the President’s ideal point. To this end, the President must be able to commit to not changing the Bureaucrat’s ideal point after the Legislature establishes the discretion level. Constitutional or statutory provisions that limit the power of the President to remove bureaucrats from office are one obvious source of such precommitment.

Ting (2001) examines the possibility that a different sort of commitment problem can be overcome in a repeated-play setting. His model describes why we might expect ex ante and ex post strategies to be complements rather than substitutes. In Ting’s model, the Politician establishes a budget, which determines bureaucratic discretion by constraining the types of policy that the Bureaucrat can implement. After the Bureaucrat implements a policy, the Politician can invest resources in ex post monitoring. Without repeated play, the Politician’s auditing strategy is independent of the outcome she observes because she cannot precommit to conditioning intensity of audits on this outcome. If the game is repeated, with a new ɛ drawn in each period, the Politician can indirectly condition auditing on outcomes by conditioning future budget levels (which influence audit strategies) on outcomes. This can only occur, however, if the Politician and Bureaucrat have sufficiently similar preferences. In such cases, ex ante discretion and ex post monitoring are not substitutes, but complements. Larger budgets provide greater discretion, and since without precommitment politicians cannot condition monitoring on observing outcomes, they will increase discretion (i.e. budgets) only when they can increase monitoring.

(p. 862) 5 Conclusion

Considerable energy has been devoted to developing models of the strategies that politicians use to shape bureaucratic behavior. Central to these models are two assumptions: first, bureaucrats, left to their own devices, will pursue outcomes that are not necessarily in the interest of politicians, and second, bureaucrats have more policy expertise than politicians. From these two assumptions, four theoretical arguments consistently emerge regarding delegation strategies. First, as policy uncertainty increases, politicians can achieve better outcomes by giving more discretion to bureaucrats. Second, as the policy preferences of politicians and bureaucrats converge, politicians can achieve better outcomes by giving more discretion to bureaucrats. Third, as ex post mechanisms to achieve policy goals become more effective, politicians should give bureaucrats more discretion. And fourth, greater political uncertainty provides current politicians with the incentive to shield bureaucratic actions from future political influence.

Empirical research, much of it on the US Congress or the US states, has begun to build support for these arguments. We can therefore have some confidence that the predictions from these models—not simply about delegation strategies, but also about delegation outcomes—are on the right track. This has allowed us to understand the strategies that politicians can use to create incentives for bureaucrats to act on behalf of politicians, and also to understand the circumstances under which exercising effective control over bureaucrats is most difficult.

As impressive as the research has been at bringing politics and political institutions into the study of bureaucracy, many questions remain unanswered. First, as we noted, the ally principle and the uncertainty principle are foundations for the study of delegation, but recent formal models suggest that both principles are fragile. The models, for example, identify specific conditions that lead to the failure of the ally principle, conditions related to endogenous specialization, low bureaucratic capacity, or separation of powers with extreme status quos. Yet most empirical work focuses on straightforward tests of the two principles, without taking into account the factors that models suggest could lead to their failure. By taking seriously some of the nuances in recent theoretical models, future empirical research could shed new light on the extent to which these two simple but pervasive principles shape delegation processes. One clear domain for testing models where bureaucrats are uninformed or of low capacity is in developing democracies. In any event, these most recent models require empirical scholars to consider carefully the precise circumstances under which it is appropriate to test the four arguments that have largely dominated the literature.

Second, courts represent one of the central yet neglected ways in which politicians can influence policy outcomes and agency actions. Empirical work (e.g. Spriggs 1996; Hanssen 2000; Howard and Nixon 2002; Canes-Wrone 2003) has examined how judges influence bureaucratic actions, or how legal structures influence ex ante delegation strategies (e.g. Huber and Shipan 2002). And spatial models have begun to examine similar issues (e.g. Shipan 2000). Yet given the relevance of the (p. 863) legislative–judicial–bureaucratic relationship to political science and to administrative law (e.g. Shapiro 2002), this is clearly an area that is ripe for further research.

Finally, studies of delegation—particularly formal models—typically rest on the premiss that politicians adopt strategies that lead to the best obtainable policy choice by a bureaucrat in some current policy space. But the choices politicians make in delegating today can have ramifications long into the future. As Lewis (2004) points out, for example, a politicized bureaucracy, where politicians can make numerous bureaucratic appointments, might allow politicians to ensure the presence of their allies in the bureaucracy. This benefit, however, could be accompanied by a nontrivial cost, which is that it will be more difficult to attract quality personnel to work in the non-politicized parts of bureaucracies. Similarly, scholars have noted that basic features of bureaucratic organization, such as the existence of meritocratic recruitment, influence corruption levels (Rauch and Evans 2000) or economic growth (Evans and Rauch 1999).

