Political Tenure, Power and the Size of Government Bureaucracy

Political Tenure, Power and the Size of Government Bureaucracy Marina Dodlova Very Preliminary Abstract This paper develops a model of a scope of bureaucratic discretion under different institutional structures of government. We contrast the effect of bureaucratic power over policy decisions in a political persistent system with that in a political instable one as well as in a separation of powers system with that of a parliamentary one. By bureaucratic power we mean rubber-stamping by politicians of their bureaucrats policy propositions without effective control even when politicians could not verify the payoffs of these projects ex ante. We demonstrate that this type of relationships is likely to develop in more political persistent systems with a minimal number of veto players: politicians stay in power for a longer time and have sufficient instruments to punish bureaucrats for an undesirable behavior. Therefore, in such systems politicians are tempted to delegate less responsibility in a credible way to their bureaucratic agents, but instead they are willing to grant more discretion in an informal way. We focus on two effects. On the one hand, the more politicians stay in office, the less decision-making power they delegate formally to bureaucrats. On the other hand, they rely on bureaucrats supporting them, so that they are likely to increase the size of bureaucracy. Using the data on government administration employment from Schiavo-Campo et al. (1997) and the database on political institutions from Beck et al. (2006) we find motivating evidence that systems with less veto players have a smaller number of bureaucrats, while more political persistent regimes tend to have bigger governments in terms of administration employment. JEL: D72, D73, D78, H00 Key words: Political tenure, veto player, bureaucracy, delegation, informal authority, political regime, government administration employment Université Paris Ouest - Nanterre la Défense, CNRS. EconomiX, Bureau 112, Batiment K: Maison Max Weber, 200 avenue de la Republique, F-92001 Nanterre Cedex, France. E-mail: mdodlova@gmail.com 1

1 Introduction The expertise advantage of the bureaucratic agency is a corner stone in political models of delegation. Typically, the agency has or can acquire for some costs perfect knowledge of the connection between policy choices and policy outcomes, the legislature is uninformed on this count, and the models go on to show why it is beneficial for the legislature to delegate discretion in order to take advantage of what the agency knows. Formal models have repeatedly shown that legislative control is quite imperfect and that agencies often have substantial autonomy. In a democracy political superiors are elected; and this means that, bureaucrats getting organized to take a collective action in politics (e.g. through public sector unions) can exert political power over their own superiors. Bureaucrats can influence on election outcomes and participate in choosing the policy course (Moe, 2006). Furthermore, very often political superiors in seeking loyalty tend to favor bureaucratic interests: promote job security, enhance autonomy, protect established programs. The normal way of thinking about bureaucrats, however, assumes that they have no political power at all. But bureaucrats are not only experts who have the information leverage over their political superiors but also successful lobbyists exploiting the privileges of the institutional structure of democratic systems (Bennedsen and Feldmann, 2004). Moreover, there are known extreme cases when politicians informally act as the agents of the top bureaucrats and the latter exert real control over policy decisions. These aspects of delegation in political control models have been never seriously considered. As Moe (2006) stresses this important part of the delegation story is missed, and theorists tend to underestimate what bureaucrats can do to get their way. In this paper, we present a model of political control where a bureaucrat may exert political power over a politician. A politician is limited in acquiring information about policy outcomes and he cannot learn a share of information (even for some costs), which, however, can be communicated to him by his bureaucratic agent. This paper contributes to the main trade-off of delegation in political agency models - a trade-off between expertise and political control but we focus on several points, which have not been seriously addressed in previous studies all together. The first one is political power of bureaucrats that might be revealed through mechanisms of rubber-stamping and informal delegation of legislative proposals by politicians. We consider the model of different forms of delegation in the relationship between the politicians and their bureaucratic agent. We are interested in the process of public policymaking in which political representatives who are mainly uninformed about the repercussions of policy choices appoint bureaucrats to gather information necessary to make informed decisions. In some cases, the agent is given formal authority to decide on policy (e.g. central banks), and in other cases, the agent merely makes recommendations or give advice on policy (e.g. investigatory commission). These are models on communi- 2

