Ministerial Typology and Political Appointments: Where and How Do Presidents Politicize the Bureaucracy?*

Does ministry type influence profiles in upper-level bureaucracy? To govern, presidents need to 01. maintain control over the content of public policies, 02. make political concessions that earn them enough parliamentary support to see those policies approved, and 03. build or keep in place the bureaucratic competencies necessary to implement them. We argue that the president makes key appointments according to the nature of the policies of each ministry and their centrality in the executive branch’s decision-making process. Employing cluster analysis, we propose an objective classification of ministries into four types: ‘coordination’, ‘redistribution’ (social policy and income), ‘regulation’, and ‘distribution’. We also identify their relationship with the profiles of those who have occupied positions in the middle and upper echelons of Brazil’s federal bureaucracy – what are referred to in Brazilian nomenclature as ‘DAS roles’ (in which ‘DAS’ stands for Direção e Assessoramento Superior) – in the period from 1999 to 2016. Our findings indicate that presidents choose to professionalize the bureaucracy of the ‘coordination’ and ‘redistribution’ ministries, to appoint partisans to the bureaucracy of the ‘distribution’ and ‘regulation’ ministries, and to allocate partisans of the coalition partners to the ministries controlled by those parties.

In coalition governments, heterogeneity is almost the norm. Ministries are responsible for policy decisions and distributing resources that vary significantly from one another. The policies of each ministry are aimed at specific target audiences and render their unique influence on the decision-making process, the government's popularity, and the electoral chances of the parties involved.
For example, while Finance and Planning are central to coordinating government activities and are present in most intersectoral policies (BATISTA, 2013;GAYLORD and RENNÓ, 2015), Local or National Integration have a narrower scope and more discretionary and territorially decentralized budgets, which can be directed towards electoral bases, as well as policies that don't require coordination with other ministries (BATISTA, 2017).
There are ministries responsible for implementing policies with a broad reach, such as Health, while others regulate sectors that often place interest groups on opposing sides, such as Mines and Energy. This division of labor results in ministerial clusters whose appointees would have profiles derived, to some extent, from the three presidential strategies presented above.
Examples of potential choices about top-level bureaucracy profiles to be made based on ministerial policies 2 frequently appear in the politics pages of newspapers. State superintendencies at entities such as Incra, or those linked to the Ministry of Cities (Local), or the Ministry of National Integration, come with an ability to influence political decisions at the local level, which makes them highly desired politically. To cite an example, it has been stated that 'Congress members seek positions at Incra because each superintendency has the power to define how funds will be used and other agrarian questions -as well as the power to make decisions regarding the rural settlements established by Incra. If a Congressman has an ally in the driving seat, it makes it much easier for him to influence decisions that affect his electoral stronghold'.
Some government sectors, such as tax collection, are more impervious to party influence. On hearing that a general coordinator with no experience had been ______________________________________________________________________________________________ 2 We employ the terms 'top-level bureaucracy/bureaucrats', 'upper-level bureaucracy/bureaucrats', 'top policymaking positions', 'top-level appointees', 'political bureaucracy' and 'political appointees' as synonyms. All of them try to convey the meaning attached, in Portuguese, to the notions of 'burocracia dirigente' or 'burocracia de alto-escalão' in the so-called 'cargos de confiança'.
The key feature is to be in a top-level position inside the decision-making bureaucracy after a discretionary appointment by politicians or the President. appointed to the head of one of the Federal Revenue's audit agencies, permanent staff at the agency " [complained] that the appointment of the new general coordinator [...] has broken the tradition followed since the agency's foundation [...] of having an auditor chosen from among agency staff at its head" (O GLOBO, 2019).
The new RFB chief of intelligence remained in office for only four months. To systematically analyze the relationship between ministry type and toplevel bureaucracy's profile, we propose a theoretical classification based on groupings of ministries that considers the types of policies they enact and the budgetary resources at their disposal. We argue that the upper-level bureaucracies' profiles vary according to the four basic activities of the government: coordination, regulation, distribution, and redistribution. Depending on the centrality and visibility of the ministry in question, the president decides between professional or politicized bureaucracies -with a bias favorable to his/her party or allies.
