The STF figures prominently in the Brazilian political landscape. As the court daily addresses controversial cases and issues of great repercussion, the behavior of the court – and its justices individually – is the subject of much discussion, especially in the current context of political instability. The debate on the justices' inclinations when voting is related to the discussion on the appointment and nomination processes taking place whenever vacancies open in the court. Analyses have sought to uncover possible links between presidential appointments confirmed by the Senate and the behavior of new justices in the court.

The Brazilian Constitution only sets minimum criteria for choosing STF justices, thus granting the president and the Senate significant discretion to, respectively, appoint and confirm (or not) new members to the court. According to the Constitution, STF justices must be citizens over 35 and under 65 years of age, with remarkable legal knowledge and an unblemished reputation. They are nominated by the President of the Republic after being confirmed by an absolute majority in the Federal Senate (article 101 and sole paragraph of CF88). But first, the Committee on Constitution and Justice must hold a public hearing (known as ‘sabatina’) to scrutinize the appointee, a process that has become increasingly important in recent years. Candidates thus may be recruited from very different legal careers, from the judiciary or other legal professions; they may even have a partisan/political profile.

Despite the importance of the position of justice, we know little about how presidents choose a new justice and how the Senate examines and decides whether to confirm the appointment. Although appointees are always confirmed and sworn in, presidents still may face several constraints and pressures in the process. Lemos and Llanos (2007) argue that “presidents do what they want, but not when they want” ( LEMOS and LLANOS, 2007 , p. 117 ), that is, the amount of time it will take to confirm a new justice depends on how challenging the political context is, although this process is faster in Brazil than in Argentina and the United States, as shown, respectively, by Lemos and Llanos (2007) and Almeida (2015) . Llanos and Lemos (2013) argue that such rapidity and greater success in confirming presidential appointments in Brazil would be more the result of the executive's strategy of anticipating the interests of its governing coalition in the Senate than of an apparent dominance of the president over the legislature. In their comparative study, Jardim and Garoupa (2011) have also shown that the Senate in the United States is historically much less deferential to presidential appointments than in Brazil. On the other hand, whenever a vacancy opens on the side of the court opposite to the Senate (taking the median on both cases as a reference) and opposite to the preference of the American president, the Senate and the presidency succeed in appointing and confirming a candidate capable of moving the median axis of the court in the direction of their interests ( Cottrell et al., 2019 ). This capacity to successfully nominate justices, if it exists in Brazil, has never been demonstrated by the analyses.

In any case, once confirmed, justices have great stability in their positions, and so they can be highly independent of the president who appointed them. In addition, they become part of a court that is simultaneously a constitutional court and the higher body of the Judiciary, which gives justices special powers over all courts and over the other branches. For these reasons, the relationship between presidential appointment and the individual behavior of justices cannot be taken as unquestionable or unequivocal – the topic, therefore, is still one of the most disputed not only among scholars but also among those who are publicly discussing how the court operates.

Some of the studies conducted so far have not found evidence that presidential appointments influence the behavior of the STF. Even though different parties have been elected to the presidential office – a diversity that would be reflected in the successive distinct appointees to the STF –, this variation would not lead to significant changes in the behavior of the justices or the court ( Jaloretto and Mueller, 2011 ; Lopes, 2013 ); not even polarization has been identified between justices appointed by different political parties ( Ferreira and Mueller, 2014 ). However, recent studies have challenged this conclusion, as they found new evidence of bias in justices’ behavior resulting from presidential appointments. According to Desposato, Ingram, and Lannes Jr. (2014) , the justices appointed during Lula's government are more ideologically aligned with the government, tending to cluster in opposition to those appointed in previous governments. Oliveira (2018 , 2012a ), in turn, has demonstrated that there is greater cohesion among justices appointed by the same presidents compared to the average level of cohesion in the court. Martins (2018) and Mariano Silva (2016) , each revisiting the court's decisions in the post-88 period in their own way, found that there was a significant division among justices, showing that there were different coalitions in different periods.

Judicial behavior is a recent research field in Brazil, still lacking investment, consistent approaches, and firm conclusions. There is relative consensus about the standing and behavior of the STF as a collective, as an institution² , but little about the individual behavior of its members. Most studies have investigated whether there is a relationship between justices’ individual decisions in Direct Actions of Unconstitutionality (Ações Diretas de Inconstitutcionalidade, ADIs) and the political party of their appointing presidents, leaving aside other aspects that could affect the decision-making in various types of cases. Altogether, there are good descriptive analyses of aggregated data, but they are inconclusive about the determinants of the individual decisions of justices.

Oliveira's work (2018 , 2017 , 2012a , 2012b , 2011 , 2008 , 2002 ) has opened a research path for explaining the judicial behavior of justices in the STF, especially the relationship between justices' decisions and their professional lives before joining the court. However, despite the important advances made, these studies did not offer conclusive results about that relationship. In her first study, in 2002, Oliveira found that the decisions (on direct constitutional review) that were more aligned with the constitutional text were issued by judges who had made their careers in the judiciary; therefore, by more conservative justices, in Oliveira’s terms. Votes less bounded by the letter of the Constitution were found to be issued by judges who have not made a career in the judiciary. Subsequently, working with a sample of 300 ADIs, and among these, another sample of 52 nonunanimous decisions, Oliveira (2011) found evidence of a counterintuitive relationship: justices with career paths more linked to politics were more likely to make more conservative and restrictive decisions (in Oliveira’s terms), while those with a background in the Public Prosecution – a comparatively more technical career – were making more political decisions³ . However, in the above-mentioned study, justices are classified ‘a posteriori’ as restrictive or activists based on whether their decisions are classified as technical or political – therefore, without any measure that would allow them to be codified ‘ex ante’ in that way. Moreover, the evidence found in the analysis of voting networks is not confirmed when the individual behavior of the justices is examined, which leaves open the question of to what extent justices’ career paths actually influence how they vote. In a 2018 study, Oliveira identified that as the number of justices who had made their careers in the judiciary and are participating in each ruling increases, consensus is more frequent, provided that the issues are not nationally relevant, in which case dissension among justices increases ( OLIVEIRA, 2018 , p. 272). Justices with this profile also tend to vote as a bloc in favor of demands from the national judges' association (Associação dos Magistrados Brasileiros) and for the interests of the judiciary ( OLIVEIRA, 2017) . Martins (2018) uses the classifications of justices’ profiles created by Olivera when trying to explain the division among justices when judging ADIs and claims of non-compliance with a fundamental precept (Arguições de Descumprimento de Preceito Fundamental, ADPFs). His conclusion, however, is that, in what concerns the justices' behavior in these types of cases, their career paths have less explanatory power than the party by which they were appointed.