cation versus delegation with uninformed principals and informed agents (Dessein, 2002; Ludema and Olofsgard, 2008). We argue that there are other cases when politicians refrain from credible delegation but transfer their authority to take policy decisions in an informal way. We follow the literature and distinguish between simple rubber-stamping of the agent s proposition (Aghion and Tirole (1997)) and informal delegation (Baker et al. (1999)) where the latter implies that the boss can promise to ratify all the agent s proposed projects regardless of whether a given project yields high or low benefits to the boss. Following to Baker et al. (1999) we contrast informal delegation with informal authority, when it is the bureaucratic agent who is tempted to renege. Our analysis is also related to the concept of partial or transferable control by Aghion et al. (2002, 2004) where the principal can transfer control to the agent but cannot commit herself to do it. Second, we assume that actors look forward and there is political uncertainty. So when the legislature makes its delegation decisions it does not ignore the fact that the president or the major coalition may soon be out of office, replaced by another person or coalition. In this case under political uncertainty the logic of delegation differs sharply. Moreover, unconditionally transferable control and informal delegation also imply that agents are forward-looking because such relationships are based on future returns and are considered as the way of building trust and inducing cooperation in future. Nevertheless, the main intuition behind the model with forward-looking actors is that under political uncertainty today s enacting coalition wants to ensure that tomorrow s legislature cannot control the bureaucracy so that it grants more discretion to bureaucratic agencies. On the other hand, it understands that today s delegation of discretion will be difficult to reverse tomorrow prompting to choose less discretion today. These contrary effects are taken into account in the model. Therefore, we stress two key determinants for evolving relational contracting agreements between elected and appointed officials: the period of staying in office and the capability to punish bureaucrats. Third, the literature pays more and more attention to the connection of political institutions and the delegation decisions as different political systems tend to have their own distinctive structures and vary in incentives that shape control relationships. We emphasize that the institutional structure of government should be take into account in analyzing the way of delegation and a scope of bureaucratic discretion. Epstein and O Halloran (1999) and Huber and Shipan (2002) are the keystones of the literature on political control of the public bureaucracy. Our analysis is related to the standard model of delegation developed by Epstein and O Halloran (1999) and adapted by Huber and Shipan (2002) for the comparative context of legislative delegation and bureaucratic discretion under different political regimes. But they do not considers such important factors as policy uncertainty and forward-looking rationality that are critical for understanding of why delegation differs across presidential and parliamentary systems. We follow to Horn (1995) in recognizing the importance of forward-looking rationality and political uncertainty but he 3

does not pay attention to the difference between presidential and parliamentary regimes. So our model uses the logic of Moe and Caldwell (1994) of comparing the delegation with forward-looking actors in different political systems. But in contrast to them we allow for the political power of bureaucrats and the incapacity of politicians to learn a share of information about the connections between policy choices and outcomes. We stress that agency problems differ in various institutional bargaining environments. This paper contrasts the effect of bureaucratic political power over political superiors in a separation of powers system with those of a parliamentary system and compares the way of delegation and the scope of bureaucratic discretion in different political regimes. We argue that there are two main features of political regimes that affect the gravity of agency problems within government bureaucracy: first, institutional bargaining environment which defines the possibility of politicians to punish bureaucrats for an undesirable behavior, and second, the period during which politicians stay in office because it is important for the share of decision making power they delegate to bureaucrats. Thus, these two characteristics of political regimes are crucial for shaping the structure and describing the features of government bureaucracies. This paper explains why these political institutions are important for the structure and size of bureaucracies through the mechanism of enlarging informal authority of bureaucrats. We argue that in political agency the information scarcity of political principals lead to the variety of ways to delegate control over policy decisions to their bureaucratic agents, and it is not necessarily contractible delegation. Indeed, politicians might rubber-stamp their bureaucrats policy propositions without effective control, even when politicians could not verify the payoffs of these projects ex ante. Thus, some relational contracting agreements could evolve, and this depends on characteristics of political regimes aforementioned. We demonstrate that this type of relationships is likely to develop in more political persistent systems with a minimal number of veto players: politicians stay in power for a longer time and have sufficient instruments to punish bureaucrats for an undesirable behavior. Therefore, in presidential systems with minimum veto players politicians are tempted to delegate less responsibility in a credible way to their bureaucratic agents, but instead they are willing to grant more discretion in an informal way. Thus, we contradict the conclusion of Huber and Shipan (2002) but confirm the result of Moe and Caldwell (1994) who compare the delegation in the United States and Britain. If politicians prove to be engaged in some relational contract agreements with their bureaucratic agents, they do not delegate the formal decision-making power but rely on bureaucrats informally thus enlarging their informal authority as defined in Baker et al. (1999). We focus on two effects concerning political agency. On the one hand, the more politicians stay in office, the less decision-making power they delegate formally to bureaucrats. On the other hand, they rely on bureaucrats supporting them, so that they are likely to increase the size of bureaucracy. Using the data on government administration 4

