sign in sign up
<<
Home , Statute

American Political Science Review Vol. 98, No. 3 August 2004

Bureaucratic Capacity, , and Political Reform JOHN D. HUBER Columbia University NOLAN MCCARTY Princeton University e analyze a model of delegation and policymaking in polities where bureaucratic capacity is low. Our analysis suggests that low bureaucratic capacity diminishes incentives for bureaucrats W to comply with , making it more difficult for politicians to induce bureaucrats to take actions that politicians desire. Consequently, when bureaucratic capacity is low, standard principles in the theoretical literature on delegation no longer hold. We also use the model to examine the issue of political reform in polities with low bureaucratic capacity. The model indicates that politicians in such polities will be trapped in a situation whereby they have little incentive to undertake reforms of either the or other (such as ) that are crucial for successful policymaking.

he central tension motivating contemporary the- corruption. For a variety of reasons, then, it may be ories of delegation from politicians to bureau- very difficult for to implement policies Tcrats lies in the potential conflict between the effectively, even when leaders within the bureaucracy value of bureaucratic expertise, on one hand, and the have sufficient expertise to understand which policies desire for political control, on the other. This tension will yield desired outcomes. The problem for politi- is well-known, and has been debated at least since cians in such contexts is not how to create appropriate Weber’s classic work on bureaucracy. Modern bureau- incentives for high-powered bureaucrats. Instead, the cracies are staffed with individuals who, by virtue of problem is how to make policy when bureaucracy is “rational” bureaucratic organization, are highly skilled known to lack capacity. policy experts who in principle should be able to We develop a model of delegation that can help help less knowledgeable politicians achieve their goals. understand policymaking processes when bureaucratic However, the very skills and expertise that bureau- capacity is lacking. A central feature of our approach is crats enjoy create the possibility that bureaucrats will to distinguish the problem of information (knowledge usurp the rightful role of politicians in policymaking about the consequences of different actions) from the processes. Because the way this tension is resolved de- problem of capacity (the ability to accomplish intended termines the ultimate compatibility of democracy and actions). The information problem has dominated the bureaucracy, existing theories of delegation typically existing delegation literature, whereas the capacity is- focus on understanding how and to what extent politi- sue has been essentially ignored. We argue that the cians can resolve it.1 problems of information and capacity are conceptually In many contexts, particularly developing democra- distinct. Senior bureaucrats, for example, may have pol- cies, bureaucracy does not fit the “Weberian” charac- icy expertise but, because of the problems described teristics that are assumed in the existing delegation in the previous paragraph, be unable to execute re- literature. The bureaucracies may be staffed with in- liably the policies they intend. By disentangling the dividuals who lack the personal capacity to execute effects of expertise and capacity on delegation, we can orders effectively. The organizational structure may achieve important insights into policymaking in situa- lead to breakdowns within bureaucratic hierarchies, tions where bureaucratic capacity is low. making it difficult for senior administrators to control Specifically, we use the model of delegation to low- their subordinates. Departments and ministries may capacity bureaucracies to explore three substantive lack basic budgets and resources to undertake imple- issues. First, we examine how low bureaucratic ca- mentation actions. They may suffer from incentives for pacity diminishes the ability of politicians to achieve their policy goals. Our main argument is that although John D. Huber is Professor of Political Science, Department of Po- low bureaucratic capacity reduces the general qual- litical Science, Columbia University, New York, NY 10025 (jdh39@ ity of bureaucratic outcomes though a straightforward columbia.edu). efficiency loss, it also changes policymaking in a less Nolan McCarty is Professor of Politics and Public Affairs, Woodrow Wilson School of International and Public Affairs, Prince- obvious way. In particular, it diminishes incentives for ton University, Princeton, NJ 08540 ([email protected]). bureaucrats to comply with legislative statutes. Put dif- We would like to thank John Carey, David Epstein, David Lewis, ferently, bad bureaucracies are not only inefficient (i.e., and Eric Schickler for comments on early versions of the manuscript. less successful at implementing the policies they in- We are also grateful for feedback from participants in seminars at UC—–Berkeley and the University of Michigan and from participants tend), but also harder to control because their incom- in conferences on the Macropolitics of Congress and Brazilian Po- petence diminishes their incentives to implement the litical Institutions in Comparative Perspective. Huber is grateful to policies politicians describe in legislation. For this rea- Princeton’s University’s Center for the Study of Democratic Politics, son, politicians can often induce more desirable ac- where he was a visiting fellow during work on this project. tions from highly competent “enemy” bureaucrats (i.e., 1 Influential works include informal arguments by McCubbins, Noll, and Weingast (1987), the formal model in Bawn 1995, and the for- bureaucrats with policy preferences that differ from mal model and empirical tests in Epstein and O’Halloran 1999. See the politician’s) than from less competent “friendly” Huber and Shipan 2002, chap. 2, for a recent review of the literature. ones (who share the politician’s policy preferences).