Other studies on the profile of judges – some of them closer to the field of sociology of professions, others to institutional studies – have investigated how professional ideologies are built in the legal field and which values and beliefs guide judges⁴ . As for the STF, some studies have described the justices' profiles, identifying relevant elements in their career paths, assessing how autonomous the court is in relation to the political elites, also classifying the justices as technical or political⁵ . Almeida (2010) has shown how legal capital and political capital may combine into four ideal types of career paths leading future justices to the STF⁶ . Examining justices' interventions in the public debate between 2013 and 2016, Almeida et al. (2017) describe groups of justices who, with few exceptions, are marked by career paths either more centered on politics or more insulated in the legal world ( ALMEIDA et al., 2017, p , pp. 79-80). On the other hand, Santos and Ros (2008) contend that, despite the increase in the proportion of justices with exclusively legal careers, the borders between the political and legal fields are still far from being impervious. Wagnitz (2014) classified into four categories 149 justices who were members of the court between 1891 and 2012⁷ ; however, when confronting these categories with two important STF rulings (same-sex union and party loyalty), she did not find the expected relationship between professional background and the legal rationale in justices’ votes. Fontainha et al. (2017) , starting from a typology very similar to the one we present next, demonstrate that justices' career paths have strong traces of politicization since many of them ascended professionally through appointments before joining the STF. However, such politicization does not mean that justices take part in the partisan dynamics, as few justices have actually occupied elected positions ( FONTAINHA et al., 2018) , as we will see below. Studies on career paths are found in reference to other judicial institutions, such as Bonelli's (2001) work on the São Paulo Court of Appeals (Tribunal de Justiça de São Paulo), and Wagnitz and Moraes' (2014) study on the Paraná Court of Appeals (Tribunal de Justiça do Paraná).

We intend to present a new classification of career paths of STF justices and analyze decisions that are illustrative of how such classification could help understand judicial behavior – we do so by considering the rules and contexts in which such decisions are made. In addition, we seek to specify which aspects of justices’ professional background are decisive in shaping their profiles and judicial behavior, particularly 01. the type of legal expertise developed throughout their careers; 02. the means through which they ascended professionally, and 03. the types of interests to which each justice was professionally associated. Therefore, we abandon, to a large extent, the ‘political’ x ‘technical’ dichotomy that has featured so prominently in this scholarship in order to create ‘ex ante’ criteria for classifying justices and more properly operationalize the independent variables – and their complexity – when explaining justices’ votes.

Quantitative studies on the STF – which are almost always focused on direct actions of constitutional control – have already revealed that unanimous decisions are numerous in the court, that the rapporteur's vote is significantly important, that Justice Marco Aurélio prefers being in the minority (the justice of the ‘defeated vote’), that public hearings have little bearing on the justices’ votes, among other findings. The search for the individual ‘decision algorithm’ of the justices still has not led to conclusive results. Qualitative analyses, in turn, usually rely on exemplary case studies that are not necessarily based on direct actions of constitutional control; however, ‘habeas corpus’ cases, extraordinary appeals, and, more recently, criminal cases can be analyzed qualitatively⁸ . This class of studies tends to criticize the court's modus operandi, pointing out that justices have an individualistic behavior⁹ and suggesting that there is not much of a deliberative nature to their collective decisions¹⁰ . Even hard cases or exemplary decisions in the court, those of great social and political repercussion, are criticized for not exactly presenting a ‘ratio decidendi’ and for being unable to produce solid jurisprudence¹¹ . This line of research has also failed to find explanatory variables for the behavior of STF justices.

This study combines quantitative and qualitative approaches to expand our knowledge of the decision-making process in the Brazilian high court. The article focuses on developing and applying a new typology related to the justices' career paths, without disregarding the internal and external institutional rules of the court, the context in which decisions are made, and the influence of other actors' interests. This new typology is based on our own empirical definitions and data collection but is in line with pioneer studies on this topic in the field of sociology of professions and elites.

First, we examine the justices who joined the court between 1988 and 2021, totaling 35 judges in our classification¹² . Then, we consider the number of years each justice dedicated to professional activities before joining the STF, using the year of the first relevant job as a starting point. The average and median years of previous professional activities were high: 28.1 and 27 years, respectively, ranging from the less experienced, Francisco Rezek I (still, with 13 years of activity between his first job and his first term in the court in 1983) to the more experienced, Paulo Brossard and Eros Grau, who were both appointed after 41 years of professional activity (in 1989 and 2004, respectively). Although the variation in age is therefore wide, one cannot infer, without further research, that age is a decisive factor in presidential choices, as Prado and Turner (2010) assert. On the other hand, we will show next that presidential appointments may be aiming at more specific goals than those of the ‘bargain’ within the governing coalition or of the ‘signaling’ to society, as considered by Arguelhes and Ribeiro (2010) .

As for the classification of justices' professions before they joined the STF, instead of codifying justices in one category (either technical or political), as was done by Oliveira (2011) , we proceeded as Fontainha et al. (2017) and counted the number of years dedicated to different types of professional and political activities, a procedure that has allowed us to create hybrid and more complex profiles. In a first formulation, made by Arantes (2018) , five main ideal types were identified – they are represented in Graph 01 , which already presents the career paths of each of the 35 justices we consider in our study. Although justices may have more than one professional activity throughout their careers, simultaneous activities were not added together, that is, we assigned only one code per year in their career paths that corresponded to the activity representing their strongest occupational link. Multiple codifications for the same year would likely blur the profiles and put an end to building the typology itself. Also, ‘strongest occupational link’ refers to the occupation that gives justices the greatest stability in their professional life – one of the most important aspects in our typology – and not necessarily to the payment, reputation, or popularity involved. We should stress that this typology is built for analyzing STF justices’ decision-making, not for assessing how justices built their condition of ‘supremeables’ (‘supremávei s ’) ( FONTAINHA et al., 2018, p , p. 121). More details about the typology and access to the database for replication purposes can be found at Harvard Dataverse¹³ .