employment from Schiavo-Campo et al. (1997) and the database on political institutions from Beck et al. (2006) we find evidence that presidential systems with less veto players have a smaller number of bureaucrats, while more political persistent regimes tend to have bigger governments in terms of administration employment. The paper is organized as follows. Section 2 introduces general set up. Then we present the results of the model. Section 3 describes our motivating evidence for our theoretical predictions. The last section concludes and suggests some directions for further research. 2 Model 2.1 General set up Following Aghion and Tirole (1997), Baker et al. (1999) and Tirole (1999) we consider the rubber-stamping, informal delegation and informal authority phenomena in various institutional bargaining environments, like in presidential and parliamentary regimes. We study the principal-bureaucrat-agent hierarchy (Guriev, 2004) where politicians are accountable to citizens and bureaucrats are accountable to politicians. The bureaucracy helps the politician to take policy decisions by acquiring information about different legislative projects and policy outcomes (e.g., Egorov, Guriev, and Sonin, 2008) so that finally the policy course is shaped by outcomes of approved projects. The main intuition of the model is to demonstrate how the information asymmetry under various bargaining environment defines the scope of bureaucratic discretion and the way politicians delegate some authority to their bureaucratic agents. Policy choices are made by the three incumbent legislators in legislative bargaining. So we study a repeated game involving three politicians who are engaged in different authority relationships subject to the political regime. Following Tirole (2009) we distinguish between collegial authority and veto collegiality. The first one requires that both parties agree to depart from the status quo in order for an alternative policy to be implemented. It thus imposes a strong status quo bias. We refer this case to parliamentary regime. The other regime implies veto collegiality under which one party has authority, except that the other party has the right to impose the status quo if he does not like the other party s choice. Thus, we assume that in the parliamentary regime three politicians have collegial authority between themselves, while in the presidential regime the authority between two politicians is allocated collegially and the third politician has the right to veto. In this version we do not consider elections as they are but we suppose that the repeated relations between politicians occur until they are in office, so that these relations are finite and could be broken if one of them leaves. The bureaucracy helps the politicians to take policy decisions by screening policy projects and acquiring information about their outcomes. Thus, the bureaucracy is the agent of political principals and we consider it as 5

a single whole that is an administrative machine working for the elected officials. Each politician has the formal authority over the bureaucracy s propositions. Figure 1: Government Hierarchy The politicians and bureaucrats can implement a status quo project 0 that yields known profit B0 i > 0 to politician i together with a private benefit to b 0 to the bureaucracy. This status quo may be interpreted either as doing nothing (the parties decide not to interact, in which case B0 i = b 0 = 0 is a natural assumption), or as pursuing current policy or renewing last year s budget, that is, as the absence of policy innovation. The agent s private benefit can be thought of as a perk or as (minus) the disutility attached to implementing the project. The bureaucracy screens among n alternative projects k = 1...n. Project k yields verifiable profit Bk i together with private benefit b k to the agent. While these profits and private benefits differ among projects, alternative projects all look ex ante identical. However, it is known that among all the projects there are the politicians preferred projects, which if it is chosen, yields the benefit B i > 0 for them, i = 1, 2, 3. Similarly, there is the agent s preferred project, which if it is chosen, yields for the agent the private gain of b > 0. Thus, each project is associated with four different benefits for three politicians and the bureaucracy, i.e. (X, Y, Z, θ). The idea is that if the first politician s preferred project is chosen then X = B 1 and the others benefits can be any with different probabilities. If the bureaucracy s preferred project is approved then θ = B 1. For simplicity, let us assume that there are some congruence in government actors preferences and there are 16 projects which give the positive benefits for government actors with different congruent parameters, while the others yield the negative payoff for them. Therefore, we do not consider the latter below. All these gains and benefits are a common knowledge. For simplicity, let us assume that the agent is very risk averse and therefore does not respond to monetary incentives. He then receives a constant wage, normalized at zero, and the principal receives the profit. The agent has reservation utility 0. Suppose that the nature of projects is unknown to the politicians and bureaucracy. The agent at private cost g A (e) perfectly learns the payoffs of all candidate projects with 6