We expect that the bodies that make up the government's decision-making core, as they are priorities, will present a more professional bureaucracy.
Conversely, the president will exert more political control over the ministries that are more susceptible to variations due to ideological preferences and that are visible to the national electorate, such as policies related to social inclusion and reduction of inequalities. The allocation of top-level management positions to allied parties is nearly uniform across different types of ministries. It is stronger in ministries controlled by coalition partners, which are mainly distribution ministries.
To analyze the compartmentalization strategies implemented by presidents, we consider four characteristics: 01. the proportion of top-level bureaucrats coming from federal career paths -i.e., whose entry into the public sector was meritocratic; 02. the percentage of appointees affiliated to political parties. Among party appointees, we specifically consider: 03. appointees affiliated to the president's party; and 04. appointments of coalition party-members. We expect that the top-level bureaucracy's profiles will vary according to the type of ministry. This is based on the argument that the characteristics of public policy influence politics because they affect the expectations of actors, the structure of conflicts, and political debates (Lowi, 1964).
This article contributes to the understanding of the appointment patterns of the federal bureaucracy in three ways. First, we test the classic argument that public policies influence political strategies. This relationship has not yet been tested to explain the association between thematic areas of State intervention and bureaucratic profiles. Second, we propose a ministerial typology that is theoretically grounded and based on objective characteristics of the organizations, and that improves replicability and the possibility of comparisons in different contexts.
Finally, the methodological approach enhances understanding of the bureaucracy's appointment strategies by exploring the variation in time between thematic areas and types of clientele involved.
In the next section, we discuss the theory and the proposed argument. In the third section, we describe the methods and data, and in the fourth section, we cover the results. In conclusion, we summarize the argument and the main contributions of this article.

Presidents and the composition of the bureaucracy
At the beginning of their terms, presidents need to select top-level bureaucrats according to their preferences but are constrained by the political concessions inherent in the process of forming the government. What criteria should they follow to fill these positions? What is the best bureaucratic profile to maximize a president's success?
A frequent solution is to appoint top policymaking positions based on the criteria suggested by Weber for career bureaucracies ([1947]  decisions. Presidents concerned with policy outcomes should make appointments based on merit and give priority to technical competencies. Rauch and Evans (2000) argue that even in high-ranking positions, the fact that those picked have gone through competitive selection processes tends to make them, on average, better than those appointed from the private sector. Stability in a given role leads to the development of skills and accumulation of knowledge for work performance (LEWIS, 2008). Much of the recent discussion on State capacities, the quality of government or governance, involves to some extent the professionalization of bureaucracy -including top and middle-level bureaucratsas a condition for a quality public policy (CÁRDENAS, 2010;EVANS and RAUCH, 1999;FUKUYAMA, 2013;HUBER and McCARTY, 2004;PETERS and PIERRE, 2004;RAUCH and EVANS, 2000).
From this perspective, the development of quality public policies is the parameter for evaluating presidential decisions regarding the composition of the bureaucracy. But creating quality public policies is not the president's only motivation. Political ambitions encourage presidents to align the content of public policies with their political objectives. One of the president's strategies to control his or her government's public policy decisions is to politicize the bureaucracy by appointing people he or she trusts (BONVECCHI, 2014;LEWIS, 2008;MOE, 1985;PEREIRA et al., 2017;RUDALEVIGE and LEWIS, 2005).
Presidents may also politicize public policy development through patronage and distribute positions to expand party support or co-opt interest groups. It is assumed that this use of patronage tends to h ave a deleterious effect on the quality of public policies because the criteria of merit and the respective technical skills required become less relevant (HOLLIBAUGH JR, HORTON and LEWIS, 2014;LEWIS, 2009;PRAÇA, FREITAS and HOEPERS, 2011).