Graph 01 Source: Created by the authors based on database available at Harvard Dataverse ˂https://doi.org/10.7910/DVN/N69VJP˃.

‘Legal type I’ refers to holding a judgeship as the main professional activity. There were very few justices who had worked as judges throughout their professional lives before being appointed to the STF. Among the 33 justices, only Cesar Peluso and Carlos Velloso began their professional life as lower court judges and remained in the judiciary until they were appointed to the STF. However, only Peluso entered a judgeship through public competition and there remained without engaging in direct political interactions to ascend in his career: Velloso became a federal judge in 1967 after being directly appointed by the President of the Republic, not after passing a civil service entrance examination. Ten years later he would be appointed to the former Federal Court of Appeals (Tribunal Federal de Recursos), transformed in 1988 into the Superior Court of Justice (Superior Tribunal de Justiça, STJ in the Portuguese acronym), from where Velloso was later appointed to the STF. Sixteen justices had held prior judgeships, but fourteen of them had dedicated to other professional activities throughout their lives, and five became federal judges by presidential appointment when there was still no civil service entrance examination, that is, when recruitment was not done through public competition (until 1977). As we will further discuss, a justice does not turn into a ‘pureblood’ judge simply by holding a judgeship for some time – even if for a long time – before joining the STF. Sometimes, entering and ascending in the judicial career relies on political appointments to positions reserved to the ‘quinto constitutional’¹⁴ or on appointments to superior courts in Brasilia, which require good connections from the beneficiaries, either through professional associations or directly in the political sphere. For this reason, our typology will be expanded to go beyond the five main types of career paths, as we will see next. All in all, type I encompasses the greatest number of years: 313, or 31.8% of the total number of years we examined (983).

‘Legal type II’ concerns legal careers (other than the judicial career) in which candidates are admitted through public competition, not political appointment, such as careers in the Public Prosecution (Ministério Público), Offices of the State Attorney (Procuradorias de Estado), the Office of the Solicitor General (Advocacia Geral da União), among others. Sixteen of the 35 justices held positions of this type, but only Carmen Lucia remained as state attorney in Minas Gerais all her professional life before joining the STF. We take her case to exemplify one of our methodological decisions. It is known that Carmen Lucia became prominent more because of her active participation in legal debates and her teaching experience than because of her position as state attorney (although such activity was also relevant since she became head of the institution). However, as in the cases of several justices who combine their main professional activity with teaching positions, we prioritized the activities that represent their strongest and most stable professional link, even if their legal reputation was built in the university or the broader legal world. A similar case is that of Roberto Barroso: in addition to being a distinguished scholar, he worked as an attorney in highly important cases in the STF. Nevertheless, to be consistent with our criteria, in our classification we prioritized the fact that he had been a state attorney in the state of Rio de Janeiro since 1983 after passing an entrance examination – a position he insisted on keeping until the eve of being sworn into the STF. In other words, although Barroso had been very successful as a lawyer and professor – and always bearing in mind that our typology only selects one occupation per year in the career path classification –, Barroso did not break his professional link with the state attorney’s office until he became a justice. In a recent book, he said that, after passing the entrance examination, “with a comfortable financial condition, I started to dedicate myself to the state attorney’s office and to my true passion, which has always been the academic life”, and then he told how it took him a few years to establish himself as a professor ( BARROSO, 2020 , p. 28).

Thus, legal type II highlights not only legal careers but also entrance through public competition and the professional stability provided by tenured positions, even if combined with other prestigious activities in which future justices had been engaged. To be sure, holding a civil service position, after passing an entrance examination, may have provided the security necessary for devising more ambitious plans and building the appropriate legal profile and the required respectability, perhaps even with an eye on a future vacancy on the STF. Legal type II totaled 200 years, or 20.3% of the total.

‘Legal type III’ covers private law, teaching (when it is the main professional activity), and entry-level assistant jobs in private law firms. Of the 35 justices, only Eros Grau was in this category all his professional life before joining the STF. He became a respectful professor and attorney, known for writing legal opinions. However, not less than 24 of the 35 justices had some experience in legal type III at some point in their professional life, totaling 280 years, or 28.5% of the total. Although some justices were attorneys only at the beginning of their careers – until they assumed civil service positions –, almost half of those who had a professional experience of type III remained there for 40% or more of their professional life before joining the STF (11 cases). Nunes Marques, Ayres Brito, Nelson Jobim, Maurício Corrêa, Sepúlveda Pertence, and Célio Borja, among others, are examples of justices whose career paths follow this pattern. What distinguishes this category, therefore, is that the professional activities are pursued outside of the legal careers in the state; this category thus focuses more on private law (thus, on the logic of private market relations) and/or on academic teaching. Also, there is not much stability, and ascending professionally depends on the private, associative, and partisan relations of those who will one day become justices. It should be stressed that those who become professors through public competition do not equate with those who passed entrance examinations for the careers included in types I and II, either because academic activities are usually combined with working as an attorney, or because we do not have any case of a professor exclusively dedicated to teaching (em regime de dedicação exclusiva). This is also why we prefer not to separate them into distinct categories.

Type IV, ‘Reliable Politician’ (Político de Confiança), includes holding positions within the state bureaucracy, generally focused on legal activities and always accessed through political nomination (e.g., state secretariats, ministries, Solicitor General, etc). Although the Constitution establishes that STF justices must have ‘remarkable legal knowledge’, this condition can be fulfilled in less traditional legal careers. Although 22 justices fall into this category at some point in their professional life – and many were heavily scrutinized in public hearings in the Senate because of their proximity to the political sphere (cases of Alexander de Moraes, Dias Toffoli, and Gilmar Mendes) –, most of them remained in this type of position for less than half of their careers. Type IV thus highlights the fact that not only justices ascend professionally through political connections and unstable positions but also defend partisan interests and/or the executive authority to which they are subordinated. Considering the total number of years we examined, this category totaled 139 years, or 14,1% of the total.

Finally, ‘type V’ includes those with a law degree who held ‘elected positions’. Strictly speaking, these are the politicians. The fact that the Constitution grants the president great formal discretion in choosing new justices would make us believe that the governing coalition would frequently appoint their fellow members as new justices based on their self-interest; however, only four of the 35 post-1988 justices had been involved with electoral politics and held elected positions: Nelson Jobim (appointed by Fernando Henrique Cardoso in 1997), Maurício Corrêa (appointed by Itamar Franco in 1994), and Paulo Brossard and Célio Borja (appointed by José Sarney in 1989 and 1986, respectively). Therefore, it has been more than two decades since a politician who had previously held an elected position is appointed to the court. Of all five categories, this is the one with the least number of years of professional experience: only 51, or 5.2% of the total of 983 years.