481 Bureaucratic Capacity, Delegation, and Political Reform August 2004

Importantly, however, politicians need some level of public-sector salaries that are sufficiently high to attract policy expertise to take advantage of bureaucratic com- skilled individuals into public service (e.g., Besley and petence. Without it, competent bureaucrats can often McLaren 1993), the nature of social networks among work successfully to achieve outcomes that politicians bureaucrats (e.g., Carpenter 2001), insulation of agency oppose. budgets from excessive political influence (e.g., Geddes Second, we use the model to understand how delega- 1994), and training programs (e.g., McCourt and Sola tion strategies in situations of low bureaucratic capacity 1999). differ from delegation strategies that have been de- Importantly,low bureaucratic capacity can exist even scribed in previous research on high-capacity bureau- when high-level bureaucrats know what they would cracies. Our model suggests that when bureaucratic like to accomplish. A senior administrator in a depart- capacity is low, well-established principles in the ex- ment of education, for example, may have a clear idea isting literature no longer hold. In models that assume of which curriculum for students will yield the best test high-capacity bureaucrats, for example, politicians typ- scores, but implementation of this curriculum may be ically delegate more discretion to bureaucrats when difficult and uneven given variation in the quality of the bureaucrats are ideological allies and when ex post school principals and teachers. Or it may be the case monitoring possibilities are most effective. We find that a senior bureaucrat charged with administering a the opposite in low capacity systems. Thus, predictions pension system can establish a clear set of rules about about discretion in legislative statutes may not apply in eligibility and payments. But this bureaucrat may be political systems where bureaucratic capacity is low. unable to enforce adequately the application of these Finally, we use the model to understand incentives rules, as subordinates may favor particular individuals for political reform in polities with low bureaucratic over others, refuse pensions where they are deserved, capacity. We argue that politicians in such polities are or provide them to individuals who are ineligible. Or trapped in a situation where they have little incen- a senior bureaucrat may be assisted by international tive to reform not only the bureaucracy, but other development agencies regarding how to build a water institutions as well. The incentives of politicians to project but be unable to motivate subordinates to get reform low-capacity bureaucracies increase as politi- the job done right. We want to focus, then, on problems cians gain more policy expertise, for instance, but the that senior bureaucrats often face when attempting to incentives of politicians to gain policy expertise are have their subordinates implement policy and, there- smallest when bureaucratic capacity is low. Similarly, fore, to separate this problem of “capacity” in imple- incentives to reform low-capacity bureaucrats increase mentation from the problem of “expertise.” when enforcement of statutes—–which is often done Our approach to modeling policymaking in the pres- by courts—–becomes more effective, but incentives to ence of capacity and expertise problems is as follows. create effective enforcement are diminished by low- The model has two players, a Politician who writes bureaucratic capacity. In addition to these problems, and a Bureaucrat who implements them.∗ Interactions we also find that the incentives of politicians to “politi- occur in a one-dimensional policy space where the cize” the bureaucracy are largest when bureaucratic Politician has an ideal point at xP = 0 and the Bureau- capacity is low and that, when bureaucrats are “politi- crat has an ideal point at xB > xP. Each player has a cized,” incentives to reform bureaucracy are lowest. quadratic utility function over policy outcomes. The game begins with the Politician adopting a statute that establishes the domain of implementation THE MODEL actions that the Bureaucrat can take while complying with the . The Politician may lack expertise (i.e., be The model we describe in this section focuses on a uncertain about which specific policies best serve her specific notion of bureaucratic capacity that is related interests). Following the adoption of the statute, the to the ability of senior members of a bureaucracy to Bureaucrat, who has policy expertise, attempts to im- implement desired actions. We are motivated, in par- plement policy. The Bureaucrat’s success at executing ticular, by previous descriptions of inefficiency and in- the desired action depends on the Bureaucrat’s capac- competence in policy implementation in developing ity. In the final stage, if the Bureaucrat’s action does not countries. Geddes (1994, 17), for example, points out comply with the statute, he is caught and punished with that in the early 1970s, leaders in Chile’s bureaucracy some probability. We interpret this probability as the clearly had the knowledge and expertise for successful effectiveness of oversight and of judicial enforcement. nationalization of Chilean copper mines, but general This general structure is therefore similar to the model mismanagement, prompted by political incentives to developed by Huber and Shipan (2002). A central treat important political positions as patronage plums, difference is that the model here adopts assumptions led to very poor performance of these mines. Geddes about bureaucratic capacity that follow Chang, Lewis, (1990) similarly describes a very incompetent bureau- and McCarty (2000). cracy in the 1930s in Brazil, when the government could To model bureaucratic capacity in the sense de- not even locate the documents that were necessary scribed above, we assume that if the Bureaucrat at- to determine how much money Brazil owed foreign tempts action a, then the realization of this action is lenders. Other scholars describe specific factors that influence basic bureaucratic capacity, such as merit- ∗ We use in our model a female politician and male bureaucrat to based pay and promotion (e.g., Evans and Rauch 1999), avoid confusion in the use of pronouns.

482 American Political Science Review Vol. 98, No. 3 a − ω, where ω is a random variable with a probabil- As noted, the game begins when the Politician adopts ity density function f (ω) = ( −|ω|)/2 on the inter- a law. This law specifies the upper and lower bound val [−, ]. Therefore, ω is distributed symmetrically on policies that the Bureaucrat can implement while 2 = 2 2 around a mean of zero with a variance σω ( /6). remaining compliant with the law. Formally, this law is Because it is directly related to the variance of ω and x = [x, x¯]. Because we assume that xP < xB, the Politi- therefore to the Bureaucrat’s ability to control the real- cian will set x as low as possible because the Bureau- ized action,  represents bureaucratic incapacity in our crat never has an incentive to take actions that are model. When  is very small, the Bureaucrat’s capac- lower than what an informed would have ity is large. As  increases, so too does the possibility wanted. To simplify the analysis, we therefore assume of large errors in implementation. Because we wish that x =−∞and focus our analysis on the location of to understand how bureaucratic incapacity influences x¯. Following the adoption of the law, the Bureaucrat institutional reform and policymaking in developing attempts to implement a, the realization of this action polities, we focus on cases where  is sufficiently large is a − ω, and the policy outcome is a − ω − ε. (i.e., where bureaucratic capacity is reasonably small). A central feature of our model is the assumption that We make this explicit below. the policy that the Bureaucrat implements (i.e., a − ω) Our assumption about policy expertise follows the may or may not comply with the statute (i.e., it may or standard approach in the literature. Politicians of- may not be the case that a − ω ∈ [x, x¯]). We assume that ten understand what policy outcomes they want to if the policy is noncompliant (i.e., a − ω/∈ [x, x¯]), then achieve—–such as an efficient pension system or a re- with some probability, γ, the Bureaucrat is caught in liable supply of clean water—–but may be uncertain noncompliance and pays a penalty, δ. Our model does about which specific policy will achieve this objective. not assume, then, that the Bureaucrat must comply A pension system might be accomplished through pri- with the law. The Bureaucrat may try to do so, but vate contributions to insurance schemes or payments fail, because of a large implementation error. Or he from general revenues, for example, and water supply may not even try, instead attempting an action that is might be assured by digging wells or making dams. In known to be out of compliance with the statute. The our analysis, we wish to treat the Politician’s degree of parameters γ and δ therefore capture the effectiveness policy uncertainty as a variable. Specifically, we assume of the political system, particularly the courts, at un- that if some policy a − ω is implemented, the ultimate covering and punishing actions by the Bureaucrat that policy outcome is a − ω − ε, where ε ∼ U [−E, E] with do not comply with statutes. 2 = 2 variance σε (E /3). Each player j ’s preferred policy It is useful to note that the punishment of bureau- action is therefore xj + ε. In this paper, we assume that crats for noncompliance is based on the realization of the Bureaucrat knows ε, capturing the fact that in most the Bureaucrat’s attempted action, a − ω, rather than political systems, senior-level bureaucrats have more on the policy outcome itself, a − ω − ε. We assume that resources for obtaining policy expertise than politi- it is never possible to observe a, the Bureaucrat’s in- cians. An important avenue for future research would tended action. We also assume that the Bureaucrat be to examine cases where expertise in the bureaucracy cannot be punished if the realization of his action com- is also variable.3 plies with the statute (a − ω ∈ [x, x¯]). It may be the case The smaller is E, the lower the Bureaucrat’s expertise that a − ω − ε/∈ [x, x¯]. However, the Bureaucrat cannot advantage over the Politician. At the extreme, where be punished for this of the world if he actually E = 0, both the Politician and the Bureaucrat know the complied with what the statute says. consequences of adopting any specific policy. We ma- Politicians, for example, may want to contain health- nipulate the value of E to examine two issues. The first care costs. A law may state that global budgeting concerns how the Politician’s level of policy uncertainty procedures should be adopted for individual hospitals (the value of E) influences policymaking processes. The to achieve this goal. It may be that senior bureaucrats second concerns political development: How does bu- instruct their team to develop these budgeting systems. reaucratic capacity influence politicians’ incentives to If in fact some bureaucrats do not implement the re- create institutions that help them to develop expertise quired budgeting systems for their hospitals, this would (e.g., that make E smaller), and how does the level be a situation where the policy is not compliant with of expertise affect incentives to improve bureaucratic the statute (and a − ω/∈ [x, x¯]). In this case, if caught in capacity? noncompliance, the bureaucrats could be punished. By It is important to bear in mind that the ultimate out- contrast, it may be the case that all of the budgeting pro- come from any action attempted by the Bureaucrat is a cedures were put in place (a − ω ∈ [x, x¯]), but the end function of capacity (ω) and the policy shock (ε). Thus, result did not save money, as was expected. In this sec- the final outcome from the action, a, attempted by the ond case, a − ω − ε/∈ [x, x¯], yet the bureaucrat should Bureaucrat is a − ω − ε. not be punished for doing what the law prescribed. This possibility that faithful execution of the law leads to bad outcomes represents one of the risks to an uninformed 2 For our results, only the unimodality and symmetry of ω is crucial. legislator (one who is uncertain about the value of ε) The density of ω is that of the sum or difference of two uniform of placing policy constraints on the Bureaucrat—–that random variables with zero mean. 3 Bendor and Meirowitz (2004) provide a model where levels of ex- is, the Politician may force the Bureaucrat to take pertise vary across bureaucracies, but each is assumed to implement actions that, given the actual value of ε, are bad from its intended policy choice perfectly. both the Bureaucrat’s and the Politician’s perspective.

