Specialists and Politics: The Recruitment of Presidents and Directors of BNDES in the PSDB and PT Administrations

This text deals with BNDES between 1995 and 2016. It intends to answer two questions: 01. Which institutions did presidents and directors of BNDES pass through in the period before they achieved their positions in the Bank?; 02. Are there any differences, in this regard, between the PSDB and PT administrations? The article uses Social Network Analysis to capture the prior trajectory of individuals before they reached the positions of president and director of the Bank. The data is presented in four sociograms, one for the FHC administrations, another for the Lula administrations, a third for the Dilma Rousseff administrations, and a last one of directors and presidents of the Bank. Our data reveals important differences in relation to the structure of the trajectories of BNDES directors between different administrations, which could be linked to their different economic projects. By way of conclusion we make some final considerations about the data and its possible theoretical implications.

he National Bank of Social and Economic Development (Banco Nacional de Desenvolvimento Econômico e Social -BNDES) was created in 1952, during the second Vargas administration. It was the institution responsible for planning, raising funds, and developing an industrialization project in Brazil. The Bank, which stood out historically due to the role it played in Kubistchek's Target Plan, is currently 1 one of the most important development agencies in the world and the 'operational arm' of the Brazilian government as the provider of funds for investment 2 .
This text is concerned with BNDES between 1995 and 2016, covering the two Fernando Henrique Cardoso administrations (FHC I and FHC II), the two Luiz Inácio Lula da Silva administrations (Lula I and Lula II), and the two Dilma Rousseff administrations (Dilma I and Dilma II). We intend to answer two questions: 01. What is the institutional trajectory of presidents and directors of BNDES in the period analyzed before they reached their positions in the Bank?; 02. and in relation to this point, are there differences between PSDB and PT administrations?
In order to answer these two questions, the article is structured as follows. After this presentation, we first discuss the specialized literature on the nature of BNDES as an institution and we locate our work within this. Next, we present the data collected and the methodology used for the analysis of the previous career structure of all BNDES directors and presidents during the period studied. In the third part, we concentrate our analyses on the institutional trajectory of the Bank's presidents and directors during Fernando Henrique Cardoso's two administrations. Following this, the same analysis is carried out for the presidents and directors of BNDES during the PT administrations. The Lula administrations and the Dilma administrations are presented separately due to the change in the head of the Executive Power. In a final part we analyze the network of directors and presidents with regard to their external or internal origin (in relation to BNDES). The ______________________________________________________________________________________________ data is presented in four sociograms. In order to strengthen our thesis about the political role of presidents in the shaping of directorates of the Bank. Finally, by way of conclusion, we will make some final considerations about the data presented in this article, emphasizing that the opposition between 'technical' and 'political' has to be qualified not with the purpose of refusing the clear existence of highly bureaucratic agencies in the Brazilian state, as is the case of BNDES, but of avoiding understanding the relationship between both as a zero-sum relationship.