probability e. With probability 1 e the agent learns nothing and still look at the projects as identical. Similarly, the principals choose how much time or effort to expend for learning payoffs. At private cost g Pi (E i ), i = 1, 2, 3, the politicians become perfectly informed about the payoffs with probability E i and learn nothing with probability 1 E i. We focus on the simultaneous case of acquiring information by principals and agents. The function g k ( ) is increasing and strictly convex and satisfy g k (0) = 0, g k (0) = 0, g k (1) =, where k = A, P 1, P 2, P 3. Thus, for example, the project with higher benefits for all government actors is approved with the probability ee 1 E 2 E 3. For others there are combinations of probabilities originating from the extent of awareness 1. In the simple case one can assume that the politicians can learn the payoffs of projects with the same probability E. Below we proceed from just this proposition as well as from that the benefit for all politicians are defined exogenously in spite of their common awareness with probability E. In practice this refers to political uncertainty when parties are informed with E but the policy projects could finally be different for all politicians. In the model we assume that information is soft, that is it cannot be verified by the other party. specific project. The communication is thus interpreted as a pure suggestion to choose a The following four possibilities should be considered: Both parties are informed (probability ee): each recommends his preferred project. Only the bureaucracy is informed (probability e(1 E)). Only the politicians are informed (probability E(1 e)). No one is informed (probability (1 e)(1 E)): the status quo is implemented with probability 1. To simplify we consider the following case when each project has four dimensions X, Y, Z, θ but each dimension has only two values, high - B H or b H and low - B L or b L. Then the list of all possible projects can be listed as it has been done in Appendix A. We assume that 2B H + B L > 3B 0, B H + B L > 2B 0, B H + 2B L > 3B 0 and B L < B 0 < B H so that some bargaining will be possible in different political regimes. Then, the inequality B H + 2B L > 3B 0 allows to capture the difference between the parliamentary regime and presidential one. By this example we demonstrate which projects will be approved and on the basis of which mechanisms subject to the bargaining environment and the type of information 1 These probabilities can be defined as the matrix 16 16. It would be interesting to consider different situations by simulations. 7

asymmetry. At this moment we abstract from endogenous search intensity and so from the choice of the type of information structure. Below we assume that it is the first politician who has the right to veto. Case 1: Both parties are informed 2 In this case the politicians are informed so they will choose only those projects that yield high benefit for them. In the parliamentary regime all projects which give 2B H + B L > 3B 0 and B H + 2B L > 3B 0 are approved subject to the high private benefit for the bureaucracy. These are projects by number 1,2,3,5,9,10,11. In the presidential regime only projects 1,2,3 will be chosen as the first politician has the right to veto and he chooses only those projects which yield no less than B 0. Thus, in the parliamentary case the status quo is sustained with the less probability than in the presidential regime 3. The mechanism of informal delegation might be launched in this case if the politicians promise to ratify all projects proposed by the bureaucracy even if they are not in the best interest of the politicians. This induces superior effort from the agent (Baker et al. 1999) and these benefits from increased effort can outweigh the expected cost of approved poor projects with low benefit for politicians. Under informal delegation the bureaucracy will propose the projects with high benefit for him that is b H and in the parliamentary regime politicians are willing ratify all these proposed projects so that projects by number 1,2,3,5,9,10,11,13 might be approved. However, in order to sustain this relationship the bureaucracy will propose only projects which give high benefit for the politicians. In the presidential regime under informal delegation projects by number 1,2,3,5 can be chosen so that under informal delegation the bureaucratic discretion is higher. Case 2: Only the bureaucracy is informed In the case of the only bureaucracy informed there are two cases possible: the politicians can just rubber stamp all proposed projects or veto them. If the expected payoff is higher than 3B 0 the politicians rubber-stamp the legislative project, otherwise they veto the proposed projects. In both presidential and parliamentary regimes the same projects by number 1,2,3,5,9,10,11,13 will be ratified, because the presidential cannot distinguish the poor project for him. However, he approves them if and only if his expected payoff is higher than B 0. According Baker et al. (1999) there may exist an informal agreement that could dominate both vetoing and rubber-stamping: the bureaucracy could be granted informal authority to recommend only projects yielding B H, and the politicians could ratify all proposed projects but threaten to retract the bureaucracy s future authority if B L is realized. Since the politicians benefit B is not contractible (and is not even observable to the politicians until after implementation), this informal agreement must be enforced 2 Appendix A summarizes all cases. 3 1 w 1 w 2 w 3 w 5 w 9 w 10 w 11 < 1 w 1 w 2 w 3 8