The three motives mentioned above are all used in practice but with differing potential emphasis depending on policy area. Faced with the dilemma between building bureaucratic capacities, exerting control over policy development and implementation, and distributing positions to build political support, the president establishes priorities with differing political and meritocratic criteria depending on the government sector. The result is a 'compartmentalized' government: some areas have a more professional bureaucracy, while in others, there are more political appointees and allies. The sectors most relevant to presidential success are protected from patronage, and more peripheral sectors are leveraged for party support. In other words, there are sectors with less politicization and greater bureaucratic capacity, the so-called 'islands of excellence', and there are more politicized sectors with less bureaucratic capacity (BERSCH, PRAÇA and TAYLOR, 2017;EVANS, 1995EVANS, , 1979GEDDES, 1994;NUNES, 1997). Bersch, Praça, and Taylor (2017) have demonstrated that federal agencies present significant differences in terms of bureaucratic capacity and decisionmaking autonomy. Our contribution addresses how differences in appointment patterns relate to the nature of policies and the political constraints of multiparty presidentialism. We propose a new classification for the different sectors of government, and we identify the reasons for differences in the appointment patterns between these sectors. In other words, we describe the variations in bureaucratic profiles between the ministries and why they exist.

Presidents, coalitions, and the bureaucracy
In Brazilian presidentialism, as in many Latin American presidential systems, the formation of coalitions is necessary for the president to rule with a legislative majority. Bargaining leadership positions is a tried and true method for forming those coalitions and a practice that generates a set of specific incentives.
Presidents in shared governments contend with demands for policy control and patronage within their party and among coalition partners. 1964) argued that the type of public policy determines the corresponding political arena. The characteristics of public policy shape politics because they influence expectations, political exchange, and the structure of the conflict between actors. In our formulation, ministries are unevenly relevant in the Executive's decisionmaking process and in controlling the budget. The objective is to propose a classification of ministries based on objective criteria and to explain why some agencies are more politicized. We are introducing a taxonomy of policy areas to better understand the functioning of politics in the federal government.
Different criteria explain the differences between governmental agencies.
For North America, Lewis (2008)  'Coordination' ministries are highly important in formulating policy, but they do not oversee large budgets. 'Redistribution' ministries are divided between those that steer social policies and those that control social security and labor market policies. The first group plays an important role in the policy design process and controls significant budgets. The second group manages a large budget, but has little discretion in execution, and is less active in formulating new policies.
'Distribution' ministries are peripheral to the policymaking process but have a large budget, mainly discretionary, which can be leveraged for electoral gains. Finally, 'regulation' ministries are not central to the formulation of policies, nor do they have large budgets, but they appeal to specific interest groups due to their control over the regulation of sectors of the economy or society.
The ministry type creates demand for a specific profile of top-level bureaucrats. Coordination ministries are the ministries responsible for elaborating the rules of the game, or the rules about the rules; they constitute the core of government policy design and are not burdened with specific interest groups since they do not control resources that are relevant to interest groups.
Because they control complex and important decisions for the success of the government, they require a predominantly technical profile to achieve their objectives. For these reasons, we expect that these ministries will present greater meritocratic recruitment and a less politicized top-level bureaucracy.
H1: Coordination ministries present higher recruitment of public servants and lower recruitment of party members.
Regulatory and distribution ministries are ministries that target specific clientele or interest groups. In the case of distributive policy, ministries play a marginal role in public policy design but distribute larger, discretionary budgets with a more concentrated geographical impact, making them more visible to voters. They allocate resources to states or municipalities, usually for public works projects (BERRY, BURDEN and HOWELL, 2010;HUDAK, 2014;KRINER and REEVES, 2015;LOWI, 1972).
Regulation ministries are responsible for formulating rules that can pit interest groups against each other and are prized for their relationship with economic interest groups. Given the proximity of the distribution and regulation ministries to specific interest groups, especially in economic terms, we expect them to be more open to the recruitment of personnel from the private sector.
