Three researchers have recently attempted to identify the factors that influence the decisions of the Supreme Federal Court (STF), the highest court of justice in Brazil. In search of answers to problems that have challenged specialists in the field for decades, they gathered information about 2 million individual votes and 423,000 decisions made by the STF since 1988 and created the most complete database on the topic ever built.
Composed of two law professors from INSPER in São Paulo, Diego Werneck Arguelhes and Ivar Hartmann, and political scientist Evan Rosevear, of the University of Southampton, in the UK, the trio set out to separate the most controversial cases to better observe the differences among the Court’s 11 justices. The researchers therefore focused on the 8,000 cases that caused the greatest division within the Court, in which at least two justices went against their colleagues.
Statistical techniques were used to assess the effects associated with the personal characteristics of the magistrates, such as career trajectory and gender, Court rules, and other factors that could influence the outcomes of the rulings, without considering interpretations of the law and the Court’s jurisprudence. For example, they confirmed a longstanding hypothesis about the impact of publicizing the STFs deliberations, whose plenary sessions began being broadcast live on television and YouTube in the early 2000s. The analysis showed that disagreements are more frequent in the rulings broadcast on TV Justiça.
Evidence emerged that the career trajectories of the justices also matter. Data revealed that those who served as judges in other courts before joining the Supreme Court tend to disagree more. This behavior goes against expectations created by studies examining constitutional courts in other countries. The study found a deviation in full bench rulings of the STF, where all 11 justices vote, but not in those of the two panels, where the justices are divided and where the majority of cases are judged.
Bruno Moura / SCO / STFA plenary session of the Court in FebruaryBruno Moura / SCO / STF
There are also indications that issues of gender affect the daily operations of the Court. The trio discovered that women disagree less than their male colleagues, and they dissent less in cases where the rapporteur is a woman. However, a study conducted by Werneck and other researchers last year revealed that men frequently interrupt their female colleagues during the Court’s plenary sessions. As only three women have reached the STF to date, and only one remains there, there is a lack of information to better understand this dynamic.
“Explaining the behavior of the justices of a court such as the STF is challenging, because there are many factors in play,” says Werneck. “We were able to test a large number of hypotheses with a broader set of data than those used in previous studies, but there is still much work to be done to understand what the data reveals to us.” A scientific article with the trio’s initial findings was published last year in the Journal of Law & Empirical Analysis.
Researchers dedicated to studying the Judicial Branch have increasingly invested in quantitative surveys like this to examine the workings of the Supreme Court, which gained strength after the 1988 Constitution was promulgated and assumed a central role in Brazil’s political system. They argue that only this way will it be possible to go beyond the theoretical assumptions about the actions of the Court’s justices and uncover patterns that explain their decisions.
The STF is a challenging subject to study due to the multiple functions it performs and the heavy workload associated with them. The new Constitution expanded the rights and guarantees of the citizens, which increased the demand of the Judiciary Branch and the number of cases in the STF, where appeals against rulings from lower courts are heard. The Constitution also expanded access to the Court by enabling political parties, professional associations, and other organizations to appeal to it in order to challenge the constitutionality of laws and acts enforced by the Executive and Legislative branches. Additionally, the STF is responsible for investigating, prosecuting, and judging politicians and hundreds of other authorities who have the right to special jurisdiction.
More than 40,000 new cases were assigned to the justices of the Supreme Court last year alone, a period during which they issued 115,000 decisions (see graph). Four out of every five were issued individually, without the participation of the other justices, according to the statistics available on the Court’s website. In contrast, the Supreme Court of the USA decides on fewer than a hundred cases a year, selected from among thousands of petitions and the Federal Constitutional Court of Germany rules on fewer than 6,000 cases.
The interest in quantitative studies is a natural result of the large number of cases but also reflects the concerns scholars have about the individual power of the members of the STF, without parallel in the courts of other countries. “Our justices have become major political players, and we need to better understand how they exercise their powers, because it affects the legitimacy of the Court,” assesses political scientist Rogério Bastos Arantes, of the School of Philosophy, Languages and Literature, and Human Sciences of the University of São Paulo (FFLCH-USP) and coordinator of the Judiciary and Democracy study group.
In the first years under the 1988 Constitution, researchers in the field prioritized examining the Court’s response to the flood of new cases. Studies conducted during this period showed that the Supreme Court exercised its veto power over the actions of the other branches on few occasions, contrary to what experts had expected. Over time, Congress expanded the powers of the STF, creating instruments for the Court to enforce its rulings on other levels of the Judiciary. As it became more assertive, academic attention shifted toward the individual behavior of the justices and their strategic calculations.