483 Bureaucratic Capacity, Delegation, and Political Reform August 2004

FIGURE 1. Sequence of Interactions in the Model

Politician Bureaucrat Realization of If realized action, Final policy takes action, enacts action occurs, a –ω ∉ [ x, x], then outcome is statute, a, to a –ω , where − ω − ε Bureaucrat is caught a , [,]xx implement ω ∈[]−Ω, Ω in non-compliance where policy ε ∈ [ − ] with probability γ , EE, and, if caught, pays penalty δ

FIGURE 2. The Bureaucrat’s Optimal Response tox ¯

f ()ω

* = x +Ω a x −Ω x xaB 2 B x 1 a3 B 1 Ω x − ε

Note: The example in this figure assumes that ε = 0.

Figure 1 summarizes the model. In the next section, environment.4 This responsiveness can work both for we examine how bureaucratic capacity influences the and against the Politician’s interests, for reasons we Bureaucrat’s implementation strategy. spell out below. In what follows, we describe the logic of these two effects and state formally the optimal re- sponse of the Bureaucrat to any statute. The Bureaucrat reacts to a statute by attempting an action, a. He knows that the realization of this effort BUREAUCRATIC CAPACITY AND POLICY − IMPLEMENTATION is a ω but does not know the value of ω (only that ω ∈ [−, ]). In choosing the optimal a, then, the Bu- In the model, one reason the Politician is harmed by a reaucrat must weigh the (potential) costs and benefits low capacity Bureaucrat is because of a straightforward of moving a toward his ideal point. Such moves can 2 = 2 efficiency loss. This efficiency loss is simply σω  /6, lead to the benefit of better policy outcomes, but at the which is increasing in . In this section, by exam- cost of possibly being punished for noncompliance. ining the Bureaucrat’s best response to any statute, Figure 2 illustrates the Bureaucrat’s decision prob- we show that bureaucratic capacity has two distinct lem given a statute,x ¯, and given an arbitrary level of ca- effects on policymaking in addition to this efficiency pacity, , which has a probability density function f (ω). effect. On one hand, the best action the Bureaucrat In Figure 2 we assume that the Bureaucrat knows that can be induced to take improves for the Politician as ε = 0. The figure depicts three different actions—–a1, a2, bureaucratic capacity increases. In fact, the Politician and a3—–along with the distributions of realizations can often induce a better action from a high-capacity of these actions, which are denoted by the triangles “enemy” Bureaucrat (with an ideal point far from the centered above each action. The attempted action Politician’s) than from a low-capacity “friendly” Bu- reaucrat (with an ideal point close to the Politician’s). On the other hand, a high-capacity Bureaucrat’s action 4 Bendor and Meirowitz (2004) make a similar point in a model with is more responsive to his information about the policy variable expertise.

484 American Political Science Review Vol. 98, No. 3

= − a1 x¯  would ensure compliance with the statute FIGURE 3. Bureaucratic Capacity and for any possible ω because even if ω =−, the real- Implementation Actions ization of the Bureaucrat’s action complies withx ¯.By contrast, by attempting action a2 = xB, the Bureaucrat f ()ω completely ignores the statute by attempting the action Ω small at his ideal point. This will often lead to a better policy outcome for the Bureaucrat, but it will also lead to substantial noncompliance. In this example, whenever ωx¯). Thus, the expected benefit for the low- ever risking noncompliance. Ifx ¯ is sufficiently far to capacity Bureaucrat of attempting greater compliance the left of xB (relative to the cost of noncompliance, with the statute (by moving from a1 to a2)ismuch 5 i.e.,x ¯ ≤ xB − ), the Bureaucrat prefers taking his less than this expected benefit for the higher-capacity most preferred action over taking any action that at- Bureaucrat. The policy loss (when noncompliance does tempts to comply with the statute. In general, for any not occur or is not detected) of moving from a1 to a2 is x¯ ∈/ [xB + ε − , xB + ε + ], the Bureaucrat will at- the same for the high-capacity and the low-capacity tempt an action at his ideal point. We can therefore see Bureaucrat, but the low-capacity Bureaucrat is less that as the bureaucratic capacity improves ( declines), willing to pay this cost because he reduces the risk of the range of statutes that result in a = xB + ε increases noncompliance much less than does the high-capacity (i.e., the interval [xB + ε − , xB + ε + ] shrinks). Bureaucrat. Thus, all else equal, the lower-capacity Bu- Although a high-capacity Bureaucrat will respond to reaucrat is harder for the Politician to control. Because a wider range of “extreme” statutes by attempting an the results of the low-capacity Bureaucrat’s actions are action at his ideal point, a higher-capacity bureaucrat is less sensitive to his efforts, he has less to lose from also easier to “control” with nonextreme statutes. That risking additional noncompliance and thus, for a given is, whenx ¯ ∈ [xB + ε − , xB + ε + ], a higher-capacity statute, will attempt actions that are closer to his pre- Bureaucrat can be induced to take actions that are ferred policy than will the high-capacity Bureaucrat. closer to the Politician’s preferred policy than can a To establish formally the precise location of the low-capacity Bureaucrat. The reason can be seen in Bureaucrat’s optimal action, recall that we assume =−∞ Figure 3, which depicts two different levels of bureau- xP < xB, x , and we multiply the quadratic policy cratic capacity, a higher-capacity Bureaucrat in the top utility by one-half. With these assumptions, the Bu- plot and a lower-capacity Bureaucrat in the bottom reaucrat’s expected utility of attempting to implement plot. the policy a in response to statutex ¯ is given by Given thex ¯ depicted in Figure 3, consider the con-  sequences for the Bureaucrat of moving his attempted 1 2 − (a − ω − ε − xB) f (ω) dω action from a1 to a2. For both levels of capacity, if the 2 − Bureaucrat attempts an action at a1 = x¯, the probability a x¯ a−x¯ 5 1 Note that this boundary point implies that the cost of punishment − (a − ω − ε − x )2 + γδ f (ω) dω for noncompliance is sufficiently small because we assume in this 2 B paper that δγ < 2 (see discussion below). −