H2: Distribution and regulation ministries present less recruitment of public servants.
Redistribution ministries control on-the-ground public policies and are important for policy formulation and in the allocation of expenditures. The policies developed are salient for voters as well as ideologically divisive. Policies that redistribute income divide society into large groups or social classes (Lowi, 2015(Lowi, , 1964, and that is why they antagonize parties on the left and the right. Lowi (2015Lowi ( , 1964) calls these 'partisan policies'. Politicians with expectations of directing these policies will appoint fellow party members, and, in these ministries, a greater partisan appointment is expected for purposes of political control.
H3: Redistribution ministries present more partisan recruitment from the president's party for purposes of policy control.
The expectation for coalition partners is that the allocation of positions to their party-members is less related to the content of public policies and more associated with the political bargaining needed to garner legislative support for the government. We expect coalition party members to be allocated more evenly across different types of ministries, and most frequently in ministries controlled by coalition partners.  Another potentially important factor is presidential popularity. More popular presidents can leverage political capital and appoint their own people despite party objections. We also included the electoral cycle in the analysis, due to the potential influence of officials entering and exiting their posts, as candidates or campaign workers. Finally, we created a control variable for each president. The next section presents the case, the data and variables, and then the results.

Case, data and variables
We Additionally, the middle echelon is considered to be the segment responsible for the connection and dialogue between the upper echelon, which formulates the policies, and the street level bureaucracy. It is, therefore, the difference in the conception of 'the middle of the hierarchical structure', and the level of political influence at level 05, which distinguishes the way in which we and ENAP define the 'high' and 'middle' level of bureaucracy federal. We considered in our definition the evidence collected in the qualitative studies on DAS positions, available in Lopez (2015) and the fact that level 05 has a level of political influence in the intra-ministerial decision-making process that is much higher than the levels immediately below, as pointed out by Freire, Vianna and Palotti (2015).  Our analysis covers the administrations of four presidents over six terms, from 1999 to 2016 6 . The primary independent variable is the type of ministry. To build the typology, we used four pieces of information: two related to the importance 7 of the ministry in the formulation of policies and two associated with the importance of ministries in the execution of the annual budget.
The first measure is the structural and functional differentiation of the ministries, which indicates the complexity and degree of specialization of the agency (INÁCIO and LLANOS, 2016;MOE, 1985). We use as a proxy the number of The second dimension of importance is the ministry budget. We separate the total budget -measure 03 -from the investment budget -measure 04 8 . The total budget includes all resources -mandatory and discretionary -controlled by the ministry. The investment budget is the portion of the annual resources ______________________________________________________________________________________________ 6 The analysis period starts in 1999 due to the lack of systematic data on previous DAS. We include all ministries, with the exception of the Ministry of Foreign Affairs. The exclusion is based on the different nature of the policies formulated by that ministry, which are not related to domestic policy and which make decisions of a different kind such as international treaties and agreements and not comparable with other ministries. We also exclude the governing bodies of the Presidency, which often do not have their own staff and are an amalgamation of organizations that are now organically connected with coordination activities of the Presidency and now have a life of their own and even have ministerial status, although linked to the core of the government. 7 The term 'importance' is used to quantify participation by the ministry in policy making and public spending. Ministries that produce more politicians are considered more important in that dimension and ministries that spend more resources are considered more important in that dimension. 8 (1) e0004 -14/34 whose execution is discretionary, freely applicable, and which can be more easily oriented towards electoral purposes.
To identify the groups of ministries based on the averages of these variables, we used the K-means cluster analysis technique. The algorithm classifies the observations into groups so that the observations in a given group are as similar as possible (intra-class similarity), and observations from different groups are as dissimilar as possible (inter-class dissimilarity). We use the Hartigan-Wong algorithm, which mobilizes the sum of the square of the Euclidean distances between the observations and the centroid as a measure of intra-cluster variation (KASSAMBARA, 2017).