BNDES in the literature: from insulation to recruitment
In this section, we present in a summarized form the literature about BNDES. In our opinion, we can divide these works into three groups. The first one emphasizes that the Bank is an insulated entity in relation to politics and seeks to explain how this condition was historically formed; the second set of works emphasize the place of the Bank in the financial system and its adaptative capacity to changes in Brazilian economic development; finally, the third set of works focus on the relationship between the Bank and the Executive, and by extension on the recruitment process of its directors. As we have stated, this division is a question of emphasis, since the last two groups also recognize that the autonomy of the technical staff of the Bank is one of its most striking characteristics. However, alongside this classic theme, these works add other questions which enrich the analysis of the institution.
The first set of studies about BNDES is the best known, and is formed of analyses which, by being essentially concerned with organizational aspects of the institution, assign a central place to the autonomy of the Bank in relation to the necessarily unstable and immediate injunctions of Brazilian political life. According to this argument, BNDES is an example of a modern bureaucratic organization, which by institutionalizing itself managed to create barriers against the political and patrimonialist pressures resulting from the electoral and political party struggle. This capacity to protect the internal logic of the functioning of the institution is seen as a fundamental characteristic of an ideal type of relations between government and political institutions, predominant in some parts of the Brazilian state apparatus constructed after the 1930s and enshrined in the literature by the phrase 'bureaucratic insulation' (NUNES, 1997, p. 34) 3 .
______________________________________________________________________________________________ 3 Throughout the text we use the words 'autonomy' and 'bureaucratic insulation' as interchangeable terms.  1950s, 1960s, and 1970s (BRESSER-PEREIRA, 2014GEDDES, 1994;JAGUARIBE;1977;LAFER;1975;LEOPOLDI, 1992). In these studies bureaucratic insulation is itself a political project which, due to the historical conditions in operation from the 1930s onwards, was victorious -at least in part -, affecting certain dimensions of the Brazilian state and leaving others at the mercy of the patrimonialism of national politics 4 .
Taking into account the problem of 'bureaucratic insulation', some studies have sought to identify which institutional aspects allow the construction of an administration seen as of 'excellence' within BNDES, revealing the historical emergence of a model of innovative management within the Brazilian state apparatus, with recruitment tools based on meritocracy, from the early years of the bank until more recent times (KLÜGER, 2015;SIKKINK, 1998;WILLIS, 1995). Seeking to explain the expansion process of the Brazilian public apparatus, Geddes (1990) argues that reforms with the purpose of increasing state capacity and the condition of insulation only occur when this strategy provides greater advantages to the leaders of the Executive in relation to the alternative form of political patronage. In this way, for her, the construction of agencies such as BNDES was the result of the strategy assumed principally by the second Vargas administration and by Juscelino Kubistchek.
The second group of studies sees the Bank as an institution formed by highly qualified technical staff with a high level of autonomy, but at the same time calls attention to the fact that the institution has managed to readapt to new stages of national development and, for this reason, has been capable of remaining as a relevant ______________________________________________________________________________________________ 4 The paradigmatic example in this sense is the creation of the Administrative Department of Public Service (DASP, in Portuguese) in 1937. DASP materialized at the same time the advances and limits of the modernization of the Brazilian state in this period. On the one hand, it carried out a series of bureaucratic changes in the public service and, on the other, it had to deal with the permanence of officials whose nomination was essentially political, as was the case of the so-called extra-'numerários'. In relation to this, see Daland, 1967. (2021) 15 (2) e0006 -5/30 institution throughout its existence. Luciano Martins' (1985) work is the first to emphasize the capacity of BNDES to expand its scope of action in an environment of constant competition between public agencies. Seeking to increase its fund raising capacity, in the middle of the 1960s the Bank moved from being a public financier to the position of the most important development agency 5 . Colby (2013) argues that the technical capacity of bank staff was a central aspect in the maintenance of its relevance and its high level of autonomy, even with the different governmental macroeconomic orientations during its institutional history. However, as Dias (2017) highlights, the continuity of BNDES' centrality is also associated with the bank's capacity to preserve its space of action in the Brazilian financial system. In general,  Martins (1985) also refers to the influence of state directors who had connections with the private sector as the relevant factor for understanding the business logic adopted by 'state companies' during this period. 6 An extensive literature discusses the economic nature and the impact of the Bank in relation to the financial system. See, for example, the work of Stallings and Studart (2006), Musacchio and Lazzarini (2014), and Ferraz and Coutinho (2017). As for the transformations of the Bank and its relations to the economic and political changes that took place in Brazil over the last two decades, see Gambacurta-Scopello (2020). According to the data of Barboza, Furtado, and Gabrielli (2019), the greatest reduction in the institution's disbursements occurred during the 1990-1994 period, when the total support provided by BNDES corresponded to 0.6% of Brazilian GDP. 8 Barboza, Furtado, and Gabrielli (2019), based on official documents, reveals to be false the idea that, during the Lula and Dilma administration, the Bank acted exclusively as supporter of large companies. BNDES made the so-called 'choice of champions' with the expansion of its funding capacity, but this process was also accompanied by the growth of the disbursement policy for micro, small, and mid-sized companies, which reached 33.4% in the 2015-2017 period. See also Gambacurta-Scopello, 2020. In an interview to Luciano Martins, Maria Conceição Tavares (BNDES, 2016) emphasizes the relevance and influence of the President of the Republic's support for the Bank to implement a specific economic policy. According to her, Delfim Netto's failure to submit the Bank to a privatizing orientation and FHC's success in doing so has everything to do with presidential support for this policy. This article dialogues preferentially with this last set of studies. We do not intend, obviously, to deny the BNDES 'bureaucratic insulation' thesis.
Nevertheless, we believe that our data makes it licit to suggest that the Bank, despite being an institution strongly structured on meritocratic recruitment criteria, suffers the impact of political factors which end up affecting its 'mission' in different governments. In this sense, as we have said, the article intends to show inextricably linked aspects: that the profile of the composition of the Bank's board of directors changes depending on the government of the day and that this change is expressed in the institutional trajectory of those who become directors of the institution. The data presented in this text refers to the directors and presidents of  (GRANOVETTER, 1994;MINTZ and SCHWARTZ, 1985;SCOTT, 1988), since it permits diagnostics of the structure of economic groups and their shareholding and directorship connections, as well as the power of the financial system in this business complex. In Brazil, network analysis gained strength at the end of the 1990s and the beginning of the 2000s, and has been used in studies of connections between the political world and social organizations (ANSELL et al., 2016;CARBONAI, 2019;CERVI et al., 2015;HOROCHOVSKI et al., 2016;LAZZARINI, 2011;MARQUES, 2006MARQUES, , 1999OLIVIERI, 2007;PERISSINOTTO et al., 2017). A review of the use of this method in recent years in Brazilian academic research was carried out by Marques, Bichir, and Moya (2014). Social networks systematized a set of relations (edges) established between certain units (nodes). SNA is an analytical perspective which had the great advantage of including in the analysis the relations between the units observed and not only their attributes. Cf. Wasserman and Faust (1994, pp. 04-06). studies such as those of Loureiro (1997) and Olivieri (2007)  shown in the tables was produced in Gephi 12 .