485 Bureaucratic Capacity, Delegation, and Political Reform August 2004

The first term in this expression is the expected nonextreme statutes, the Bureaucrat’s best response utility over the range of ω for which a leads to a lies at the intersection of the marginal benefit curve compliant policy by the Bureaucrat. The second term with the relevant marginal cost curve. Givenx ¯ 1,for ∗ ∗ in this expression is the expected utility over the example, the optimal action is a1. For any a < a1,the range of ω’s that lead to noncompliant policies. Eval- marginal policy benefits of increasing the policy action uating the integrals reveals that the expected utility (toward the Bureaucrat’s most preferred) exceed the | = ∗ for the Bureaucrat from an action, a,isEUB(a x¯) marginal compliance costs, and for any a > a1,there- − 1 − − 2 − − − 2 duction in compliance costs of moving the action away 2 (a ε xB) γδF(a x¯) σω, where F(ω)isthe cumulative distribution function for ω. Thus, the Bu- from the Bureaucrat’s most preferred action exceed reaucrat’s optimal implementation strategy a∗ solves the policy losses. ∗ the first-order condition ∂EUB(a | x¯)/∂a = xB − a + ε − We can see from Figure 4 that the effect of changes γδf (a∗ − x¯) = 0. inx ¯ on the Bureaucrat’s best response depends on The first-order condition has a very intuitive inter- whether the apex of the “compliance cost” triangle is to pretation. It says that in equilibrium the marginal utility the left or right of the policy benefit line (i.e., to the left ∗ ∗ of increasing the intended action toward xB, xB − a + ε or right ofx ¯ 2 in Figure 4). Forx ¯ 1 < x¯ < x¯ 2, a > x¯.Inthis must equal the marginal increase in the likelihood of range, increases inx ¯ increase the marginal compliance punishment γ f (a∗ − x¯) times the penalty δ. Below we costs of any a > x¯, inducing the Bureaucrat to move ∗ ∗ refer to x − a + ε as the marginal policy benefit and a toward the Politician’s ideal point. Atx ¯ 2, however, B ∗ γδf (a − x¯) as the marginal compliance cost of action a. this effect reverses. Forx ¯ > x¯ 2, a < x¯, and increases in ∗ Thus, the key feature that determines how much the x¯ decrease the marginal compliance cost of any a < x¯, Bureaucrat is willing to choose actions far from his inducing the Bureaucrat to adjust his action closer to ideal point is f (a∗ − x¯), the value of the density of ω at his ideal point. a∗ − x¯. We focus in this paper on the case where 2 >γδ. Figure 4 plots the marginal policy benefit curve and This simplifies the analysis by ensuring that the the marginal compliance costs for three values ofx ¯ as marginal benefit and cost lines intersect only once and a function of a. The marginal benefit line is the thick guarantees second-order conditions for a maximum downward-sloping line. The marginal policy benefit is hold at that solution. Substantively, this implies that independent of the location ofx ¯ and declines as the we are focusing on cases where the variance in the Bureaucrat’s action approaches the Bureaucrat’s most implementation errors is relatively large compared to preferred action, xB + ε. The marginal cost curves de- the Bureaucrat’s expected costs of noncompliance. This pend on the location ofx ¯ and are depicted in Figure 4 restriction is reasonable given our focus on democ- by the three triangles centered atx ¯ 1,¯x2, andx ¯ 3. Outside racies where bureaucratic capacity is limited and the these triangles the marginal cost is zero. Ifx ¯ were too probability of the courts rigorously enforcing statutes high or too low, the marginal costs would be zero at the is relatively small. Bureaucrat’s ideal intended action, xB + ε. This would Proposition 1 specifies the best-response function for lead the Bureaucrat to choose his ideal action. For the Bureaucrat.

FIGURE 4. Effect of Statute on Intended Action

+ε x B

Marginal Policy Marginal policy benefit of increasing a Benefits xa−+ε and B Compliance Costs Marginal compliance cost of increasing a γδfa() − x

0 ε = * **= + ε x1 xa22 aa13 xB x3 Statutes and Intended Actions

486 American Political Science Review Vol. 98, No. 3

FIGURE 5. The Bureaucrat’s Best Response

x * ()− ε − ε ax γδ + Ω x −Ω x − xB xB B B Ω

xB

**( ,−Ε) + Ε **( , Ε )−Ε a x 2 ax2

γδ x − B Ω

= xP 0 − Ε * − Ε * x + Ε * + Ε −ε x1 x2 x1 x2 1 x2 x

Proposition 1. When 2 >γδ, the following intended Second, the Politician can induce any action ∗ actions by the Bureaucrat are optimal given x:¯ a − ε ∈ [xB − (δγ/), xB] by adopting intermediate values ofx ¯.For¯x − ε ∈ [x − , x − (δγ/)], the Bu- − ≤ − ∗ = + B B (i) If x¯ ε xB ,a xB ε. reaucrat’s equilibrium action moves farther from his − − ≤ − ∗ = 2 + (ii) If xB 

6 ∗ then the best action that the Politician can induce is For largex ¯, a − ε = xB is always compliant so the Bureaucrat will ∗ ∗ a − ε = xB − (δγ/). Because we wish to focus in this choose it. For lowx ¯, a − ε = xB is never compliant and marginal concessions have no impact on compliance, so the Bureaucrat is paper on cases where Bureaucratic capacity is low, we willing to choose it. will concentrate only on cases where the Politician