Overall, despite the variation in professional profiles among the 35 justices represented in Graph 01 , our typology demonstrates that the idealized images of a ‘pure judge’ or a ‘pure politician’ we see in the presidential appointment folklore are rare types among those nominated to the STF. This result suggests that presidents and Senate majorities are not using the full scope of their constitutional prerogative, despite calls from some groups for exclusively career judges to be appointed and the desire of others to choose just among party members. Everything suggests that this profile does not represent a path to the court. On the other hand, the results reveal that even justices with legal careers – including those who have held judgeships – interacted with the political world in their corporations, class associations, and party politics. The hands that reach the STF were not used only to handle books or forge unblemished reputations, but they also left their fingerprints in various fields of political activity.

As previously mentioned, our typology goes beyond the formal aspects commonly observed in dichotomous classifications of career paths in order to include more substantive elements that could lead to new hypotheses about the judicial behavior of justices. Also important is that we do not derive the profiles of justices from their votes; we have built the profiles ‘ex ante’ to try to explain their votes. Thus, if the justices' visions of the world and the law are to some extent shaped by the time they dedicated to certain professions and activities before entering the court, it is important to specify which factors, within these professions, can produce this long-term effect. So, we expand our typology to distinguish three other fundamental aspects in the career paths of justices: 01. The type of legal expertise, if focused on prosecuting, defending, or judging; 02. Whether professional rise derives from the stability provided by legal career positions assumed through public competition, from appointed positions – among which those known as ‘cargos de confiança’ (when the appointee may keep its position only as long as the appointing authority wishes), but also other forms of political appointments, such as the ‘quinto constitucional’ (a tenured position following appointment) –, or from private law (market) and teaching; 03. Whether the professional experience is more linked to the defense of private and partisan interests of the parties, the defense of public interests, or the adjudication of cases/disputes between the interested parties. These three highlighted elements in our typology can shed some light on the following aspects: 01. the most long-lasting role played by each justice when processing cases in their careers before entering the court (defending, prosecuting, or judging), 02. the condition of either independence due to tenured positions assumed through public competition or political dependence derived from relying on political appointments to ascend professionally, and 03. the type of interest defended for the longest period, whether partial/private/partisan or public. As with the five types of career paths we presented before, details of this second step in our typology, as well as the replication data, are found at Harvard Dataverse¹⁵ .

The STF combines the role of the court of last instance of the Judiciary and the role and characteristics typical of a constitutional court. Its justices, therefore, are expected to decide on simple and complex cases, of minor or greater political, economic, and social repercussions, cases more or less adherent to legal standards, more or less open to constitutional interpretation. All this variation in cases and types of outcomes challenges the professional expertise developed by justices before joining the court, as well as the dependence-based relationships built (or not) during their careers and the types of interests they were used to defending. Given that the rule for nominating and appointing justices to the STF allows for various profiles to enter the court, our general hypothesis is that these different profiles will be reflected in the act of judging.

To be more specific, our typology considers that 01. STF nominees bring to the court the expertise in prosecuting, defending, or judging they developed before joining the court. Through these previous activities, future justices learn specific procedural techniques, law interpretations, and probably different worldviews. We expect that the various inclinations among justices would influence how they vote and justify their votes since the types of legal actions heard in the STF allow for these various inclinations not only to be expressed but also to organize the judicial decision-making in the court. Therefore, ‘ceteris paribus’, as an example, it is likely that justices with a background in the Public Prosecution will stand out for how they mobilize prosecutorial elements, while those who worked as attorneys will be more sensitive to defense arguments. Our typology also considers that 02. although security of tenure until mandatory retirement can neutralize the effect of previous ties and reshape the behavior of a new justice, it will not completely neutralize the effect of having career paths dependent on political appointments or based exclusively on civil service positions secured by competition. Even if there are no debts derived from appointments to pay off, such experiences result in different combinations of the Weberian ethics of conviction and responsibility among justices. Thus, ‘ceteris paribus’, as an example, it is likely that justices who have held ‘cargos de confiança’ or ascended professionally through other types of appointments (of a political or even corporatist nature) would be more sensitive to the consequences of their decisions than those with more insulated careers. As an extension of type 02, our empirical model specifies that holding a judgeship in superior courts – which represents a mix of judgeship and political appointment – and living in Brasilia when holding high-level federal positions at the time of the presidential appointment – which does not include, therefore, simply living in Brasilia or holding positions in courts of second instance, such as the Tribunal Regional Federal da 1^a Região (TRF1) – would be an indication of professional rise through judicial channels or ‘cargos de confiança’ in the federal executive. Finally, 03. security of tenure does not erase the type of interests defended by new justices throughout their careers. We distinguish three types of experiences related to interests: the defense of private/partisan interests or public interests and the adjudication of cases/disputes between the interested parties. Thus, ‘ceteris paribus’, as an example, we consider that attorneys – but also state attorneys and solicitors general – are inclined to pursue the private or partisan interest of the party they represent, while members of the Public Prosecution are encouraged to pursue public interests, and judges are inclined to adjudicate cases/disputes between the interested parties in specific cases. Given the variety of cases that reaches the STF, exploring the interest dimension seems essential, if not in substantive terms, at least in procedural terms; we postulate that previous professional experiences have somehow shaped justices’ profiles, which allows them to identify and defend interests based on their own preferences. Finally, this typology further supports the idea that the STF is not simply a court of justice, operating under predictable technical standards, protected from political interests and dynamics. Above all, the STF is embedded in the heart of the political system, an active participant in the country's governability, sometimes with a leading role, sometimes conducted by the ‘judicial machine’ of which it is the court of last instance. The justices’ various profiles channel multiple competencies into the STF rulings and bring to the court different levels of previous exposure to politics, to notions and practices of judicial independence, and to public, partisan, corporate, and private interests they might have previously defended and that often appear in the cases decided by the court.