Network analysis applied to the study of the prior institutional trajectories of BNDES directors
The organization of the network in Gephi 0.9.2. used the 'Force Atlas' distribution algorithm which is based on mechanisms of repulsion force, attraction force, and gravity (CHERVEN, 2015, p. 72-73). In this algorithm, we selected the options 'stabilization function', 'distribution of attraction', and 'adjust by size', nevertheless some nodes still remain superimposed. In order to make the visualization friendlier, we made some adjustments of distances and positions trying as much as possible to preserve the distribution oriented by the algorithm.
Based on this procedure, three types of analysis were carried out on the sociograms presented: first, a grouping analysis, second, centrality analysis, and finally, cohesion analysis. For the grouping analysis an algorithm called Girvan-Newman was used. Through the identification of the 'betweenness' centralities (which will be explained below) of all connections (edges) of the sociogram, the algorithm uses the connection with the highest centrality as the marker to divide the network into two ______________________________________________________________________________________________ The second type of analysis was based on metrics of centrality. Instead of identifying groups of actors, they measure the importance of a determined node within the network. The first metric of centrality we highlight is called 'betweenness'. This measure is produced through an algorithm which calculates the total number of shortest paths between all the nodes (geodesic distances) and assigns a determined value to each node (and in the case of the Girvan-Newman algorithm cited above, also for each 'edge') by the frequency with which they appear on these paths (SCOTT, 2000, pp. 86-87). Institutions with high 'betwenness' are 'brokers', in other words, spaces in which directors with different trajectories and/or long trajectories in various institutions pass preferentially before becoming directors of the Bank.
Along with 'betweenness', we use another measure of centrality, 'degree', which defines the size of the nodes in the sociograms. This refers to the sum of the connections which a determined node has in the network. In relation to the 'one-  We will first look at the characteristics of the network of connections of presidents and directors of BNDES during the two FHC administrations (Figure 01). The only node which competes in size with BNDES in the 'government' sector is the Central Bank, which, in turn, in the case of the two FHC administrations, allows the   Alessandro Tokumoto, Rodolfo Dias, Eric Gil Dantas, Renato Perissinoto (2021) 15 (2) e0006 -17/30