through rent sharing in a long-run relationship. That is, the bureaucracy will receive an efficiency wage payoff so long as he does not abuse his informal authority (i.e., as long as he recommends only B H projects), and the agreement will be terminated if he abuses his authority. Following termination, the bureaucracy and politicians will interact through one-shot games characterized by either veto or rubber stamp, depending on the sign of the expected payoff of the politicians subject to high benefit for the bureaucracy E(B b H ). Case 3: Only the politicians are informed The bureaucracy looks at projects as identical so it proposes all projects, whereas politicians could see which of them yield benefit greater than B 0. Thus, in the parliamentary regime almost all projects will be approved - 1,2,3,4,5,6,7,8,9,10,11,12,14,15. In the presidential regime, when there is no bargaining between three of them only six projects are chosen - 1,2,3,4,6,7. In case when politicians are informed and the bureaucracy is uninformed there exist another interesting mechanism which we call informal concern. It consists in that the informed principal (in our case politicians) ratifies only those projects which give high benefit for the bureaucracy in order to increase their search intensity in future. Thus, we add Baker et al. s (1999) classification of informal phenomena by this additional informal agreement. However, if B H + 2B L < B 0, then the other range of projects will be approved. Appendix A lists these projects in both cases. In analyzing these cases we assume that E is exogenous but in fact this is endogenous and further all probabilities of projects are conditional and defined by the search intensity of the bureaucracy (e) and politicians (E). Moreover, we consider the same search intensity for all politicians but in general it could be different E i, where i = 1, 2, 3 as if we mentioned in general set up. 2.2 The choice of information structure and the type of relational agreement This section approaches to our findings in a closer way, although in order to demonstrate a number of simple results we introduce some simplifying assumptions. We need to abstract from the complex probability structures so that let us assume the following. Let all politicians have the same high and low benefit, B H and B L, respectively. They also choose the same search intensity E, that is E P1 = E P2 = E P3 = E. Then, the probability of discovering payoffs of projects is equal to E. And then bargaining occurs between politicians according to the authority structure in a regime. Let us assume that B H + 2B L < B 0, then the projects listed in case 2 of Appendix A will be chosen. We can compute the utilities of government actors in the basis of the 9

approved policy projects. Let B H = B L and B 0 = 0, then these utilities are presented in the last column. Now we can see our first result that in the parliamentary regime the range of approved projects is greater than in the presidential regime so that we can conclude that the discretion is greater in the parliamentary regime. We get the following utilities for four cases: in the parliamentary regime: contractible case: 6B H E(2 e) and 4b H e(2 E) relational case: 6B H (e + E 2eE) and 4b H (e + E) in the presidential regime: contractible case: 5B H E(2 e) and b H e(8 5E) relational case: B H (5e + 5E 6eE) and b H (3e + 3E 2eE) The conditions when the politicians and the bureaucracy choose to be engaged in relational agreements is traditionally defined and can be written as in the parliamentary regime: B L + δ6b H (e + E 2eE) δ(6b H E(2 e)) and 4b H (e + E) 4b H e(2 E). These inequalities can be rewritten as B L + δ6b H (e E ee) 0 and E e + ee 0. in the presidential regime: B L + δ5b H E(2 e) δb H (5e + 5E 6eE) and b H (3e + 3E 2eE) b H (8e 5eE). These inequalities can be rewritten as B L + δ5b H (E e + 1/5eE) 0 and 3E + 3eE 5e 0. where δ is a discount factor depending from the number of reelection periods n, that is δ = 1 q 1 q n. Hence, we get in the parliamentary regime: E e 1 6δ 1+e and E e 1+e. in the presidential regime: E e+ 1 5δ 1+1/5e and E 5e 3(1+e). It follows that in the parliamentary regime informal agreement in not feasible, while in the presidential regime it is feasible. Moreover, in the presidential regime the search intensity of the politicians is higher than a certain threshold so that the search intensity of the bureaucracy, on the contrary, is less than a certain threshold. The main intuition behind this is that the higher search intensity from the bureaucracy requires the higher level of administration employment, that is the bureaucracy enlarges. The other interesting result concerns the number of reelection periods. Indeed, we get that our thresholds depend on the discount factor. One can see that in the presidential 10

regime the more the number of reelection periods, this threshold decreases. This lead to the less search intensity of the politicians. Thus, the longer the politicians stay in office, the less their search intensity that induces the higher search intensity of the bureaucracy so that the bureaucracy enlarges. 3 Motivating Evidence We focus on two main predictions from our model. First, that in presidential regimes governments are smaller in terms of administration employment. Second, the longer politicians stay in office, the higher is the level of government employment. 3.1 Data For measuring government administration employment and wage we use the data from the paper of Schiavo-Campo, de Tommaso and Mukherjee (1997) on government labor and wage. They extract these data through statistical yearbooks, yearly budget documents, personnel ministries or agencies, inquiries to embassy personnel. Government administration includes executive and legislative administration by departments directly dependent on the Head of State or the Parliament, together with all other ministries and administrative departments. Consequently, government administration in their definition is general government less teaching and health personnel (p. 47). The armed force is considered separately, so that the data we use concerns only the civilian government employment. We consider separately the central government employment which is by definition of Schiavo- Campo et al. a part of government administration which includes all employees paid by the central government budget. From those countries for which the data on government administration employment is available, we select 50 countries that have a degree of democracy no less than 4 according to Jaggers and Marshall s (2000) Polity score. To depict the characteristics of political regimes along with the Polity IV data from Jaggers and Marshall (2000) we use the database on political institutions (DPI) from Beck et al. (2006). The latter one is more detailed in describing political institutions. It covers 177 countries over 21 years, 1975 1995. We extract the data from the beginning, that is from 1975, and analyse how the changes in political institutions for this period have influenced on the size of government in terms of administration employment when the latter is fixed for the beginning of 90th years. We pay attention to the following indicators of political regimes. The first one is the type of political regime, presidential or parliamentary system, as well as the number of veto players. The DPI contains variables that help to capture the institutional features of various bargaining environment of these systems. This fundamental characteristics of 11