487 Bureaucratic Capacity, Delegation, and Political Reform August 2004 cannot induce his most preferred action (i.e., where THE POLITICIAN’S OPTIMAL DELEGATION xB > (δγ/)). STRATEGY Third, the Bureaucrat’s V-shaped best response is asymmetric, with a steeper downward slope in the first We now consider the Politician’s optimal statute. First, assume that the Politician has complete policy exper- interval than the upward slope in the second interval. = This asymmetry can also be seen in Figure 4, where tise, corresponding to the case where E 0. From Proposition 1, the best action the Politician can althoughx ¯ 1 andx ¯ 3 produce the same action,x ¯ 1 is much ∗ induce is a − ε = xB − (δγ/), which occurs when closer tox ¯ 2 than isx ¯ 3. To understand why the Bureau- − = − crat’s action is more responsive to changes inx ¯ for x¯ ε xB (δγ/). The more difficult case occurs − when the Politician does not know ε (which occurs low values ofx ¯ ε than for high values, note that the − effect on a∗ of a change inx ¯ can be decomposed into when E > 0). The Politician’s goal is to setx ¯ ε so that the Agent’s expected response is as close as possible two components. The first-order effect is that increas- − ingx ¯ changes the marginal compliance costs, giving to xB (δγ/), but policy uncertainty precludes the the Bureaucrat an incentive to adjust its action. The Politician from doing this effectively. second-order effect is that the change in the action Return to Figure 5 for an example of how the Politi- itself changes the marginal policy benefits and com- cian solves the problem. The Politician wishes to po- pliance costs, leading to an additional adjustment in sitionx ¯ so as to maximize his utility over the range ∗ 7 of possible ε. If the Politician choosesx ¯ 1, for exam- a . When these first- and second-order effects move − ∈ − + in the same direction, the total effect ofx ¯ is larger, lead- ple, thenx ¯ 1 ε [¯x1 E, x¯ 1 E], as depicted by the ing to policy choices that are more responsive to the two dashed vertical lines. The expected utility for the statutes. Politician fromx ¯ 1 is represented by the area that is If a∗ > x¯, the marginal compliance cost is decreasing bounded by the dashed lines on each side, the horizon- in a and increasing inx ¯. The first-order effect of increas- tal axis on the bottom, and the best response function ingx ¯ is therefore a decrease in a∗. The second-order on the top. The Politician wants to choosex ¯ to min- effect is that the decrease in a∗ also increases marginal imize this area (and, of course, would maximize her ∗ utility if the best response was a flat horizontal line at compliance costs further decreasing a . Because both ∗ − = effects work in the same direction, the intended ac- a (¯x ε) 0). tion is very responsive tox ¯ when a∗ > x¯.Fora∗ < x¯, We can see thatx ¯ 1 cannot be an equilibrium statute the marginal compliance cost is increasing in a and de- for the Politician. By moving fromx ¯ 1 tox ¯ 2, for ex- creasing inx ¯. Thus, the first-order effect of increasingx ¯ ample, the Politician eliminates the outcomes in the is an increase in a∗, whereas the second-order effect is shaded area at the left in the diagram and adds the increased compliance costs, which depresses a∗. Since outcomes in the shaded area at the right in the diagram. the first- and second-order effects move in opposite Because the shaded area at the left is larger than the directions, the Bureaucrat is less responsive to changes shaded area at the right (and the Politician wants to in the statute.8 minimize the area under the best-response function), As Proposition 1 and Figure 5 make clear, as the the Politician does better adoptingx ¯ 2 than adopting Bureaucrat’s ideal point moves away from the Politi- x¯ 1. cian’s, so too does the best action that the Politician In equilibrium, the Politician will choosex ¯ such that if can obtain. Consistent with many theories of delega- the Politician movesx ¯ in either direction, the expected tion, then, when we ignore the impact of bureaucratic utility of the policy losses created on the right side of capacity, the Politician in our model always prefers a the V in the Bureaucrat’s best response is exactly equal to that of the expected losses on the left side of this V. Bureaucrat with an ideal point closer to the Politician’s ∗ − = ∗ − + to one with an ideal point farther away. But if we take This occurs only when a (¯x, E) E a (¯x, E) E, as bureaucratic capacity into account, the Politician may withx ¯ 2 in Figure 5. not prefer a Bureaucrat who has preferences similar Note that if the policy uncertainty is so large that to the Politician’s. Note that the best action the Politi- E >, there is no unique optimal statute for the Politi- ∗ cian. In such cases of extreme uncertainty, the Politi- cian can induce, a − ε = xB − (δγ/), moves closer to the Politician’s ideal point as bureaucratic capacity im- cian’s decision problem in designing her statute is triv- proves ( gets smaller). Thus, when our assumption ial because all she can do is centerx ¯ so as to en- sure that the interval [xB − , xB + ] is spanned by that xB > (δγ/) is satisfied, in comparing any two − + Bureaucrats, there is always a level of bureaucratic [¯x E, x¯ E]. For this reason, in what follows, we fo- capacity that will make the Politician prefer the “less cus on the case where >E, an assumption that is also friendly” one to the “more friendly” one (in terms of consistent with our focus on polities with low-capacity preferences) because the high-capacity Bureaucrat is bureaucrats. easier to control. This intuition about the optimal statute is formal- ized in Proposition 2, which also states the Politician’s expected utility from this statute. √ 7 More technically inclined readers will recognize this intuition as Proposition 2. If > min{E, (δγ/xB), γδ}, then that underlying the implicit function theorem. 8 This asymmetry in the response function does not depend on our ∗ γδ γδ specification of the distribution of ω. It is true for any distribution x¯ = xB − + E function that is symmetric and unimodal at zero.  2

488 American Political Science Review Vol. 98, No. 3 and values of ε. Because politicians with divergent prefer- ences value this net increase in concessions the most, 1 γδ 2 γδ γδ they are the most willing to grant additional discretion EU (¯x∗) =− x − + x − E P 2 B  B  2 to bureaucrats. The mechanism that leads to a violation of the mon- itoring principle is similar. As ex post enforcement im- 1 γδ 2 + 2 + 2 proves, the Bureaucrat will make more concessions E σω 3 2 to the Politician’s preferences due to the increased cost of noncompliance. In particular, these increased Proof. See the Appendix. concessions are larger for low values of ε than for high values. By reducing discretion, the Politician can The results described in Proposition 2 can be used to transfer some of the gains for low-ε to high-ε states of explore two substantive issues. On one hand, since the the world. Thus, effective ex post enforcement creates ∗ Proposition describesx ¯ , it can be used to examine del- opportunities for the Politician to use policy details egation incentives when bureaucratic capacity is low. in statutes to induce greater responsiveness by a low- We explore this issue in the remainder of this section. capacity Bureaucrat. This turns the standard monitor- On the other hand, since it describes the Politician’s ing principle from studies of high-capacity bureaucrats expected utility from optimal behavior, the Proposition on its head. With high-capacity bureaucracies, policy can be used to examine the payoff to the Politician of details and ex post enforcement are substitutes for each particular types of political reform, such as improved other, but with low-capacity bureaucracies, they are ex post enforcement (which occurs when δγ increases). complements. We explore this issue in the next section. Given these findings, our empirical predictions about The contemporary literature on bureaucratic policy- delegation in developing systems are quite different making has developed a number of standard predic- from those of the advanced democracies for whom the tions about delegation to highly competent bureaucra- bulk of the empirical knowledge about delegation has cies, as follows. originated. Our model clearly suggests that findings consistent with the ally and monitoring principals are The uncertainty principle: Politicians should dele- gate more policymaking autonomy to bureaucrats contingent on a high bureaucratic capacity and that these principles should be applied to developing sys- when politicians are more uncertain about which tems with a great degree of caution. policy will yield the best outcome. The ally principle: Politicians should delegate more policymaking autonomy to bureaucrats when POLITICAL REFORM IN POLITIES WITH politicians and bureaucrats share similar policy LOW-CAPACITY BUREAUCRACIES preferences. The monitoring principle: Politicians should dele- Proposition 2 states that the Politician’s expected utility gate more policymaking autonomy to bureau- from adoptingx ¯ ∗ is crats when politicians have more opportunities for 2 ex post monitoring and sanctions. ∗ 1 γδ γδ γδ EUP(¯x ) =− xB − + xB − E Our model suggests that neither the ally nor the 2   2 monitoring principle should hold in polities with low- capacity bureaucrats. Note that an increase inx ¯ ∗ 1 γδ 2 + E2 + σ2 implies greater policymaking discretion for the Bu- 3 2 ω reaucrat in our model. Direct inspection ofx ¯ ∗ in Propo- sition 2 shows that the uncertainty principle holds in ∗ We can use the equation to examine how bureaucratic our model (i.e.,x ¯ is increasing in E). However, neither capacity influences the Politician’s willingness to pay the ally principle nor the monitoring principle holds ∗ for different types of institutional reforms. In the Ap- sincex ¯ is increasing x and decreasing in γδ. B pendix, we define these propositions formally and pro- The reason the allied principle fails is straightfor- vide proofs. In what follows, we state the Propositions ward. As we have demonstrated, low-capacity bureau- informally and discuss their intuitions. crats are make fewer concessions toward the ideal point Our first proposition concerns the interaction of pol- of the Politician because the impact of their policy icy expertise and bureaucratic capacity. choice on the probability of compliance is low. Thus, low capacity is especially harmful to the Politician when Proposition 3. When bureaucratic capacity is low, the Bureaucrat has divergent preferences. So how does r the Politician mitigate the harm and make the Bureau- the Politician’s benefit from increasing bureaucratic crat more responsive? One way is to increasex ¯.This r capacity increases as policy expertise increases, and leads to a higher marginal impact of policy choice on the Politician’s benefit from increasing policy ex- compliance for low values of ε and a lower impact for pertise decreases as bureaucratic capacity declines. high values. However, as we know from Figure 5, this leads to a sharper increase in concessions for lower Proof. See the Appendix for a formal statement of values of ε than the decrease in concessions for higher the proposition and its proof.