The PT networks
We will now turn to the configuration of the networks in the PT administrations, analyzing first the two Lula administrations and, afterwards, Dilma's two mandates (Figure 02). In the networks of the Lula administrations there is a density of 0.10292, which represents a little more than 10% potential connections. The most significant difference in the network which synthesizes the data from the two Lula administrations is that the 'market' sector is significantly weakened and, in this process, the financial sector practically loses importance. The connection with the market, when it appears, occurs with the productive sector. As we will see, however, none of these companies appear with significant measures of centrality, except for Sebrae. The network of the Lula administration for the recruitment of BNDES' directors  (Table 02). In both metrics, once again BNDES occupies a central place, revealing that it was also an important path in the career of directors and presidents in the There is yet another particularly important difference. We can say that the network of the Lula administrations is state-academic. It is possible to add another qualifier. It is state-academic-political party. Only here the party of the President (PT) acquires a certain centrality and is especially important in the capacity to create intermediation within the network. In the FHC network, the PSDB does not have any 'betweeness' value (it does not belong to any geodesic) against 0.094 (9.4% of geodesics) for the PT in the Lula network. We will see below if these characteristics are maintained in the network of Dilma's administrations (Figure 03).

Figure 03. One-mode sociogram of institutions belonging to trajectories of BNDES presidents and directors in the Dilma administrations
Source: prepared by the authors. Note: Key: Circles represent the 'government' sector, triangles represent the 'market' sector, and squares represent the 'academic' sector. Colors represent 'communitarian structures' identified by the Girvan-Newman algorithm partitioned in 05 structures. Node and label size are defined by their 'degree' level.
The network of Dilma's administrations has a density of 0.11347, in other words, a little more than 11% of possible connections. This data reveals that the network of President Dilma is slightly more cohesive and, as will be seen, strongly concentrated in certain institutions, showing the greatest proximity of the professional trajectory of the directors and presidents in this administration. Even lasting only six years, the network of directors and presidents of BNDES of Dilma 's administrations is surprisingly 'lean' (as well as the shorter period, the rotativity of directors in this administration was lower 15 ), which explains its higher density (since in a network with so few agents it is easier to fill the latent connections with real connections). While the FHC network was strongly centered on the market and the Lula  one is a mixture of the state, academic and research institutions, and the party, Dilma's network is basically state centered. As will be seen, there are institutions linked to the market, but very few. Academic and research institutions are also present, but again in conditions of clear inferiority. The only similarity with the others is the importance of BNDES. Based on this, we can find in Dilma's network an almost total predominance of state agencies, corroborated by the numbers, as can be seen in Table   03.

Technical directors and technical-political presidents
As we have seen above, in all the networks analyzed here, from all administrations, regarding the professional trajectory of its directors, BNDES occupies a leading role in the two measures of centrality. In other words, in all networks BNDES is the institution which establishes most connections with the other points on the networks and which has the greatest intermediation capacity for nodes in the networks. This suggests that passing through BNDES is an important condition to reach the position of director. In this sense, as we have stated, our data shows that the recruitment of the Bank's directors follows the logic of bureaucratic insulation and, therefore, the absence of individuals who use 'side entrances' (a political indication, for example) to reach this position is expected. In conclusion, the Bank recruits and promotes its own staff.
However, this does not function for presidents. In relation to this, see sociogram below ( Figure 04). The above network is formed, unlike the others, only of individuals, containing all the directors and presidents of the Bank during the period analyzed. In this network it is possible to see an intense tangle of connections between the Bank's directors (identified as blue), indicating, as we have said, that they tend to pass through the same place that connects them all, in other words, BNDES. To the contrary, the presidents of the Bank (identified as red) appear as external nodes to the mesh of connections which characterizes the universe of directors, indicating that they tend to be recruited outside

