The CJEU Database Platform: Decisions and Decision-Makers

Abstract

This article presents the CJEU Database Platform, which provides scholars with an extensive collection of easily accessible, research-ready data on the the universe of cases, decisions, and judges at the Court of Justice of the European Union (CJEU). The CJEU Database Platform provides a foundation for the broader CJEU Database currently being developed by The IUROPA Project, a multidisciplinary group of scholars researching judicial politics in the European Union (EU). In this article, we illustrate how the CJEU Database Platform opens the door to new areas of theoretical and empirical research on judicial politics in the EU.

The judicialization of politics and the rise of courts as important allocators of societal values have repeatedly been identified as key features of the 21st century (Vallinder 1994; Ferejohn 2002; Hirschl 2004; Stone Sweet 2004; Alter 2014). Judicialization is an international trend, but anyone who surveys the literature on judicial politics will be struck by the overwhelming dominance of research originating in — and focusing on — the United States. For example, a recent handbook on judicial behavior includes nineteen chapters exclusively concentrating on US courts, five chapters on comparative research, and three chapters on international courts (Howard and Randazzo 2017).

One of the main reasons why research on European courts is lagging behind research on American courts — besides different traditions with respect to law and politics on the two sides of the Atlantic (Kelemen 2011) — is the lack of accessible multi-user databases that provide scholars visibility into the decision-making of European courts.¹ The US Supreme Court Database, for instance, is one of the most successful social science databases and facilitates quantitative research on judicial politics in the United States (Epstein et al. 2015). There is no equivalent for the main court of the European Union, the Court of Justice of the EU (CJEU). This article introduces The CJEU Database Platform: Decisions and Decision-Makers, which aims to fill that gap.

An easily accessible and comprehensive database of CJEU decisions and decision-makers is critical to the development of the empirical literature on the judicial politics of the EU. There has been an increase in the availability of data on the CJEU in recent years, but these data-collection efforts have been overlapping, uncoordinated, and not always easily accessible. The CJEU Database Platform provides a central platform for data on the CJEU and data from this database can easily be integrated with existing datasets on EU law and politics, such as Derlén and Lindholm (2014), Carrubba and Gabel (2015), Larsson and Naurin (2016), Fjelstul (2019), and Ovádek (2021).

By providing data on the key features of cases, decisions, and decision-makers at the CJEU, which is the court at the heart of the process of judicialization in Europe (Alter 2010), we aim to improve the opportunities to study a range of questions central to the field of judicial politics. We believe this will facilitate better empirical research on the CJEU and open the door to studying new questions about legal mobilisation in the EU (e.g., Hofmann and Naurin 2020; Pavone 2022), judicial bargaining and the internal politics of the Court (e.g., Frankenreiter 2017, 2018; Cheruvu 2019; Hermansen 2020; Fjelstul 2022), the politics of judicial appointments (e.g., Bobek 2015), the sources and impact of judges’ attitudes (Vauchez 2012; Larsson and Naurin 2019), and the sensitivity of the Court to its political and social environment, including the politics of noncompliance and legislative override (e.g., Carrubba, Gabel, and Hankla 2008; Alter 2010; Carrubba and Gabel 2015; Martinsen 2015; Larsson and Naurin 2016; Larsson et al. 2017; Fjelstul and Carrubba 2018; Blauberger and Martinsen 2020). Previous research on the CJEU has long debated the power and autonomy of the Court in relation to the EU member states, but theory development has been hampered by the treatment of the Court as a unitary actor. The CJEU Database Platform provides opportunities for scholars to make progress by combining data on the external relations, internal processes, and members of the Court.

An important obstacle to research on the decision-making of the CJEU is the fact that the Court does not publish judges’ votes. The Court conducts its deliberations in secret, and no votes or dissenting opinions are ever published. This means that it is difficult to estimate the ideological positions of individual judges. As a result, the CJEU has almost invariably been studied as a unitary actor (Pollack 2003, 2013).² However, the CJEU does publish information on its individual members that, with some creativity, can be used by scholars to address core research questions relating to judicial behavior. The Court’s database of cases (InfoCuria) and the EU’s database of legal documents (EUR-Lex) include information on chamber presidents, Judge-Rapporteurs, Advocates General (AGs), and the composition of chambers. Furthermore, the Court provides background information about the previous careers of judges and AGs. The opportunity to combine this individual-level information with data on the Court’s internal procedures, and with other information that connects the Court’s decisions to features of the social, political, and legal context in which it is situated, in our view, deprives scholars of any remaining excuses for not studying central questions to judicial politics in the European context.