489 Bureaucratic Capacity, Delegation, and Political Reform August 2004

When politicians are relatively expert (E low), they model speaks to this issue in the following respect. The can optimally choose statutes that capture large con- preferences of bureaucrats are largely determined by cessions from the bureaucrat for most values of ε.For the rules and structures that politicians adopt. If the example, in the extreme case of E = 0, they can induce rules and procedures give politicians great latitude to ∗ the best possible action of a − ε = xB − (γδ/). Be- appoint, promote, and dismiss bureaucrats, then we cause these concessions get larger as capacity improves, might expect bureaucrats to have preferences that are highly informed politicians have strong incentives to relatively close to the politicians’. By contrast, if bu- increase capacity. reaucratic hiring and promotion are largely governed On the other hand, consider the incentives to ac- by neutral rules and procedures of , quire information when capacity is low. As we have then we might expect bureaucrats to have preferences shown, low-capacity bureaucrats tend to bias policy to- that are more divergent from the politicians (at least ward their ideal point and the expected policy outcome in comparison to bureaucrats who are chosen by a a∗ − ε tends to be less responsive to ε. This change in more politicized process). We can therefore ask how responsive of the expected policy outcome to the policy bureaucratic capacity influences the value of having a shock means that politicians are unable to generate “politicized” bureaucracy—–that is, of having a Bureau- substantially more concessions by learning ε.Thus,in crat with preferences aligned with the Politician. systems with low-capacity bureaucrats, politicians are Our main result is summarized in Proposition 5. least likely to want to develop policy expertise. Next consider reforms related to enforcement. In Proposition 5. When bureaucratic capacity is low, r our model, this enforcement, which is typically done the Politician’s benefit from decreasing policy diver- by courts but may also be done by politicians them- gence with the Bureaucrat increases as bureaucratic selves, is represented by δγ, the probability that bu- r capacity declines, and reaucratic actions that do not comply with a statute the Politician’s benefit from increasing bureaucratic are discovered, multiplied by the punishment cost for capacity declines as policy divergence with the Bu- noncompliance. A country with a that rigor- reaucrat decreases. ously enforces statutes will have a large value of δγ. We want to understand how reforms of bureaucracy Proof. See the Appendix for a formal statement of are related to the quality of the judiciary and how in- the proposition and its proof. centives to create strong are influenced by bureaucratic capacity. Our main result is summarized in Proposition 4. Proposition 5 describes another trap that is cre- ated by low bureaucratic capacity. When bureaucratic Proposition 4. When bureaucratic capacity is low, capacity is low, the benefit to the Politician of an r “allied” bureaucrat (with policy preferences similar the Politician’s benefit from increasing bureaucratic to the Politician’s) is greatest. This is because low- capacity increases as enforcement becomes more capacity bureaucrats bias policy choices so heavily to- r effective, and ward their own ideal points. Thus, one way to eliminate the Politician’s benefit from increasing the effec- those losses of political control is to try to politicize tiveness of enforcement decreases as bureaucratic the bureaucracy by moving it to the Politician’s ideal capacity declines. point. But consider how this logic affects the decision of a Proof. See the Appendix for a formal statement of politician to reform the capacity of an allied bureau- the proposition and its proof. cracy. Since a low-capacity ally produces policies close to the Politician’s ideal point, there is significantly less We have noted that the actions of a high-capacity Bu- reason to engage in capacity enhancing reforms. The reaucrat are more sensitive to the location of statutes. Politician’s benefit from reforming the bureaucracy The Politician can therefore best take advantage of a (decreasing ) declines when xB is close to xP. high-capacity Bureaucrat when strong ex post enforce- ment intensifies the relationship between actions and CONCLUSION punishment. But the Politician has the least to gain from improving ex post enforcement when bureau- In any polity governed by the , bureaucratic cratic capacity is low. One way to see this is to recall behavior during policy implementation is conditioned that the best policy the Politician can ever induce the by the bureaucrats’ fear that they could be punished if Bureaucrat to take is given by xB − (δγ/). Thus, the caught taking actions that the law forbids (or not taking Politician can push policy toward its ideal point by im- actions that the law requires). This fear is central to a proving enforcement (increasing δγ) or by increasing politician’s ability to use laws to successfully delegate bureaucratic capacity (decreasing ). But as the equa- policymaking authority to bureaucrats. Without it, bu- tion makes clear, the benefit of improving enforcement reaucrats could ignore the policy goals of politicians, will decline as  increases, and the benefit of improving thereby severing any democratic link between citizens bureaucratic capacity will decline as δγ decreases. and the policymaking process. Finally, consider how bureaucratic capacity influ- Factors that influence bureaucrats’ concern about ences incentives to “politicize” the bureaucracy. Our the consequences of noncompliance with legislation

490 American Political Science Review Vol. 98, No. 3 therefore have a significant influence on delegation and simultaneous efforts focusing on bureaucracies, courts, policymaking. Our analysis identifies one such factor— and politicians themselves. Although such comprehen- –low bureaucratic capacity—–that has received little siveness is a tall order, our argument suggests the con- attention in formal studies of delegation. Our model trary as well—–reforms that are not comprehensive are indicates that low bureaucratic capacity diminishes the likely to fail. Efforts to improve bureaucratic capacity ability of politicians to influence the actions of bureau- in polities with ineffective enforcement by judiciaries, crats. As bureaucratic capacity declines, bureaucrats for example, are likely to flounder and unravel. This recognize that their ability to take actions that comply need for comprehensiveness, we suspect, may help ex- with legislation also declines, diminishing their incen- plain the substantial difficulties polities face as they tive to try to do so. Politicians, then, are less able to attempt to develop politically. use legislation to influence bureaucratic actions when Because there has been little effort to model pol- bureaucratic capacity is low. icymaking in polities with low bureaucratic capacity, We argue that this diminished ability of politicians to there is clearly a need to develop additional theoret- control bureaucrats with low capacity has a significant ical models that can further explore the issues raised impact on how we understand delegation and polit- here. Bureaucratic capacity can be conceptualized in ical reform in developing democracies. Our analysis a number of ways, and our analysis focuses narrowly suggests, for example, that when bureaucratic capacity on only one of the possibilities. A particularly interest- is low, well-established principles in the existing dele- ing issue to introduce would be policy uncertainty by gation literature no longer operate. In particular, the bureaucrats themselves. In addition, our policymaking “monitoring principle” is turned on its head, as is the model is “-free,” and it would be useful to “ally principle.” consider how the constitutional setting, such as the sep- With respect to political reform, we argue that poli- aration of legislative and power, influences ties with low bureaucratic capacity are caught in a trap. delegation to low-capacity bureaucrats. And we have Although politicians always benefit from reforming obviously offered a rather reduced-form specification low-capacity bureaucracies, they will be most willing of the factors within bureaucracies that create low to pay for such reforms when politicians themselves capacity. Modeling these factors within bureaucracies have technical expertise, and when ex post enforce- explicitly should have a central place in future studies. ment is effective. But politicians will be least willing to The further development of such models should signif- improve their technical expertise or ex post enforce- icantly improve our understanding of challenges faced ment when bureaucratic capacity is low. In addition, in governing those polities that are in the process of when bureaucratic capacity is low, politicians have the political development. greatest incentives to politicize the bureaucracy, and this politicization diminishes incentives to improve bu- reaucratic capacity. APPENDIX The existence of this trap suggests that “compre- hensiveness” is crucial to successful political reform. Proposition 1. The following intended actions by the Bu- In studies of political development, scholars often un- reaucrat are optimal given x:¯ derline the importance of reforming courts to enhance ∗ (v) If x¯ − ε ≤ xB − ,a = xB + ε. enforcement, or reforming bureaucracies to enhance ∗ 2 (vi) If xB − 