a political system is described by the variable SYSTEM that captures whether countries are presidential, assembly-elected presidential, or parliamentary. Second, we take the variables that describe the impact of leader horizon. These are most important for us executive turnover measures. However, in systems in which the prime minister or president is not the only veto player, these turnover measures present a biased picture of the horizons of all veto players. In particular, in countries with multiple veto players, the replacement of a president, prime minister, or even a cabinet, does not always mean that all veto players from the previous government have changed. Often, some of the parties that exercised a veto role in earlier governments reappear in successor coalitions. The DPI allows to control for more nuanced turnover indicators. Nevertheless, to capture a general tendency we take the variable TENSYS, which indicates how long has the country had not consolidated democratic institutions and leadership was personalitybased, in other words, how long has the country been autocratic or democratic. Another important indicators we use are YRSOFFC and PRTYIN, which refer to the executive s years in office and the executive s party in office, respectively. We use the log of gross national income per capita based on the purchasing power parity to take into account that richer countries are characterized by bigger governments. We also include the log of total population to control for two effects considered in the literature. Along with the direct effect of larger bureaucracies in more populous countries the effect of economies of scale in public administration should be taken into account. As Alesina and Wacziarg (1998) argue economies of scale in supplying public goods lead to smaller government in larger countries. 3.2 Results Appendix B contains the data description statistics and country coverage. In Appendix C we present the results of simple regression analysis on the basis of ordinary least squares estimation. There are three specifications of models without controls and with the log of total population as well as with the gross national income per capita to take into account the effect of country richness and economic development. Table 2 contains the results of these models. R 2 and Adj. R 2 are presented to indicate the good fitness of models. According to OLS estimation TENSYS has a coefficient of 0.0193 which is significant at the 0.1% level and explains about 42% of the variation of total administration employment. In case of cental administration employment the variable of long-term leadership is also significant and equals 0.0153. As we can see the longer period of an autocratic regime with personal leadership has a positive effect and is strongly significant in all specifications. This confirms our first theoretical prediction: if a country experienced a long-term period of personal leadership, its government are characterized by the higher level of administration employment. 12

Including SYSTEM in the models confirms our other conclusion about the effect of the type of political regimes on the size of bureaucracy. Presidential regimes are associated with the lower level of government employment both in case of total administration and central administration. One can observe the improvement of models by about 5%-6%. The log of population size is negative and significant. Thus, we confirm the finding of Alesina nd Wacziarg (1998) of the effect of economies of scale. Further, including total population to the model considerably improves the good fitness of the model. The coefficient of income is positive so that richer countries have bigger governments. Only a weak bias in estimates of SYSTEM and TENSYS is observed in the model with income. It is interesting to note that the income variable is significant only in case of total government employment and its inclusion does not substantially improve the OLS model. It is interesting to capture the effects of other important indicators such as YRSOFFC and PRTYIN, which refer to the current executive s years in office and the executive s party in office, respectively. We do not present these results in appendix but we should report that the executive s years in office is more significant (at 10%) in the case of total government employment and non significant in the case of central administration. On the contrary, the executive party s years in office is significant only for central administration employment. 4 Concluding Remarks This paper studies the issues of informal delegation and relational agreements in political agency. We argue that different institutional bargaining environment in the presidential and parliamentary regimes defines the structure of agency problems between the politicians and their bureaucratic agents. Information asymmetry and policy uncertainty are also crucial for the type of relationship between government actors. Indeed, there is little literature, which ties the behavior and incentives of elected officials with agency problems within the government bureaucracy as well as there is little known about the informal authority of bureaucrats over politicians in policy decision making. We shed light on the scope of real effective control in the political agency under different institutional bargaining environment. Furthermore, our work contributes to the literature which emphasizes the features of government bureaucracies as crucial for economic performance. 13