The CJEU Database Platform is the foundation of a broader CJEU Database currently being developed by The IUROPA Project — a multidisciplinary group of scholars promoting research on judicial politics in the EU.³ The longer-term objective of The IUROPA Project is to help scholars improve on the state of empirical research on the CJEU by building a comprehensive CJEU Database centered around the CJEU Database Platform, including hand-coded and machine-coded measures of legal issues, positions, outcomes, legal arguments, topics, and a complete corpus of CJEU decisions (in French and English) that scholars can use for text analysis.

The CJEU Database Platform covers the CJEU’s entire history (1952–present) and includes data on the universe of cases and decisions at the three courts that have made up the CJEU: the Court of Justice (1952–present), the General Court (1989–present), and the Civil Service Tribunal (2005–2016). Our database currently includes nine datasets containing information on: (1) the universe of CJEU cases (cases); (2) the universe of CJEU proceedings, which are individual or joined cases that can result in decisions (proceedings); (3) the universe of decisions issued by the CJEU, including judgments, orders, and AG opinions (decisions); (4) data on the parties in each proceeding (parties); (5) the disposition of each decision with respect to each legal procedure, if applicable (procedures); (6) the composition of the panel that issued each decision (assignments); (7) third-party observations and interventions with respect to each decision (submissions); (8) citations in each decision to case law, the EU Treaties, and EU legislation (citations); and (9) the judges and AGs who have been appointed to the Court (judges) — the key information scholars need to study the decision-making of the CJEU.

In order to understand how the CJEU Database Platform can be used for empirical research, scholars also require knowledge of the organizational structure and internal procedures of the Court. Next, we introduce the datasets in the CJEU Database Platform and discuss how the datasets capture the key aspects of the Court’s internal procedures. Then, we provide empirical applications to illustrate what we can learn about decision-making at the CJEU using the data and how the data can facilitate empirical research on the Court and the development of EU law.

The CJEU Database Platform

In this section, we introduce the datasets in the CJEU Database Platform, review the Court’s internal procedures, and show how the datasets capture theoretically interesting institutional activity.⁴ We also discuss recent data-collection efforts, identify the gaps, and illustrate how the CJEU Database Platform will help to facilitate research on the decision-making of the Court. Our purpose is not to describe the existing literature in detail, but rather to illustrate how the CJEU Database will help scholars advance the literature. For recent reviews of the empirical literature on the CJEU, see Blauberger and Schmidt (2017) and Carrubba and Fjelstul (2021). For more information on the Court’s internal procedures, see Wägenbaur (2013) and Lenaerts, Maselis, and Gutman (2014).

The CJEU Database Platform contains nine datasets, each with a different unit of observation: cases, proceedings, decisions, parties, procedures, assignments, submissions, citations, and judges. Table 1 summarizes the contents of the CJEU Database Platform. The Supporting Information includes a codebook. As we discuss the Court’s internal procedures, we indicate which datasets include relevant data. To create the database, we have compiled data and metadata from the Registry, InfoCuria, and EUR-Lex (the three official data sources for CJEU cases and decisions) using a supervised, automated process. Each of these official sources is incomplete and contains errors. We use hand coding to fill in missing data, and we cross-validate the data wherever possible to create the most complete and most accurate collection of CJEU data currently available.⁵ The CJEU Database Platform is updated regularly as new cases are lodged, new decisions are published, and new judges and AGs are appointed to the Court.

Table 1. Overview of the CJEU Database Platform

Note. The CJEU Database Platform is updated and expanded regularly.