491 Bureaucratic Capacity, Delegation, and Political Reform August 2004 strategy for the Bureaucrat to choose an action higher than Now consider the Politician’s optimization problem, which ∗ her ideal action, we must also verify that a ≤ xB + ε. This is is to choosex ¯ to maximize guaranteed byx ¯ −  ≥ xB + ε orx ¯ ≥ xB + ε + . E 1 > − − 2 − 2 Case 2. a x.¯ In this case, the first-order condi- (a(¯x,ε) ε) dε σω = ε + − γδ/2 + γδ − /2 4E tion becomes a xB ( ) [ (a x¯) ], so that − ∗ 2 2 E a = ( (xB + ε) − γδ[¯x + ])/( − γδ). Similarly to above, we must establish that a∗ > x¯. This inequality will hold when ∗ The first-order condition is xB + ε − (γδ/) > x¯. It can be verified that a < xB + ε if and E only ifx ¯ > xB + ε − . √ 1 ∂a ∗ − (a(¯x,ε) − ε) dε = 0 Proposition 2. If > min{E, (δγ/xB), γδ}, x¯ = xB − 2E ∂x¯ (γδ/) + (γδ/2)E and −E ∗ 2 Note that ifx ¯ + (δγ/) − xB > E, then [∂a (¯x,ε)/∂x¯] = [γδ/ ∗ 1 γδ γδ γδ (2 + γδ)] > 0forallε ∈ [−E, E], which guarantees that EUP(¯x ) =− xB − + xB − E 2   2 the first-order condition is always negative. Thus, we know that the optimal policy satisfies E − (δγ/) + xB > x¯. 1 γδ 2 + E2 + σ2 By a similar argument, we can show that optimality 2 ω 3  also requires −E + xB − (γδ/) < x¯. Combining these con- ditions with our first claim, the optimalx ¯ must sat- Proof. Note that the expected policy outcome given a isfy both −E + xB − (γδ/) < x¯ < −E + xB +  and E + xB − choice ofx ¯ is ε>x¯ − x −  x¯ xB B  B  2 / ∗  xB − γδ(¯x +  − ε) Integrating both terms we can rewrite this condition (1 4)E a (¯x,ε) − ε = [(a(¯x, E)−E)2 − (a(¯x, −E) + E)2] = 0. Because both squared 2 − γδ terms are positive, the only solution is a(¯x, E) − E = γδ a(¯x, −E) + E.9 Thus, we can solve forx ¯ from ifx ¯ +  − x >ε>x¯ − x + B B  2x + γδ(¯x −  + E) 2x − γδ(¯x +  − E) ∗ − = − − ≥ B = B a (¯x,ε) ε xB ifx ¯ xB  ε 2 + γδ 2 − γδ ≥ − + or if ε x¯ xB  ∗ 2 which leads tox ¯ = xB − (γδ/) + (γδ/ )E. It is easy to ∗ − + − ∗ − + > δγ/ check thatx ¯ satisfies both E xB (γδ/) < x¯ < E Note that because we assume that ( xB), it must be the + + − ∗ + − 2 case that x ≥ a∗(¯x,ε) − ε>0sothat xB  and E xB  B γδ. Finally, givenx ¯ ∗, we can compute the Politician’s expected utility at the optimum as γδ 2 ∗ − 2 ∈ − 2 (a (¯x,ε) ε) xB , xB  1 γδ 2 γδ γδ EU (¯x∗) =− x − + x − E P 2 B  B  2 Now we claim that since >E,x ¯ ∗ ∈ [x + E − , x − E + B B ]. Suppose thatx ¯ < x + E −  orx ¯ > x − E + .We B B 1 γδ 2 can show that the distribution of utilities generated by + E2 + σ2 these choices is stochastically dominated by a distribu- 3 2 ω tion generated by some other choice ofx ¯. First, con- − + − 2 − 2 siderx ¯ > xB E . This generates a utility of xB σω Proofs of Propositions 3––5. Each of Propositions 3––5 ∈ − − − − ∗ − 2 − 2 − 2 − 2 for ε [ E, x¯  xB]and (a (¯x,ε) ε) σω > xB σω relates to the interaction of  with one other substantive for ε ∈ [¯x −  − xB, E]. Now we compare this distribu- variable in our model. Proving the propositions therefore  tion of utilities to that ofx ¯ = xB − E + . Note that involves (a) examining the first-order partial derivatives of ∗  − = ∗ + −  − − +  a (¯x ,ε) ε a (¯x,ε x¯ x¯ ) ε x¯ x¯ . Thus, because ε EUP with respect to  and with respect to the substantive is distributed uniformly, the distribution of outcomes variable of interest and (b) demonstrating that the cross- for ε ∈ [−E, E − x¯ + x¯ ] underx ¯  is the same as for ε ∈ partial derivative of these two variables has the correct sign  [¯x −  − xB, E] underx ¯. However, for ε ∈ [E − x¯ + x¯ , E] un- (given the signs of the first order partials). In Proposition 3,  − ∗  − 2 − 2 − 2 − 2 derx ¯ , (a (¯x ,ε) ε) σω > xB σω, which is strictly bet- for example, we claim that (i) the benefits to the Politician − 2 − 2 ter than the xB σω for the same width interval under x¯. Thus, the distribution of utilities underx ¯ stochastically dominates the distribution underx ¯. We can show that any 9 The second-order condition is [−γδ2(a(¯x, E) − E)]/[E(2 − γδ) 2 2 x¯ < xB + E −  is similarly dominated. ( + γδ)], which is negative because a(¯x, E) − E > 0and >γδ.