To create a more accurate description of the presidents' choices, we propose four specifications for the top-level bureaucracy profile -our dependent variable. The first is the proportion of appointed federal civil servants. This specification defines the portion of appointees that have already undergone competitive and meritocratic selection. Although federal civil servants are often chosen based on elective or partisan affinities between them and their superiors, it is reasonable to assume that, on average, having civil servants that are already part of federal careers means more familiarity with administrative protocols and the bureaucracy, which is valuable institutional knowledge that is both time-consuming and costly 9 to obtain. ______________________________________________________________________________________________ 9 It is possible to claim that career servants are as engaged politically as appointees from the private sector or those appointed directly from groups of party members. In general, however, career public servants, even if they hold a personal preference for certain presidencies, have undeniable expertise compared to those appointed from the private sector, due to their day-to-day first hand knowledge of public bureaucracy. It is worth remembering, for example, that the definition of minimum quotas for career easements to occupy DAS positions at levels 01 to 04, in 2005, occurred after the discovery of networks of influence in government agencies originating from appointments of people from the private sector, in the episode called 'mensalão scandal'. The concepts adopted here of 'professionalization', referring to the percentage of federal career employees in the total DAS positions, and 'partisanization', referring to the fact that the appointee is affiliated to a political party is stylized as such for two reasons -first, because there are career servants affiliated to parties, although the percentage is much lower than those coming from the private sector (LOPEZ and SILVA, 2019). We consider each attribute -party membership and recruitment meritocratic -separately. In considering the degree of partisanization of a ministry, it increases with more party-affiliated appointees that are unconnected with federal careers and decreases the percentage of career public servants in the ministry. The affiliated public servant is, therefore, an intermediate case, which carries both the attributes of professionalism and partisanization. Second, we believe that membership is a clearer, and more easily measurable indication of partisanization, but it is not the only one.

The different uses of ministries: building a typology
The proposed typology adopts deductive and inductive reasoning.  Among the ministries with less centrality in formulating, the Executive's legislative agenda are Fishery, Tourism, Culture, and Communications. Each one is responsible for well-defined areas of policies that have activities related to the regulation of specific areas, hence the reduced legislative production. Figure 02 shows the importance of ministries in the policymaking dimension. in other dimensions, since in addition to the formulation of policies, it is also necessary to implement them.    Having defined the five clusters, we now describe the profiles of the appointees in each type of ministry. We then analyze which ministries are politicized by the president, and how.   Party-member appointees represent the most explicit version of the politicization of bureaucracy and are a low percentage of total appointees. We emphasize that membership of the public service and membership of a party are not mutually exclusive. Career civil servants can be party members, and that affiliation can be a reason for career advancement. The data show that the average of party members in DAS positions is 13%, with averages of 12% for the middle echelon and ______________________________________________________________________________________________ 12 Table 02 in the Annex shows the allocation of ministries in each of the five clusters. 21% for the upper echelon. This finding again reveals the preference in filling highlevel positions of the ministries with partisan appointees and is consistent with the fact that these are the positions with the highest salaries and the greatest power to guide and align policies with the interests of allied politicians.

Types of ministries and the profiles of political bureaucracy
As we further distinguish the types of partisan appointments within the ministry between appointees from the president 's party or coalition parties, the results point to greater partisan appointments by the coalition partners. The percentage of appointees from the president's party is 4%, with the average for the middle echelon being 3%, versus 10% for the upper echelon. Appointees from the coalition partners are, on average, 5%, with 4% in the middle echelon and 7% in the upper echelon. In short, the participation of civil servants occupying management and advisory positions is high and partisan appointment in ministries is comparatively low. Partisan appointments, be they of the president's party or a coalition party, are more concentrated in higher-ranking positions.