The CJEU is composed of the Court of Justice and the General Court. The European Court of Justice (ECJ), the precursor to the modern CJEU, was created in 1952 by the Treaty of Paris. In 1989, in response to the Court’s growing workload, the Court of First Instance was created to handle some of the more fact-intensive and technical cases. Then, in 2005, the Civil Service Tribunal was created to hear disputes between EU civil servants and the EU institutions, which are called staff cases. Staff cases had previously been heard by the Court of First Instance, and before that, by the Court of Justice. In 2009, the Treaty of Lisbon reformed the structure of the Court. The Court of First Instance was renamed the General Court, the ECJ was renamed the Court of Justice, and all three courts together were called the CJEU. The Civil Service Tribunal was dissolved in 2016, and staff cases are now heard by the General Court again.

We focus our attention in this section on the internal procedures of the Court of Justice, rather than the General Court, although there are many similarities. However, we emphasize that the CJEU Database Platform includes data on all cases, decisions, and judges at the General Court and the Civil Service Tribunal as well.

Cases arrive at the CJEU in several ways. First, national courts can refer questions about the application of EU law to the CJEU in references for a preliminary ruling (or just preliminary rulings). In a preliminary ruling, a national court asks the Court how to interpret EU law. The Court answers the question and remands the case back to the national court for a final decision. The national court is obligated to follow the Court’s ruling. Second, direct actions are cases that are filed by applicants. Third, the CJEU has jurisdiction in staff cases (which are rarely as theoretically interesting as other types of cases). Fourth, parties can appeal decisions of the General Court to the Court of Justice on points of law. The cases dataset includes information on the universe of CJEU cases, including all decisions issued by the Court that are associated with the case.

There are four major types of direct actions: (1) actions for failure to fulfill obligations (also called infringement cases), where the European Commission sues a member state for violating EU law (Article 258 TFEU); (2) actions for annulment, where an applicant sues an EU institution to annul a measure adopted by that institution (Article 263 TFEU); (3) actions for failure to act, where an applicant sues an EU institution for failing to take action (Article 265 TFEU); and (4) actions for damages, where an applicant sues an EU institution for damages (Article 268 TFEU). Infringement cases are always heard by the Court of Justice, but other direct actions are usually heard at first instance by the General Court. We refer to these four types of direct actions, along with preliminary rulings, staff cases, and appeals, as legal procedures. Cases occasionally involve multiple legal procedures. There are also a variety of minor legal procedures (e.g., applications for measure of inquiry, applications for interim measures, applications to intervene, etc.), most of which only occur in combination with one of the major procedures. The procedures dataset indicates all of the legal procedures associated with each judgment.

The Court frequently joins similar cases together, so we make a distinction between cases and proceedings. A proceeding can refer to a group of joined cases or a single un-joined case that can lead to a decision. When cases are joined, the Court uses the case ID of the earliest case lodged to refer to all of the joined cases. In other words, the earliest case ID becomes the primary case ID for the group of joined cases. This can be confusing because then that case ID refers to both a single case and a group of joined cases. To avoid confusion, we assign a proceeding ID to all cases. All cases that are joined together will always have the same proceeding ID. For joined cases, the proceeding ID is the case ID of the earliest case. For stand-alone cases, the proceeding ID is the same as the case ID. The proceedings dataset includes one observation per unique proceeding. Every decision in the decisions dataset is associated with a single proceeding ID. The decision dataset lists the cases that each decision applies to, allowing researchers to handle joined cases appropriately, based on their research question.

The Registry receives the procedural documents from the applicant in direct actions, or from a national court in references for a preliminary ruling, and assigns a date of lodgment to the case (RoP, Art. 57).⁶ The Court then publishes a notice in the Official Journal of the European Union (RoP, Art. 21). This marks the starting point of the written procedure, which consists of the submission of all applications, statements of case, defenses, observations, and replies of parties and third parties involved in the dispute. The decisions dataset includes the date of lodgment, allowing researchers to calculate the duration of proceedings that result in a decision. The parties dataset includes information about the parties (i.e., the applicant and the defendant) in each proceeding.

In preliminary rulings, within two months of the notification, the parties to the case, the member states, the European Commission, and other relevant actors can submit written observations to the Court (Statute, Art. 23).⁷ Carrubba and Gabel (2015) and Larsson and Naurin (2016) use data on observations coded at the level of legal issues. These datasets have the advantage of including which position on the legal issue each observation supports, but the samples do not cover decisions after 2008. The submissions dataset includes information on all observations submitted to the Court.