Conclusion
This article uses social network analysis (SNA) to produce a panoramic portrait of the structure of institutional positions previously occupied by the presidents and directors of BNDES. This strategy has two advantages. First, it allows us to go beyond those studies that limit themselves to identifying an individual's last position as evidence of their connection with sectors external to BNDES. The second advantage refers to the fact that network analysis allows an overview of the previous trajectories of actors and, more importantly, the privileged place which certain institutions assume in it. The presentation of sociograms and measures of centrality makes it easier than in simple frequency tables to observe the phenomenon, and also allows, through precise visual and numerical information, to quickly identify the importance of certain paths towards command positions within the Bank.
By using this strategy, network analysis allowed the following empirical findings about the trajectories and connections of directors and presidents of BNDES between 1995 and 2016.
First, in all governments the recruitment of BNDES' directors cannot ignore the highly institutionalized character of the Bank, or in another words its insulation. Whatever the ideological orientation of the government, most directors are individuals whose careers, whether fully or partially, were based within the institution itself. This empirical finding is in full agreement with much of the literature that draws attention to its bureaucratic structure and its insulation when it comes to the degree of control which the Bank has over the recruitment process of its members. However, important differences were detected in the networks analyzed by us in both the PSDB and PT administrations.
In the FHC administrations, recruitment to the board of directors and presidency of BNDES was strongly marked by the presence of sectors linked to the market, notably financial institutions. In the Lula administrations, change is significant since the dominant points of the network come from within the state and from research institutions (universities and research centers). Furthermore, recruitment is also party based, with the PT being an important intermediation point in the network of the Lula administrations. Finally, the Dilma network is strongly state based, without significant ties with research institutions, with the president's party, or even the market. What theoretical considerations can we make based on these empirical findings? With the nature of the data presented it is difficult to securely go beyond the findings described above. We can speculate with some pertinence in three directions. First, it is always very difficult to establish a clear and direct relationship between the agent's social and professional backgrounds and their decision-making process in a given institution, above all if we do not make a detailed analysis of the concrete decisions taken by the board of the Bank in the period analyzed here. Nevertheless, we believe that our data points to some coherence between the structure of institutional positions identified and the nature of the political project of the governments analyzed. The privatization put forward by the FHC administrations, with the support of BNDES funding, went along well with directors with a pro-market bias; Lula's economic projects, with their developmentalist reminiscences, tend to be preferentially implemented by individuals from university institutions, such as Unicamp and UFRJ, with clear sympathies for this orientation and who have essentially not passed through market institutions; finally, the characteristics of Dilma's network appear to express her well known technocratic and statist perspective.
Second, it is also worth looking at the homology between the structures of the trajectories analyzed here and the nature of the institution studied. BNDES is a Brazilian state institution, structured according to the typical criteria of modern bureaucracy. At the same time, it is an institution with an evident function, which is to produce development policies using various instruments for this, from the concession of long-term credits to shareholding and direct participation in the administration of certain companies. In this sense, BNDES has been, since its foundation, an institution with an evident functional connection with the world of private companies, and, by extension, with those individuals who work in this sector. According to this, it is difficult not to analyze the three preferential types of trajectories identified by us -the market, university institutions, and the state itself -as effects of the objective position of the Bank in the system of institutions of the Brazilian state. As part of the state, the Bank has to establish itself as a public organization, create administrative staff, and, therefore enable internal careers to the top. However, as an instrument of economic policy, which necessarily varies in accordance with the project of the government of the day, the Bank cannot simply ignore groups of actors outside its walls, whether in the market, universities, or individuals from other sectors of the state. Finally, our data is important to reiterate something trivial, but almost always forgotten, if not in academic articles, at least in the common sense: there is no logical, theoretical, or empirical incompatibility between the technical qualification of an institution and its use by government officials for their political projects. Obviously, an institution may not bureaucratize itself if it is available to politicians interested in using its assets as a means of exchange in patronage relationships. However, this does not mean that the total bureaucratization of an institution makes it immune to the injunctions of politics. This technocratic dream was not achieved, as we have seen, not even in one of the most insulated institutions of the Brazilian state. Technical knowledge can remain intact and, at the same time, lend its expertise to efficiently achieve distinct political projects.

Translated by Eoin Portela Submitted on December 04, 2019
Accepted on July 28, 2020