Although our typology does not eliminate the overlap of different professional experiences in the career of a justice, it has the merit of separating and identifying elements that allow us to specify the behavior of justices based on these same elements. Table 01 shows how this typology was used to classify the subgroup of justices who participated in the ‘Mensalão’ trial and whose votes will be examined in the next section. It should be noted that this typology leads not to a dichotomous but to a multivariate classification of the main characteristics of the justices, based on the number of years dedicated to the different professional experiences, if they represent at least 30% of the justices' professional life. Thus, we only coded with ‘yes’ in Table 01 the characteristics of expertise, professional rise, and interests associated with activities that occupied at least 30% of the time of the justices' career paths before entering the court. For example, a justice may have had more than one mean of professional rise throughout his or her career, but our analysis only considered the activities that occupied at least 30% of his or her professional life.

Table 01

Justice	Appointed by	Superior Court	Brasília	Expertise			Professional rise				Interests
				Prosecuting	Defending	Judging	Public Competition	Appointed position	Elected Position	Market	Private/ Partisan	Public	Adjudication between parties
Celso de Mello	José Sarney	No	Yes	Yes	No	No	Yes	No	No	No	No	Yes	No
Marco Aurélio	Collor	Yes	Yes	No	No	Yes	Yes	No	No	No	No	No	Yes
Gilmar Mendes	FHC	No	Yes	No	Yes	No	No	Yes	No	No	Yes	No	No
Joaquim Barbosa	Lula	No	No	Yes	No	No	Yes	No	No	No	No	Yes	No
Ayres Britto	Lula	No	No	No	Yes	No	No	Yes	No	Yes	Yes	No	No
Cezar Peluso	Lula	No	No	No	No	Yes	Yes	No	No	No	No	No	Yes
Ricardo Lewandovski	Lula	No	No	No	Yes	Yes	No	Yes	No	No	Yes	No	Yes
Carmen Lucia	Lula	No	No	No	Yes	No	Yes	No	No	No	Yes	No	No
Dias Toffoli	Lula	No	Yes	No	Yes	No	No	Yes	No	No	Yes	No	No
Luiz Fux.	Dilma	Yes	Yes	No	No	Yes	Yes	No	No	No	No	No	Yes
Rosa Weber.	Dilma	Yes	Yes	No	No	Yes	Yes	No	No	No	No	No	Yes
Teori Zavascki	Dilma	Yes	Yes	No	Yes	Yes	Yes	Yes	No	No	Yes	No	Yes
Roberto Barroso	Dilma	No	No	No	Yes	No	Yes	No	No	No	Yes	No	No

Source: Created by the authors based on database available at Harvard Dataverse ˂https://doi.org/10.7910/DVN/N69VJP˃.

In descriptive terms, what emerges from this subgroup classification is a composition of justices more professionally experienced in defending (63.6%) and judging (54.5%) than in prosecuting (only 18.2%)¹⁶ , with most justices ascending through public competition to tenured positions (81.8%), while another group ascended through appointed positions (45.5%); the majority (63.3%) has experience defending private or partisan interests during most of their careers, and another share, adjudicating cases/disputes between the interested parties (54.5%). By and large, our data indicate that, in this period, the profiles least represented in the court are those of justices with experience in prosecuting, those who ascended through elected positions, those who achieved success in the private market, and those who are used to defending public interests. Next, we will see how these different profiles affected the ‘Mensalão’ trial and its final outcome.

Considered by many a watershed moment in the history of fighting corruption among the Brazilian political class, the ‘Mensalão’ case occupied the country's political scene for almost a decade – from the outset of the scandal in 2005 to the STF ruling on the last appeals in 2014. The Attorney General filed a criminal complaint against 40 defendants in the court in March 2006, but the defendants’ trial only began in August 2012.

In total, there were 106 ‘en banc’ decisions on the merits, involving 36 accused persons (divided into three groups or cells, called ‘nuclei’ (núcleos) by the Attorney General) and seven types of crimes¹⁷ . Decisions were reached in seven stages, and the average percentage of non-unanimous decisions was 57%¹⁸ . In five stages this was the approximate average. The other two stand out for their discrepant results: while the stage dedicated to money laundering by the advertising and financial cells of the scheme resulted in only 30% of non-unanimous decisions, the one dedicated to the charge of criminal conspiracy by the political, advertising, and financial cells of the scheme resulted in 92.3% of split decisions.

All in all, because of its characteristics, the ‘Mensalão’ trial offers an excellent opportunity to test hypotheses and theories. It involved a politically sensitive issue, it was under permanent media and public scrutiny, it was broadcast live by ‘TV Justiça’, in a court composed of justices appointed by presidents Sarney, Collor, FHC, Lula, and Dilma, a court exercising its original jurisdiction over the crimes of high-ranking authorities, with more than one hundred decisions keeping the plenary almost exclusively busy for the entire second half of 2012, and, finally, with a number of split decisions well above the average of what is found for the concentrated constitutional control cases. Because this is a criminal case involving leaders of the government that appointed a considerable number of justices to the court, we consider the ‘Mensalão’ trial a good test as it puts to work the justices' characteristics that were systematized in our typology regarding their different legal expertise, means of professional rise, and types of interests defended throughout their careers.

Also, although the ‘Mensalão’ trial had striking features, it was not an exception. FALCÃO et al. (2019) speak of a ‘Criminal Supreme’ (Supremo Criminal) today, referring not only to the court’s original jurisdiction over criminal cases, but also to criminal matters about which the STF has been discussing and deliberating according to the three main dimensions in which the court operates (constitutional, appellate, and original jurisdictions), which means that STF rulings range from extraordinary appeals to ‘habeas corpus’, writ of ‘ mandamus’, and cases of concentrated constitutional control ( FALCÃO et al., 2019 ). That is, this type of content has figured even in types of legal actions that are apparently unrelated to criminal matters.

It should be stressed that this article is not the first to use the ‘Mensalão’ decisions to test hypotheses and theories. Ferreira and Mueller (2014) , applying the W-NOMINATE technique, found a different decision-making pattern in the ‘Mensalão’ trial compared to that found by the same authors when analyzing concentrated constitutional control. In the ‘Mensalão’ trial, there was a clear split in the court between the justices appointed by Lula and Dilma and those appointed by previous presidents, a typical behavior according to the attitudinal model ( FERREIRA and MUELLER, 2014 , p. 292). Araujo (2018) also confirmed this pattern but revealed that the variation between justices was even more significant when comparing those appointed by PT governments before and after the scandal. Arantes (2018) shows how Teori Zavascki and Roberto Barroso – both appointed to the court after the merits judgments had been made – were decisive for reviewing sentencing and for reversing the conviction of all accused of criminal conspiracy, resulting in lighter sentences; at the end, the ‘Mensalão’ was a crime without an author, at least from a judicial point of view.