References [1] Aghion, Philippe and Jean Tirole (1997) Formal and Real Authority // Journal of Political Economy. Vol. 105, No. 1. P. 1-29. [2] Aghion, Philippe, Mathias Dewatripont and Patrick Rey (2002) On Partial Contracting // European Economic Review. Vol. 46. No. 4-5. P. 745-753. [3] Aghion, Philippe, Mathias Dewatripont and Patrick Rey (2004) Transferable Control // Journal of the European Economic Association. Vol. 2. No. 1. P. 115-138. [4] Alesina, Alberto and Guido Tabellini (2007) Bureaucrats or Politicians? Part I: A Single Policy Task // American Economic Review. Vol. 97. No. 1. P. 169-179. [5] Alesina, Alberto and Guido Tabellini (2008) Bureaucrats or Politicians? Part II: Multiple policy tasks // Journal of Public Economics. Vol. 92. No. 3-4. P. 426-447. [6] Baker, George, Robert Gibbons and Kevin Murphy (1999) Informal Authority in Organizations // Journal of Law, Economics and Organization. Vol. 15. No. 1. P. 56-73. [7] Beck Thorsten, George Clarke, Alberto Groff, Philip Keefer, and Patrick Walsh (2004) New Tools and New Tests in Comparative Political Economy: The Database of Political Institutions // World Bank Policy Research Working Paper No. 2283. [8] Bennedsen, Morten and Sven Feldmann (2004) Lobbying Bureaucrats. Delegation and Influence Under Alternative Political Structures // Discussion Paper CEBR 2004-13, Center for Economic and Business Research. [9] Boadway, Robin and Motohiro Sato (2006) Bureaucratic Advice and Political Governance. Working Paper No. 1070, Queen s University. [10] Dessein, Wouter (2002) Authority and Communication in Organizations // Review of Economic Studies. Vol. 69. No. 4. P. 811-838. [11] Epstein, David, and Sharyn O Halloran (1999) Delegating Powers. New York: Cambridge University Press. [12] Horn, Murray J. (1995) The Political Economy of Public Administration. New York: Cambridge University Press. [13] Huber, John D., and Charles R. Shipan (2002) Deliberate Discretion? New York: Cambridge University Press. [14] Jaggers, Keith and Monty G. Marshall (2000) Polity IV Project, Political Regime Characteristics and Transitions, 1800-2000, Database Codebook.

[15] Li, Hao and Wing Suen (2004) Delegating Decisions to Experts // Journal of Political Economy. Vol. 112. No. 1. Pt. 2. P. S311-S335. [16] Ludema, Rodney and Anders Olofsgard (2008) Delegation versus Communication in the Organization of Government // Journal of Public Economics, Vol. 92. No. 1-2. P. 213-235. [17] Moe, Terry and Michael Caldwell (1994) The Institutional Foundations of Democratic Government: A Comparison of Presidential and Parliamentary Systems // Journal of Institutional and Theoretical Economics. Vol. 150. No. 1. P. 171-195. [18] Moe, Terry (2006) Political Control and the Power of the Agent // Journal of Law, Economics, and Organization. Vol. 22. No. 1. P. 1-29. [19] Schiavo-Campo Salvatore, Giulio De Tommaso, and Amitabha Mukherjee (1997) An International Statistical Survey of Government Employment and Wages // World Bank Policy Research Working Paper No. 1806. [20] Tirole, Jean (1999) Incomplete Contracts: Where do we stand? // Econometrica. Vol. 67. No. 4. P. 741-781. 15

A Project selection procedure 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 Prob w 1 w 2 w 3 w 4 w 5 w 6 w 7 w 8 w 9 w 10 w 11 w 12 w 13 w 14 w 15 w 16 P 1 B H B H B H B H B H B H B H B H B L B L B L B L B L B L B L B L P 2 B H B L B H B H B L B H B L B L B H B L B H B H B L B H B L B L P 3 B H B H B L B H B L B L B H B L B H B H B L B H B L B L B H B L A b H b H b H b L b H b L b L b L b H b H b H b L b H b L b L b L w i is the probability of discovering this project (can be expressed in the efforts of both politicians and the bureaucracy). Case 1. B H + 2B L > B 0 Information Relationship Political Approved Projects structure Regime Ee Centralization Parl 1,2,3,5,9,10,11 Pres 1,2,3 Informal delegation Parl 1,2,3,5,9,10,11,13 Pres 1,2,3,5 (1-E)e Rubber-stamping Parl 1,2,3,5,9,10,11,13 Pres 1,2,3,5,9,10,11,13 Informal authority Parl 1,2,3,5,9,10,11 Pres 1,2,3,5 E(1-e) Centralization Parl 1,2,3,4,5,6,7,8,9,10,11,12,14,15 Pres 1,2,3,4,6,7 Informal concern Parl 1,2,3,5,9,10,11 Pres 1,2,3 (1-E)(1-e) Parl veto all projects Pres veto all projects The others (non mentioned projects) are vetoed. 16