Sociologist Fabiana Luci de Oliveira, of the Federal University of São Carlos (UFSCar), was one of the first to go in this direction. In studies published since 2002, she analyzed individual votes in hundreds of constitutional challenges and gathered evidence that political preferences and career trajectories influenced the justices. “In some cases, they expressed concern during their own votes about the political repercussions of their decisions for the government and the Court’s image,” says Oliveira.
Another important step was made by the Getulio Vargas Foundation (FGV) Law School in Rio de Janeiro in 2010, when the Supreme in Numbers project was launched, led by constitutional law professor Joaquim Falcão, who is now retired. The research group, in which Werneck, Hartmann, and other researchers participated, was the first to develop a comprehensive database on the Court’s decisions. The findings were released in annual reports published through 2020.
Conducting these surveys requires months of work. Most of the information is on the STF website and can be extracted using data scraping tools, which are computer programs that automatically copy the available data. The researchers from INSPER also had to create an algorithm to determine the vote of each justice in the gathered cases, because the summaries on the Court’s website only indicate which justices dissented and do not name those who aligned to form a majority. Keeping the database up to date requires repeating several procedures.
Nelson Jr. / SCO / STFJustice Rosa Weber, who retired that yearNelson Jr. / SCO / STF
The group led by Rogério Arantes at USP hopes to add new impetus to studies in the field this semester, when it is expected to provide researchers with a database full of information on all the cases available on the STF website, including party names, case movements, and decision summaries. In a second phase, the group intends to include in the database information from the Court’s virtual plenary sessions, in which there is no interaction between justices and where most decisions are currently made.
Despite the expectations created by the advance of quantitative studies, many researchers in the field see limits that not even the most comprehensive database could overcome. “The Supreme Court is an extremely complex and idiosyncratic court, with frequent oscillations of its positions on important issues,” says Eloísa Machado de Almeida, coordinator of the Supreme Court on the Agenda group at the FGV Law School in São Paulo. “Statistics do not capture the nuances that can only be understood by following the sessions and reading the votes.”
For Virgílio Afonso da Silva, constitutional law professor at the USP School of Law, empirical research has contributed to a more sophisticated understanding of the workings of the STF. He notes that the excess of information forces researchers to make choices to separate relevant cases from routine ones in the mountain of proceedings the Court handles, but this can also introduce undesirable biases into their analyses.
Separating cases that were not decided unanimously, as the researchers from INSPER did, or focusing on those that attract media attention, as others suggest, are alternative strategies that may exclude important cases for understanding the Court. “There are unanimous decisions and ones off the media radar that are also relevant,” says Silva. “We are still at a crossroads, searching for criteria to interpret the data, and it seems difficult to draw overly definitive conclusions.”
Between 2011 and 2013, as part of a research project on the Court’s deliberative process, which had funding from FAPESP, Silva interviewed 17 justices, including current and retired members of the STF. Several suggested, among the factors that influenced their decisions, elements that empirical studies struggled to capture, such as their concerns about criticism from academia and the economic impact of the Supreme Court’s decisions.
The emergence of the virtual plenary as the location where most decisions are made has created further complications. On the platform, the justices can submit written votes or indicate whether they support the opinion of the rapporteur of the case or that of another colleague. There are no debates, but any one of the justices can interrupt the judgment by requesting a review to examine the case more thoroughly, or requesting the discussion be moved to an in-person plenary session, which considerably changes the calculations justices usually make when deciding.
Carlos Humberto / SCO / STFIndigenous people follow the judgment of the temporal framework thesis, in 2023Carlos Humberto / SCO / STF
The most difficult has been identifying the reasons behind the justices’ calculations. In the USA, where many of the studies that influence experts in this field were developed, research points to the political and ideological preferences of Supreme Court justices as determining factors for explaining their decisions and finds a strong association between the judges’ preferences and the parties of the presidents who nominated them. However, it has been difficult to replicate this result in Brazil, because of the differences between the political systems of the two countries.
In a study published two years ago in Brazilian Political Science Review, Rogério Arantes and political scientist Rodrigo Martins, from USP, proposed a typology to analyze the professional trajectories of the justices and verify whether these influenced their decisions. To test the hypothesis, they examined a hundred decisions made during the judgment of those involved in the Mensalão scandal (a political bribery scandal), in 2012, and they concluded that the professional trajectories of the justices were more decisive than the party affiliations of the presidents who nominated them.
Career judges were more inclined to convict them, according to the researchers. The eight justices appointed by presidents affiliated to the Workers’ Party (PT) who participated in the judgment were split evenly between those who applied severe punishments and those that were more lenient toward the accused. Three aligned themselves with Justice Joaquim Barbosa, who was the rapporteur of the case and voted to convict the defendants in 88% of his decisions. The others voted for convictions less than 60% of the time, aligning themselves with Ricardo Lewandowski, the case reviewer, who voted to convict in only 36% of his decisions. President Luiz Inácio Lula da Silva nominated both.