In this section, we sought to overcome previous studies, exploring the above-mentioned elements and variations in order to attest that the justices’ different career paths are reflected in their behavior in the court.

Graph 02 shows the judicial behavior in the STF in the ‘Mensalão’ trial – the decisions on motions for clarification and requests for reconsideration are excluded. We considered 106 decisions, totaling 1080 individual votes (see Harvard Dataverse¹⁹ ). Graph 02 clearly shows the court split in two: on one side, those with a more punitive behavior; on the other, those most inclined to lessen the sentences. Such a division, however, cannot be explained by the partisanship of the appointing presidents. There are four justices appointed by PT presidents on each side of this division, and the other justices appointed by non-PT presidents were not clustered on the same side. Suggestively, the rapporteur and the reviewer in this case had opposite behaviors and were both appointed by the same president. While Justice Joaquim Barbosa was the one who most voted to punish the defendants (87.7% of his votes were against the defendants), Ricardo Lewandowski was the one who least voted to convict (35.8% of his votes). The proportion of votes against the defendants in the court, indicated by the dashed line in the graph, was 66.7%.

Graph 02 Source: Created by the authors based on database available at Harvard Dataverse ˂https://doi.org/10.7910/DVN/N69VJP˃.

The variable partisanship of the appointing president is clearly insufficient to explain the individual behavior of the STF justices in the ‘Mensalão’ trial. As previously mentioned, we argue that elements in justices' career paths might work better as independent variables since justices develop distinct legal expertise as they get involved with various professional experiences, defend different interests, and ascend professionally through different means – aspects that can have a great influence on decisions in cases like this.

We used different logistic regression models to verify whether and to what extent the characteristics of the career paths of the STF justices influenced their decisions. Using logistic regression is appropriate because the dependent variable to be explained is binary. The votes in favor of convicting the defendants were codified with a value of ‘01’ and those in favor of acquittal, with a value of ‘0’. The independent variables of interest are those related to the type of previous legal expertise (in prosecuting, defending, or judging); the type of interests they used to defend (related to private/partisan interests, public interests, or adjudication of cases/disputes between the interested parties); the means of professional rise (through public competition, ‘cargos de confiança’ and other political appointments, elected positions, or the market); and experience in superior courts and high-level federal positions in Brasilia. Thus, we have 11 distinct independent variables to assess how each affects the judicial behavior of the 11 justices who participated in the main trial. In all models, we used control variables related to the cases adjudicated by the STF, more specifically, the type of crime and the item of the criminal complaint to which the rulings refer.

Graph 03 shows the results for four logistic regression models, each separately estimating the four main characteristics of the justices' career paths (repeating: means of professional rise, types of interests, types of expertise, and background in Brasília/superior court)²⁰ . Each point indicates the average marginal effect of the variables, in probabilities. The horizontal lines indicate the confidence interval, and the estimated coefficient is not statistically significant if the horizontal lines cross the dashed line. To the left of the line are the variables that decrease the likelihood of justices voting to convict the defendants, and to the right are those that increase that likelihood.

Graph 03 Source: Created by the authors based on database available at Harvard Dataverse ˂https://doi.org/10.7910/DVN/N69VJP˃.

Except for the variables associated with defending public interests and having judicial expertise in prosecuting, the other professional variables are statistically significant, with a relevant, substantial impact on the justices’ likely votes. Having held a judgeship in a superior court and having ascended professionally through the market or public competition are both associated with an increase in the probability of voting to convict the defendants. On the other hand, having held a high-level federal position in Brasília, having ascended professionally through ‘cargos de confiança’, having experience defending private/partisan interests and adjudicating cases/disputes between the interested parties, and, finally, having expertise in defending and judging are all associated with a tendency to acquit the defendants. The variables concerning judicial expertise and type of interest defended are equivalent because they correlate perfectly in this sample of justices²² . That is, justices who represented private/partisan interests also have expertise in defending. We should stress that justices who worked in superior courts necessarily have a background in Brasília, so the effect of one variable practically nullifies the other. Therefore, we must pay attention to those who have a background in Brasilia but have not held a judgeship in a superior court, that is, those who have held positions through political appointments, such as Gilmar Mendes and Dias Toffoli.

When estimated separately, the four dimensions we considered in justices' career paths have proven to be relevant for explaining the justices' individual behavior during the ‘Mensalão’ trial. Nevertheless, because of how these variables were created, they correlate with each other to some degree; therefore, to properly assess the effect of each separately, we should estimate a regression with all the characteristics simultaneously. In the complete regression, all variables remain in the same direction. This model can correctly classify 78.15% of the votes. Considering that a ‘naïve’ model – one that assumes that all votes follow the direction of the most frequent result (convicting the defendants) – would have correctly classified 66.75% of the votes, our results show that the logistic regression has significant explanatory power when it comes to the individual behavior of the justices.

Even though we found a reasonably stable result and were successful in classifying justices' votes, only professional rise through the market is a statistically significant variable in the regression with all variables. That is because logistic regression is sensitive to high multicollinearity between the variables²³ , which increases standard errors ( FERNANDES et al., 2020) . There are some ways to tackle this problem. First, we could increase the number of observations, which is not possible because we work with the total number of decisions in the ‘Mensalão’ trial. The second alternative would be to use a data reduction technique, such as principal component analysis ( FIGUEIREDO FILHO et al., 2015 ). Since we want to assess the effect of each of the characteristics separately, data reduction would not be an appropriate choice.

To address this issue, we used an algorithm ( Calcagno and Mazancourt, 2010 ) that estimates hundreds of thousands of distinct regressions with several possible combinations of variables. By doing so, we can verify to what extent the estimated coefficients for the career paths of the justices change according to whether they are excluded or included in the models. We can thus evaluate the extent to which the result we had previously found will change depending on the choice of variables in the regression model; we can also assess the extent to which our choice of a specific model is arbitrary or not, given that the results vary in the face of multiple possible choices. To better compare, we chose the models with the lowest BIC (Bayesian information criterion) and lowest AIC (Akaike information criterion), measures frequently used as performance criteria when selecting various models, especially to make predictions with more parsimonious models. We found four models with the same minimum BIC value and four with the same minimum AIC value, whose estimates are presented in Graph 04 .