The following multivariate analyses aim to identify whether key appointment choice is, as we suggest, potentially explainable by the nature of policies, and by the extent to which they vary with the strategies for choosing each minister, as well as contextual characteristics like presidential popularity and the electoral cycle. To interpret the variable coefficients that indicate the clusters of the ministries, the 'coordination' cluster is the reference category, in relation to which the coefficients of the other clusters are interpreted. Table 02 first displays the specifications of the dependent variable for public service appointees, followed by the models that indicate the total number of party members. Regarding appointees from the federal public service, taking the coordination ministries as a reference category, the other clusters show a negative sign, which indicates less meritocratic recruitment and greater politicization. The expectation of less politicization compared to coordination ministries is confirmed.
The cluster of redistribution-social ministries shows no statistical difference in relation to the coordination ministries, indicating that in these areas, the appointment of public servants is also prioritized. At the other extreme, the distribution and regulation ministries are the groups where more individuals from outside public service are appointed to toplevel positions, which is consistent with our hypothesis that such sectors are especially permeable to those appointed from outside public service and are possibly linked to specific interest groups.
Concerning the appointment of party members, the most specific form of politicization that seeks to capture party alignment, the results indicate that all groups of ministries show a positive sign, indicating greater partisan appointment than the coordination ministries. At the upp er echelon, all ministry groups present a statistically significant difference, which suggests that coordination ministries are especially protected against the partisan appointment. However, in    The type of minister affects the profile of the bureaucracy in an expected way. The reference category is non-partisan ministers. When the minister is from the president's party, we see a greater share of appointees affiliated to the president's party. In coalition-controlled ministries, we see a greater share of appointees affiliated to coalition parties 13 . Figures 05 and 06  We observed a wide variation in each ministry's patterns of partisan appointments, depending on the features of its minister. When the minister is from the president's party, there are more appointments of members from the president's party in all ministry types. The ministries in the categories of redistribution-labor, regulation, and redistribution-social had a disproportionate number of appointees in the upper echelon from the president's party. In the middle echelon, the priorities are regulation, redistribution-labor, and distribution. Therefore, appointment from the president's party is closer to a standard aimed at both exerting control over divisive public policies, as we predicted, and at rewarding allies. This is because some of the appointees are placed in ministries peripheral to the decision-making process, but relevant to the party agenda and special interest groups.  In short, presidents choose to professionalize the bureaucracy of the ministries of coordination and redistribution, to make the bureaucracy in the ministries of distribution and regulation more partisan, and to concentrate the coalition's partisan appointments in the ministries controlled by those parties. The results support our general argument that there are different politicization strategies in the federal bureaucracy and that they vary according to the type of public policy that the ministry controls. The partisan appointment is limited in both the middle and upper echelons, although more accentuated in the latter. Partisan appointments from the president's party focus on redistribution and regulation ministries. Partisan appointments from coalition parties predominate in the ministries of their respective ministers. The low percentages of partisan appointees -even in the upper echelon -suggest that politicization is either less than expected or manifests itself outside formal party affiliation. In other words, politicization can be extra-party, arising from professional or friendship networks, ideological and other affinities, which motivate choices without requiring formal party membership.
With these results, we contributed on three fronts to the literature on the politicization of bureaucracy. First, we present a typology of ministerial groups that are objective and theoretically grounded. This typology contributes to increased knowledge of the subject because it is replicable in other contexts. Second, we establish predictions about how different profiles of the bureaucracy vary according to the ministry type. Additionally, that the incidence and types of politicization in each group reflect, in part, the threefold presidential strategy of controlling the policies that constitute the core of the president's agenda, building and preserving legislative support to the government -and the presidential agenda -and implementing it successfully. Finally, we present original findings regarding the different types of politicization -appointees from outside the public service, members of the president's party, and members of coalition parties -over the 21- year period we analyzed. In analyzing new connections between areas of public policy and appointment strategies in the top-level bureaucracy, we hope to have broadened understanding of the strategies for the division of power within the federal Executive branch.