492 American Political Science Review Vol. 98, No. 3 of increasing bureaucratic capacity increase as policy exper- Formal Statement tise increases, and (ii) the marginal benefits to the Politician √ { ˆ } of increasing policy expertise decrease as bureaucratic ca- For > min , (δγ/xB), δγ , ∂EUP/∂(δγ) > 0and 2 ˆ pacity declines. As we show below, (∂EUP/∂) < 0. Thus, ∂ EUP/∂(δγ)∂ < 0, where  is defined implicitly by 2 2 to show that the benefit to the Politician of increasing bu- (ˆ ) ≡ ˆ − E−(γδ/xB)(2 + (4E /3ˆ ) − (3E/ˆ )) = 0. reaucratic capacity increases as policy expertise increases, 2 we need to show that (∂ EUP/∂E∂) > 0. By a similar logic, Proof. Given Lemma 1, both parts of the proposition are true if, for sufficiently low bureaucratic capacity, given that (∂EUP/∂E) < 0, we again need to establish that  2 2 ≡ − (∂ EUP/∂E∂) > 0. For Propositions 3––5, we simply estab- ∂EUP/∂(δγ) < 0and∂ EUP/∂(δγ)∂ < 0. Note that  1 2 2 3 lish a sufficient condition on the size of  to ensure that the (γδ/xB)((3E/ˆ ) − (8E /3ˆ )) > 0 because (xB/γδ) > 1. ≡−  ˆ various partial derivatives have the correct sign. In so doing, Because (E) (γδ/3√xB) < 0and > 0, >E. Therefore,√ each of the three propositions takes advantage of Lemma 1. if > min{,ˆ (δγ/xB), δγ},then> min{E, (δγ/xB), δγ} √ so the politician’s expected utility is given by Proposition 2. Lemma 1. For > min{E, (δγ/xB), γδ}, (∂EUP/∂) < 0. Therefore, the marginal benefit of γδ is then ∂EUP/ ∂ δγ =  − / 2 − γδ/ γδ / 3 − / Proof. From proposition 2, ( ) [((2 E) 2 )(xB ( ))( E 2 ) (1 3)(E2γδ/4)]. For  ≥ Eand>(δγ/x ), this marginal B 2 utility must be positive. These conditions are also implied by ∂EUP γδ E γδ E(γδ) 4 E 1 =− 1 − xB − − 1 − −  >ˆ . ∂ 2   24 3  6 2 The cross partial of  and γδ is ∂ EUP/∂(δγ)∂ = 3 2 (1/ )[(E − )(xB − (γδ/)) + γδ + [(4E − 6)/3 ]γδE]. Note that the first term is clearly negative because >E. The 2 This cross partial is negative if  − E − (γδ/xB)(2 + (4E / sum of the second and third terms term is negative as long 2 2 6 3 ) − (3E/)) ≡ () > 0. By definition of ˆ and the fact as E(γδ) (4E − 3) < . Because >E, the left-hand side  that  > 0, this implies that the cross partial will be negative of this inequality can be no larger than 2(γδ)2. Because ˆ  2 >γδ, the inequality holds and the partial derivative is as long as >. negative.  Proposition 5 Proposition 3 Informal Statement Informal Statement When bureaucratic capacity is low, r When bureaucratic capacity is low, the Politician’s benefit from decreasing policy diver- r the benefit to the Politician of increasing bureaucratic gence with the Bureaucrat increases as bureaucratic ca- pacity declines, and r capacity increases as policy expertise increases, and r the benefit to the Politician of increasing policy expertise the Politician’s benefit from increasing bureaucratic ca- decreases as bureaucratic capacity declines. pacity declines as policy divergence with the Bureaucrat decreases.

Formal Statement √ Formal Statement { } √ If > min E, (3δγ/2xB), δγ ,(∂EUP/∂E) < 0and 2 2 If > min{E, (δγ/xB), δγ},(∂EUP/∂xB) < 0and(∂ EUP/ (∂ EUP/∂E∂) > 0. ∂xB∂) < 0. Proof. As noted above, given Lemma 1, both parts of Proof. Given Lemma 1, to prove both parts of the propo- the proposition are true if, for sufficiently low bureau- 2 2 sition we need to show that (∂EU√P/∂xB) < 0and(∂ EUP/ cratic capacity, (∂EUP/∂√E) < 0and(∂ EUP/∂E∂) > 0. Let ∗ ∗ ∂xB∂) < 0. If > min{E, (δγ/xB), δγ}, then the conditions  = min{E, (3δγ/2xB), δγ}.If ≥  , then the Politician’s for Proposition 2 hold, implying that (∂EUP/∂xB) = utilities are given by Proposition 2. Thus, 2 2 −(xB − (γδ/)) − E(γδ/2 ) < 0and(∂ EUP/∂xB∂) = − 2 + 3  ∂EU 1 γδ γδ 2 γδ (γδ/ ) E(γδ/ ) < 0 because E <. P =− x − + E < 0, ∂E 2 2 B  3 2 and REFERENCES ∂2EU γδ 3γδ 4 γδ P = x − + E > 0. ∂E∂ 3 B 2 3 2 Bawn, Kathleen. 1995. “Political Control versus Expertise: Congres- sional Choice About Administration Procedures.” American Po-  litical Science Review 89 (1): 62––73. Bendor, Jonathan, and Adam Meirowitz. 2004. “Spatial Models of Proposition 4 Delegation.” American Political Science Review. 98 (2): 293––310. Besley, Timothy, and John McLaren. 1993. “Taxes and Bribery: The Role of Wage Incentives.” Economic Journal 103: 119––41. Informal Statement Carpenter, Daniel. 2001. The Forging of Bureaucratic Capacity: Rep- utations, Networks, and Policy Innovation in Executive Agencies, When bureaucratic capacity is low, 1862––1928. Princeton, NJ: Princeton University Press. r Chang, Kelly,David E. Lewis, and Nolan McCarty.2000. “The Tenure the benefit of increasing bureaucratic capacity increases of Political Appointees.” Paper presented at the 2000 American r as enforcement becomes more effective, and Political Science Association, Washington, DC. the benefit of increasing effectiveness of enforcement Epstein, David, and Sharyn O’Halloran. 1999. Delegating Powers. decreases as bureaucratic capacity declines. New York: Cambridge University Press.

493 Bureaucratic Capacity, Delegation, and Political Reform August 2004

Evans, Peter, and James Rauch. 1999. “Bureaucracy and Growth: A Huber, John D., and Charles Shipan. 2002. Deliberate Discretion? Cross-National Analysis of the Effects of ‘Weberian’ State Struc- Institutional Foundations of Bureaucratic Autonomy. New York: tures on Economic Growth.” American Sociological Review 64: Cambridge University Press. 748––65. McCourt, Willy, and Nazar Sola. 1999. “Using Training to Pro- Geddes, Barbara. 1990. “Building ‘State’ Autonomy in Brazil, 1930–– mote Civil Service Reform: A Tanzanian Local Government Case 64.” Comparative Politics 22 (2): 217––35. Study.” Public Administration and Development 19: 63––75. Geddes, Barbara. 1994. Politician’s Dilemma: Building State Ca- McCubbins, Mathew D., Roger G. Noll, and Barry R. Weingast. 1987. pacity in Latin America. Berkeley: University of California “Administrative Procedures as Instruments of Political Control.” Press. Journal of Law, Economics, and Organization 3 (2): 243––77.

494