Graph 04 Source: Created by the authors based on database available at Harvard Dataverse ˂https://doi.org/10.7910/DVN/N69VJP˃.

The difference between the models we selected is that those with the lowest BIC use only three variables simultaneously; all these models use the variable professional rise through the market. The models with lower AIC values use the same variables but add another one of less substantive impact: holding high-level federal positions in Brasília. The results indicate that the most important professional characteristics for assessing the justices' decision-making profiles when voting in the ‘Mensalão’ trial are professional rise through the market, ‘cargos de confiança’ or public competition, legal expertise in prosecuting, experience in defending public interests, and a background in Brasília. When compared to Graph 03 , only background in Brasília changes the direction of the substantive impact on the likely vote; at the same time, defense of public interests and expertise in prosecuting become more statistically significant.

With the results we have so far, we could contend that a justice more likely to acquit the defendants would have the profile of someone who ascended professionally not through the market and/or public competition but through ‘cargos de confiança’, who has no legal expertise in prosecuting, no experience in defending public interests, and who has never held a high-level federal position in Brasilia (in this last aspect, the effect is minimal). The justice who most voted to acquit in the ‘Mensalão’ trial, Ricardo Lewandowski, fits this profile perfectly. In the opposite direction, Joaquim Barbosa has four of the six characteristics associated with the punitive profile; he only did not hold a high-level federal position in Brasilia or ascended professionally through the market.

Our analysis was limited to the ‘Mensalão’ decisions made before the motions for clarification and the requests for reconsideration. Between one phase of the trial and the other, President Dilma Rousseff had the opportunity to appoint two justices to fill the vacated seats of Ayres Britto and Cezar Peluso, who had been appointed by Lula and were part of the most punitive coalition in the court during the ‘Mensalão’ trial. So, to further test our findings, we could ask: If the president were concerned with appointing justices in order to lessen the punishment of important politicians in her party (such as José Dirceu, José Genoíno, and João Paulo Cunha), appointing Luís Roberto Barroso and Teori Zavascki would have been the right decision, according to the results we found? As a simulation, to what extent could we have ‘predicted’ the votes of these justices based on the estimates of the previous logistic regressions?

These two justices nominated by Rousseff have most of the characteristics associated with acquitting the defendants. Although Teori Zavascki passed a civil service entrance examination at the beginning of his career, he joined a court of second instance (Tribunal Federal Regional da 4^a Região, TRF-4) through the ‘quinto constitucional’ provision and was later appointed to the Superior Court of Justice (Superior Tribunal de Justiça) by the president. During a significant part of his career, he developed expertise in defending and judging, but not in prosecuting; he also worked as an attorney defending private interests or as a judge judging cases between parties, but he did not work defending public interests. Luís Roberto Barroso had the stability associated with a tenured position assumed through public competition. He did not ascend professionally through a ‘cargo de confiança’ and had not held a high-level federal position in Brasilia. He developed expertise in defending and worked representing the interests of parties. Although both justices have the characteristics more associated with acquitting than with convicting, when comparing only the two, Barroso would have a more punitive behavior. We used our complete model to ‘predict’ the votes of these two justices and found that 86.3% of their votes were correctly classified. Both voted on the decisions about the 11 motions and requests, so our model got 03 votes wrong out of 22. In fact, one was Teori Zavascki's vote to convict Breno Fischberg for money laundering – our model had predicted he would vote to acquit. Our model correctly predicts Barroso's vote to convict in this trial but was wrong about two other votes. According to the model, Barroso would be more punitive: he would convict João Cláudio de Carvalho Genu and João Paulo Cunha also for money laundering, when in fact he voted to acquit.

Indeed, one can say Dilma Rousseff was very successful if she intended to appoint justices with a less punitive profile to minimize the punishment of highly important PT politicians (José Dirceu, José Genoíno, and Delúbio Soares) in the decisions on the motions and requests in the ‘Mensalão’ trial. Although there is no evidence that the president made such a calculation, by observing the career paths of the justices who had been less inclined to convict, she would have been able to see the path to appointing ‘supremables’ with similar characteristics who could change the trial's final outcome. Thus, choosing Barroso and Zavascki might not have been intentional, but the effects of these choices were not random, as the results of our analyzes demonstrate.

We conclude this section highlighting our contribution to the theories of judicial behavior as we offer a model that helps explain judicial decisions in the STF based on elements of justices’ career paths. Studies on judicial behavior are structured in two types of groups: a larger group of studies that first focuses on the characteristics of the decisions and/or the rules and constraints of the decision-making process to distinguish how judges decide, and a smaller group that avoids this circularity or endogeneity by starting from exogenous and ‘ex ante’ measures, from the justices before they join the high court and participate in the decision making. The well-known attitudinal model – whose premise is that judges decide guided by their sincere preferences – was introduced in the United States by Glendon Schubert (1965 , 1959 ) with studies of the first type ( SEGAL, 2003 ); however, the attitudinal model evolved and moved towards the second group, establishing itself as a model for predicting judicial decisions based on a previous characterization of justices – for example, the work of Segal and Spaeth (2002 , p. 321). In Brazil, as shown in this article, decisions have frequently been used to characterize judges; or, at most, presidential appointments have been taken as a proxy for justices' ideological positions. Our career path-based typology offers an external, ‘ex ante’ measure, and because it includes a wider range of elements to create different profiles, it also avoids simply dividing justices between ‘liberal or conservative’, ‘political or technical’.

This is a quantitative study, but the qualitative analysis of the ‘Mensalão’ trial carried out by Arantes (2018) shows how the justices – whose different profiles were operationalized here – mobilized typical elements of the legal models in their votes ( SEGAL, SPAETH and BENESH, 2005 , p. 22-34). Although the law constrains judges' behavior, they still have significant latitude in their decision-making. Since legal realism emerged at the beginning of the last century, scholars have been working to understand how personal characteristics affect judges' decisions ( GEORGE and WEAVER, 2017 ). Although we still need to further explain how certain professional experiences – and career paths more broadly – lead to specific types of decisions, the evidence we gathered suggests that scholars should take career paths into consideration, more specifically, the types of expertise, means of professional rise, and types of interests with which justices became familiar throughout their careers. Last but not least, our typology is in dialogue with the relational ( DRESSEL et al., 2018) and audience ( BAUM, 2006) models since in examining justices' biographies one can reveal the relational networks in which they forged their profiles, built their sense of (in)dependence, arrived in Brasília beforehand, became familiar with different types of interests to defend or settle, and earned the respect of groups – or even the public – whose interests they will seek to serve as STF justices.