“If the developed model makes sense for analyzing judicial behavior, it could also be useful for understanding the choices that presidents make when the Court has vacancies,” states Arantes. “Unlike in the USA where the choices are more clearly aligned with a single party, the presidents in Brazil shy away from clearly partisan options and look at the professional trajectory of the candidates for justices, seeking elements that help them reach majority approval in the Senate and, who knows, influence the direction of the Court.”
Fellipe Sampaio / SCO / STFJustice Luiz Fux presides over a plenary session via videoconference in 2021Fellipe Sampaio / SCO / STF
In his PhD thesis defended at the Federal University of Minas Gerais (UFMG) in 2024, Shandor Torok Moreira, Attorney General of the state of Mato Grosso do Sul, proposed an alternative model to determine the effect of the political preferences of the justices of the STF. In the research, he selected 465 cases judged between 2010 and 2018 and divided the decisions into two groups, classifying them as conservative or progressive. Next, he verified how the justices positioned themselves in each situation and established their ideological preferences.
To classify the decisions, the researcher analyzed the issues being discussed in the cases and adopted different criteria. In criminal proceedings, for example, decisions in favor of the defendants were classified as progressive and the rest as conservative. Decisions in favor of vulnerable groups, consumers, and unions were considered progressive. In tax disputes, votes in favor of companies and taxpayers were labeled as conservative.
According to the thesis, many justices tend to align themselves with conservative positions, some of them in over 60% of the cases, but the results also suggest that the majority have a moderate profile, oscillating between different ideological orientations depending on the issue under discussion. “The worldview of the justices is one of the factors that influences their decisions, and this helps make their behavior more predictable,” affirms Moreira.
Some researchers dedicated to the topic are hopeful that the statistical models used to interpret the new databases could help predict the decisions of the STF, which would assist attorneys and their clients in better assessing the costs and risks of a case in the Court. But there are also limits to this area. “Algorithms exist capable of predicting decisions with high accuracy rates, but none are able to predict the outcome of the most complex and important cases,” states Hartmann, from INSPER.
Carlos Humberto / SCO / STFA session in 2012 that kickstarted the judgment of the Mensalão caseCarlos Humberto / SCO / STF
In 2002, US political scientists Andrew Martin, of Washington University in St. Louis, and Kevin Quinn, of Harvard, partnered with two law professors, Theodore Ruger, of the University of Pennsylvania, and Pauline Kim, also of Washington University in St. Louis, to try to predict the outcomes of all the cases heard by the US Supreme Court during the course of a year. The first two used a statistical model for their predictions, while the other two relied on the expertise of a group of specialists including attorneys and professors.
The results of the experiment were mixed: The political scientists successfully predicted the outcomes of 75% of the 68 cases analyzed, while the legal experts achieved a success rate of 59%. In the prediction of the individual votes of the justices, the two teams tied. The statistical model correctly predicted 67% of the votes and the legal experts managed 68%. The experiment from 2002 was not repeated, but other researchers have developed new algorithms since then, achieving higher accuracy rates. So far, nobody has attempted anything similar in Brazil.
The story above was published with the title “The minds of judges” in issue in issue 350 of april/2025.
Project
Deliberative practice at the Supreme Federal Court (nº 11/01066-0); Grant Mechanism Regular Research Grant; Principal Investigator and Beneficiary Luís Virgílio Afonso da Silva (USP); Investment R$20,414.79.
Scientific articles
ARANTES, R. B. & MARTINS, R. Does the before influence the after? Career paths, nominations, and votes of the STF Justices. Brazilian Political Science Review. Vol. 16, no. 3. 2022.
ARGUELHES, D. W. et al. “They don’t let us speak”: Gender, collegiality, and interruptions in deliberations in the Brazilian Supreme Court. Journal of Empirical Legal Studies. Vol. 21, no. 1. 2024.
DA ROS, L. et al. Do presente ao futuro: Novas agendas de pesquisa sobre o Supremo Tribunal Federal. Política & Sociedade. Vol. 21, no. 52. 2022.
OLIVEIRA, F. L. de. Quando a Corte se divide: Coalizões majoritárias mínimas no Supremo Tribunal Federal. Direito e Práxis. Vol. 8, no. 3. 2017.
ROSEVEAR, E. et al Dissenting votes on the Brazilian Supreme Court. Journal of Law and Empirical Analysis. Vol. 1, no. 2. 2024.
SILVA, V. A. da. Pauta, público, princípios e precedentes: Condicionantes e consequências da prática deliberativa do STF. Supreme. Vol. 1, no. 1. 2021.
Republish