Case 2. B H + 2B L < B 0 Information Relationship Political Approved Projects Utility structure Regime Ee Centralization Parl 1,2,3,9 6B H and 4b H Pres 1,2,3 5B H and 3b H Informal delegation Parl 1,2,3,5,9,10,11,13 0 and 8b H Pres 1,2,3,5 4B H and 4b H (1-E)e Rubber-stamping Parl 1,2,3,5,9,10,11,13 0 and 8b H Pres 1,2,3,5,9,10,11,13 0 and 8b H Informal authority Parl 1,2,3,9 6B H and 4b H Pres 1,2,3 5B H and 3b H E(1-e) Centralization Parl 1,2,3,4,6,7,9,12 12B H and 0 Pres 1,2,3,4,6,7 10B H and 0 Informal concern Parl 1,2,3,9 6B H and 4b H Pres 1,2,3 5B H and 3b H (1-E)(1-e) Parl veto all projects Pres veto all projects The others (non mentioned projects) are vetoed. 17

B Data Variable Description Year(s) Source(s) LNCGOVEMPP Log of Central Government Administration Employment (in % of population) LNSUBGOVEMPP Log of Non-Central Government Administration Employment (in % of population) LNTOTGOVEMPP Log of Total Government Administration Employment (in % of population) LNGNI Log of gross national income per capita (at PPP in current international $) DEMOC Level of democracy (from 1 to 10) SYSTEM Parliamentary (2), Assembly-elected President (1), Presidential (0) PRES Parliamentary (1), Presidential (0) TENSYS the period when leadership was personality-based YRSOFFS the executive s years in office PRTYIN the executive party s years in office LNPOPUL Log of total population (in thousands) 1992-1994 Schiavo-Campo et al. (1997) 1992-1994 Schiavo-Campo et al. (1997) 1992-1994 Schiavo-Campo et al. (1997) 1994 World Development Indicators (WDI) 1994 Polity IV, Jaggers and Marshall s (2000) 1975-1994 DPI (2006) 1994 Polity IV, Jaggers and Marshall s (2000) 1975-1994 DPI (2006) 1975-1994 DPI (2006) 1975-1994 DPI (2006) 1992-1994 Schiavo-Campo et al. (1997) 18

C Government administration employment Country coverage For models including central government employment N = 58: Albania, Argentina, Australia, Austria, Bangladesh, Belgium, Bolivia, Botswana, Bulgaria, Canada, Chile, Colombia, Czech Republic, Denmark, Ecuador, El Salvador, Estonia, Finland, France, Germany, Greece, Honduras, Hungary, India, Ireland, Italy, Japan, S Korea, Lebanon, Macedonia, Madagascar, Malaysia, Mauritius, Moldova, Netherlands, New Zealand, Nicaragua, Norway, Pakistan, Paraguay, Philippines, Poland, Portugal, Russia, South Africa, Slovak Republic, Spain, Sri Lanka, Sweden, Switzerland, Thailand, Turkey, Ukraine, UK, USA, Uruguay, Venezuela, Zambia. For models including total government employment N = 52: missing El Salvador, Estonia, Macedonia, Moldova, Russia, Ukraine. Variable N.Obs. Mean Std.Dev. Min Max LNTOTGOVEMPP 52.7786067.7506002 -.6931472 2.292535 LNCGOVEMPP 58.0644466.832376-2.302585 1.410987 LNSUBGOVEMPP 52 -.1618908.9928996-2.302585 2.04122 LNGNI 60 8.773263 1.05762 6.429719 10.26221 LNPOPUL 60 9.657441 1.419686 7.01661 13.72515 DEMOC 59 8.440678 1.611074 5 10 SYSTEM 60 1.166667.9596374 0 2 YRSOFFS 60 3.75 3.342713 1 14 PRTYIN 58 7.637931 10.21643 1 46 TENSYS 60 27.15 24.74605 2 64 Table 1: Descriptive Statistics 19

Total Government Employment Central Government Employment (LNTOTGOVEMPP) (LNCGOVEMPP) N=52 N=52 N=52 N=52 N=58 N=58 N=58 N=58 TENSYS.0193.0160.0173.0132.0153.0121.0140.0134 (.0032) (.0034) (.0032) (.0039) (.0039) (.0043) (.0038) (.0047) SYSTEM.2067.1685.1215.1973.1653.1578 (.0899) (.0899) (.0911) (.1128) (.1002) (.1057) LNPOPUL.1249.1182.2479.2463 (.0529) (.0517) (.0617) (.0626) LNGNI.1975.0279 (.1047) (.1143) R 2 0.4270 0.4800 0.5342 0.5669 0.2110 0.2526 0.4248 0.4254 Adj. R 2 0.4156 0.4588 0.5051 0.5301 0.1969 0.2254 0.3928 0.3821 Table 2: Regression Results for Government Administration Employment All regressions include a constant. Robust standard errors are in parentheses.,, and denote significance at the 1%, 5%, 10% and 20% levels, respectively. 20