The constitutional rule that grants presidents wide latitude in choosing new STF justices produces its effects. We identified distinct professional profiles of justices in the period we analyzed, which varied according to how much time of professional experience they had and what types of different careers they had pursued – some of a more law-centered nature, others overlapping with politics. Our analysis of the composition of the court during these years shows that some characteristics of the justices' professional profiles remain stable while others fluctuate. The proportion of judges with expertise in defending and judging, and who used to defend private/partisan interests or adjudicate cases/disputes between the interested parties remains the same. However, with the PT appointees in the court, the proportion of judges who ascended professionally through elected positions or who were not used to defending public interests decreased, while the number of those who ascended through public competition increased.

The variation we identified refers not only to the 35 profiles of justices we examined but to the fact that when presidents were able to choose two or more justices during their tenure, they did not choose candidates with the same profile. Lula, for example, chose a ‘pure-blood’ judge, Peluso, but also Eros Grau, who had been an attorney and law professor his entire career. Lula also appointed Dias Toffoli, who had been an attorney for the PT and Lula’s Solicitor General, but he also appointed the first black judge to the court, who would later become PT's tormentor in the ‘Mensalão’ case.

On the other hand, although the president has wide latitude to choose, this margin seems to shrink due to factors that we still do not know for sure how to identify. Over the period we analyzed, there were few cases of appointees who were personally or politically linked to the presidents, at least not directly linked. This suggests that appointments may be a two-way process: not only the presidents are making a choice, but a good number of candidates is also working hard throughout their careers to join the STF. No less than 24 of the 35 justices were already in high-level federal positions in Brasilia when they were appointed to the STF. In other words, they ascended through ranks until they reached the federal capital and got involved in politics, not only by holding elected positions or ‘cargos de confiança’ in the federal executive but also by engaging in legal world politics that took them to the superior courts. Except for Lula in his first term, the other presidents mostly appointed candidates who fit this profile.

A classification of career paths that categorizes judges as ‘technical’ or ‘political’ is thus too restrictive: The careers of justices are generally marked by the overlapping of law and politics throughout their lives. Even justices who spent most of their careers as judges, such as Lewandowski, Fux, or Rosa Weber, were not disconnected from politics: The first joined the São Paulo Court of Appeals (Tribunal de Justiça de São Paulo) through the ‘quinto constitucional’ provision after a major lobby of political parties and public administrators, and the other two were nominated to the Superior Court of Justice (Superior Tribunal de Justiça) and the Superior Court of Labor (Tribunal Superior do Trabalho), respectively, before joining the Supreme Court (in the case of Fux, after a blatant personal campaign to be appointed). In our typology, therefore, we sought to overcome this dichotomy in previous classifications and identify hypothetically relevant aspects for explaining justices' decisions. This typology might also be useful for future research on how presidents choose – and senators confirm – new justices.

By examining the decisions in the ‘Mensalão’ trial, this article revealed that certain characteristics in justices’ career paths have affected their behavior in the court, especially in the context of a criminal case. While we found that having ascended professionally through ‘cargos de confiança’ and other political appointments was associated with acquitting the defendants, ascending through public competition or the market, having expertise in prosecuting and experience in defending public interests, and having a background in Brasília (the latter, to a lesser extent) were associated with convicting. We have shown that career paths not only explain the justices' behavior but also predict quite successfully the behavior of those who joined the court later, when the motions for clarification and requests for reconsideration were being voted. We do not know whether presidents and the majority in the Senate know, but career-based choices could be more beneficial to them than choices based on personal or political ties. The cost is lower and future benefits may be higher and more predictable.

Dilma Rousseff did not have to turn to party allies to partially reverse the final outcome of the ‘Mensalão’ trial. In fact, she sought candidates with remarkable legal knowledge and an unblemished reputation who were aspiring candidates for a vacancy in the court. As Arantes (2018) demonstrated, given the legal inconsistencies in the Mensalão case, “the most political appointment Dilma could make was that of a candidate with a technical profile, with maturity and respectability in the art of due process” ( ARANTES, 2018 , p. 387). For it was Teori Zavascki who identified that the sentencing (more specifically, the amount of punishment (dosimetria)) was being manipulated and that there was an inconsistency in the defendants’ convictions for criminal conspiracy. With Barroso’s concurring vote, they gathered enough votes to review the sentence and acquit all defendants who had initially been convicted of criminal conspiracy. Thus, the defendants received lighter sentences based on legal justifications, not because of political appointments to the court.

Our results show that justices' career paths have tailored their decisions on the ‘Mensalão’ trial, contrasting those who voted more to convict to those voting more to acquit. Not by coincidence, in the context in which the ‘Lava Jato’ investigation is being reviewed and President Bolsonaro sees himself under pressure due to criminal allegations involving allies and members of his family, he and the Senate majority sought a candidate with a more ‘garantista’ profile – one who values due process more than punishing the defendants – to fill the vacated seat of Celso de Mello in 2020. Consistent with our findings, they chose someone who had developed expertise in defending and/or judging, but not in prosecuting; who ascended through ‘cargos de confiança’ and other political appointments, but not through tenured positions assumed via public competition; and who began his career defending private interests and later dedicated to adjudicating cases/disputes between the interested parties, but who did not defend public interests. Without ever holding high-level federal positions in Brasilia, his appointment was surprising since he was almost unknown; however, considering the above-mentioned characteristics, the appointment of Nunes Marques was not at all random, as it suits the preferences of the president and the Senate majority in that context.

Finally, this study opens at least four paths for future research dealing with the following topics and issues: 01. the use of the typology of career paths to examine decisions involving other sets of cases in the STF; 02. the use of this typology to analyze presidential appointments and confirmations by the Senate; 03. the effect of length of tenure on the characteristics forged in previous professional experiences, and, last but not least; 04. the continuous development of this typology based on career paths, an undertaking to which this article – neither the first nor the last – sought to contribute.