It is clear, then, that there exists considerable variation in such institutions within bureaucracies—that is, in the politicization of appointments, pay levels of bureaucrats, rules for hiring and promotion, and so on. But little attention has been paid to explaining the sources or consequences of this variation. Thus, linking studies of the politics of delegation to studies of structures internal to bureaucracies, which have been chiefly the domain of sociology (but see Whitford 2002), represents one of the most important avenues for future research on bureaucracy.


Aberbach, J. D. 1990. Keeping a Watchful Eye. Washington, DC: Brookings Institution.Find this resource:

Balla, S. J. 1998. Administrative procedures and political control of the bureaucracy. American Political Science Review, 92: 663–74.Find this resource:

——2000. Legislative organization and congressional review of agency regulations. Journal of Law, Economics, and Organization, 16: 424–48.Find this resource:

—and Wright, J. R. 2001. Interest groups, advisory committees, and congressional control of bureaucracy. American Journal of Political Science, 45: 799–812.Find this resource:

Bawn, K. 1995. Political control versus expertise: congressional choices about administrative procedures. American Political Science Review, 89: 62–73.Find this resource:

—1997. Choosing strategies to control the bureaucracy: statutory constraints, oversight, and the committee system. Journal of Law Economics and Organization, 13: 101–26.Find this resource:

Bendor, J. and Meirowitz, A. 2004. Spatial models of delegation. American Political Science Review, 98: 293–310.Find this resource:

Bernhard, W. 2002. Banking on Reform. Ann Arbor: University of Michigan Press.Find this resource:

Canes-Wrone, B. 2003. Bureaucratic decisions and the composition of the lower courts. American Journal of Political Science, 47: 205–14.Find this resource:

de Figueiredo, R. J. P., Jr. 2002. Electoral competition, political uncertainty, and policy insulation. Amer ican Political Science Review, 96: 321–33.Find this resource:

—2003. Budget institutions and political insulation: why states adopt the item veto. Journal of Public Economics, 87: 2677–701.Find this resource:

(p. 864) Epstein, D. and O’Halloran, S. 1994. Administrative procedures, information, and agency discretion. American Journal of Political Science, 38: 697–722.Find this resource:

——1999. Delegating Powers. New York: Cambridge University Press.Find this resource:

Evans, P. and Rauch, J. 1999. Bureaucracy and growth: a cross-national analysis of the effects of “Weberian” state structures on economic growth. American Sociological Review, 64: 748–65.Find this resource:

Ferejohn, J. and Shipan, C. 1990. Congressional influence on bureaucracy. Journal of Law Economics and Organization, 6: 1–20.Find this resource:

Franchino, F. 2001. Delegation and constraints in the national execution of the EC policies: a longitudinal and qualitative analysis. West European Politics, 24: 169–92.Find this resource:

—2004. Delegating powers in the European Community. British Journal of Political Science, 34: 269–93.Find this resource:

—n.d. The powers of the Union: who does what and why in the European Union. Manuscript, University College London.Find this resource:

Gailmard, S. 2002. Expertise, subversion, and bureaucratic discretion. Journal of Law Economics and Organization, 18: 536–55.Find this resource:

Giladi, F. 2002. Policy creditility and delegation to independent regulatory agencies: a comparative empirical analysis. Journal of European Public Policy, 9: 873–93.Find this resource:

Hamilton, J. T. and Schroeder, C. H. 1994. Strategic regulators and the choice of rulemaking procedures: the selection of formal vs. informal rules in regulating hazardous waste. Law and Contemporary Problems, 57: 111–60.Find this resource:

Hanssen, F. A. 2000. Independent courts and administrative agencies: an empirical analysis of the states. Journal of Law Economics and Organization, 16: 534–71.Find this resource:

Horn, M. J. 1995. The Political Economy of Public Administration. New York: Cambridge University Press.Find this resource:

Howard, R. M. and Nixon, D. C. 2002. Regional court influence over bureaucratic policymaking: courts, ideological preferences, and the Internal Revenue Service. Political Research Quarterly, 55: 907–22.Find this resource:

Howell, W. G. and Lewis, D. E. 2002. Agencies by presidential design. Journal of Politics, 64: 1095–114.Find this resource:

Huber, J. D. 2000. Delegation to civil servants in parliamentary democracies. European Journal of Political Research, 37: 397–413.Find this resource:

—and McCarty, N. 2004. Bureaucratic capacity, delegation and political reform. American Political Science Review, 98: 481–94.Find this resource:

—and Shipan, C. R. 2000. The costs of control: legislators, agencies, and transaction costs. Legislative Studies Quarterly, 25: 25–52.Find this resource:

——2002. Deliberate Discretion: The Institutional Foundations of Bureaucratic Autonomy. New York: Cambridge University Press.Find this resource:

Lewis, D. E. 2003. Presidents and the Politics of Agency Design. Stanford, Calif.: Stanford University Press.Find this resource:

—2004. Presidents and the politicization of the institutional presidency. Working paper, Princeton University.Find this resource:

McCarty, N. 2004. The appointments dilemma. American Journal of Political Science, 48: 413–28.Find this resource:

McCubbins, M. D., Noll, R. G., and Weingast, B. R. 1987. Administrative procedures as instruments of political control. Journal of Law Economics and Organization, 3: 243–77.Find this resource:

———1989. Structure and process, politics and policy: administrative arrangements and the political control of agencies. Virginia Law Review, 75: 431–82.Find this resource:

(p. 865) Moe, T. M. 1989. The politics of bureaucratic structure. In Can the Government Govern? ed. J. E. Chubb and P. E. Peterson. Washington, DC: Brookings Institution.Find this resource:

Nixon, D. C., Howard, R. M., and DeWitt, J. R. 2002. With friends like these: rulemaking comment submissions to the Securities and Exchange Commission. Journal of Public Administration Research and Theory, 12: 59–76.Find this resource:

Olson, M. K. 1999. Agency rulemaking, political influences, regulation and industry compliance. Journal of Law Economics and Organization, 15: 573–601.Find this resource:

Potoski, M. 1999. Managing uncertainty through bureaucratic design: administrative procedures and state air pollution control agencies. Journal of Public Administration Research and Theory, 9: 623–39.Find this resource:

—and Woods, N. 2001. Designing state clean air agencies: administrative procedures and bureaucratic autonomy. Journal of Public Administration Research and Theory, 11: 203–22.Find this resource:

Rauch, J. and Evans, P. 2000. Bureaucratic structure and bureaucratic performance in less developed countries. Journal of Public Economics, 74: 49–71.Find this resource:

Shapiro, M. 2002. Judicial delegation doctrines: the US, Britain, and France. West European Politics, 25: 173–99.Find this resource:

Shepsle, K. A. 1992. Bureaucratic drift, coalitional drift, and time inconsistency: a comment on Macey. Journal of Law, Economics and Organization, 8: 111–18.Find this resource:

Shipan, C. R. 2000. The legislative design of judicial review: a formal analysis. Journal of Theoretical Politics, 12: 269–304.Find this resource:

—2004. Regulatory regimes, agency actions, and the conditional nature of political influence. American Political Science Review, 93: 467–80.Find this resource:

Spence, D. B. 1999. Managing delegation ex ante: using law to steer administrative agencies. Journal of Legal Studies, 28: 413–59.Find this resource:

Spriggs, J. F. 1996. The Supreme Court and federal administrative agencies: a resource-based theory and analysis of judicial impact. American Journal of Political Science, 40: 1122–51.Find this resource:

Ting, M. M. 2001. The “power of the purse” and its implications for bureaucratic policymaking. Public Choice, 106: 243–74.Find this resource:

Volden, C. 2002a. A formal model of the politics of delegation in a separation of powers system. American Journal of Political Science, 46: 111–33.Find this resource:

—2002b. Delegating power to bureaucracies: evidence from the states. Journal of Law Economics and Organization, 18: 187–220.Find this resource:

Weingast, B. R. and Moran, M. J. 1983. Bureaucratic discretion or congressional control: regulatory policy-making by the FTC. Journal of Political Economy, 91: 765–800.Find this resource:

Whitford, A. B. 2002. Decentralization and political control of the bureaucracy. Journal of Theoretical Politics, 14: 167–93.Find this resource:

Wood, B. D. and Bohte, J. 2004. Political transaction costs and the politics of administrative design. Journal of Politics, 66: 176–202.Find this resource:

—and Waterman, R. W. 1991. The dynamics of political control of the bureaucracy. American Political Science Review, 85: 801–28.Find this resource:


(*) We are grateful to Mike Ting for helpful comments and to Eduardo Leoni for research assistance.

(1) Howell and Lewis 2002 and Lewis 2003 provide evidence that presidents successfully use such tools to influence agencies.

(2) Potoski 1999 finds similar effects in US states.

(3) See, however, Balla 2000.

(4) De Figueiredo 2003 also provides evidence about the effects of political uncertainty on policy insulation in US state legislatures.

(5) See also Bernhard 2002 on the political factors that lead to central bank independence.