In direct actions, the member states, the Commission, and other interested parties can intervene in support of one of the parties (Statute, Art. 40). An application to intervene must be submitted within six weeks of the publication of the notice in the Official Journal. The intervention must be limited to supporting, in whole or in part, the form of order sought by one of the parties (RoP, Art. 129). The submissions dataset also includes information on all interventions in direct actions, including data on what kind of actor is intervening (a member state, a company, an individual, etc.) and which party (application or defendant) the intervener is supporting, if available.

After a case has been lodged, the President of the Court designates a judge to act as Judge-Rapporteur in the case (RoP, Art. 15). This is the most important task of the President with respect to the judicial business of the Court (interviews). The President also represents the Court, ensures the proper functioning of the services, and presides over hearings and deliberations of the Grand Chamber, one of the formations of the Court (RoP, Art. 9). The Judge-Rapporteur plays a key role in each case. They manage the case from the moment they are appointed by the President. The Judge-Rapporteur is instrumental both to several procedural steps in the case management process (RoP, Art. 59) and as an agenda-setter in the deliberations (interviews). The decisions dataset indicates the Judge-Rapporteur for each judgment.

The judges elect the President and the Vice-President of the Court, by majority vote and secret ballot, for three-year renewable terms (Statute, Art. 16; RoP, Art. 8). They also elect the presidents of the chambers for a once-renewable three-year term (Statute, Art. 16, RoP, Art. 12). These positions are highly prestigious. They come with specific powers, such as the President’s prerogative to appoint a Judge-Rapporteur and to chair the Grand Chamber deliberations, where the most important proceedings are decided. The presidents of chambers also have specific procedural powers relating to the cases that are heard in their chamber, such as the authority to decide when to end the deliberations and call for a vote to decide the disposition of the case (interviews). The status of the position as President of a five-judge chamber was further enhanced in 2003 when, after a revision of the Rules of Procedure, they were made subject to elections among the judges. The Presidents of the five-judge chambers also have a seat in many Grand Chamber cases.

In addition to one judge from each member state, the Court of Justice employs 11 Advocates General (AGs), with similar status as the judges. The larger member states can nominate one AG each, while the smaller states nominate AGs on a rotating basis. The role of the AG is to provide an independent — and public — opinion before the judges start deliberations. An AG is appointed for each case by the First Advocate General. The position of First Advocate General rotates on a one-year basis (RoP, Art. 14; interviews). As soon as a case is lodged in the Registry, the First Advocate General assigns an AG (RoP Art. 16). The decisions dataset indicates the AG for all decisions.

Data on which AGs participated in which proceedings will help researchers investigate the extent to which AGs influence the Court’s rulings (e.g., Sadl and Sankari 2017) and the extent to which AGs opinions display political leanings (e.g., Frankenreiter 2018). There is a debate in the literature about whether AG opinions reflect the legal merits of a case (Carrubba and Gabel 2015; Frankenreiter 2018), which can be an important control variable in empirical analyses. There are also differences in citation patterns between AG opinions and judgments in the same proceeding, but it is not yet clear the extent to which these differences indicate a difference in legal reasoning or political preferences between the AG and the judges in the chamber (Fjelstul 2019). The citations dataset includes citations in CJEU decisions, including AG opinions and judgments.

When the written part of the procedure is closed, the President fixes a date on which the Judge-Rapporteur is to present a preliminary report at a General Meeting of the Court (RoP, Art. 59). The General Meeting is the central decision-making body of the Court, where all judges and AGs meet and have a vote (RoP, Art. 25; interviews). The preliminary report contains a number of proposals relating to important parts of the procedural management of the case. The Judge-Rapporteur consults with the AG on these proposals before presenting them at the General Meeting (RoP, Art. 59). The preliminary report is classified as an internal document (and is not made public), but several of the key decisions taken by the judges at the General Meeting on the basis of the report leave traces in the Court’s records of cases and decisions, which the CJEU Database Platform captures.

First, the preliminary report suggests whether further measures of inquiry, or requests for clarification from the national court (in preliminary rulings), should be undertaken. The procedures dataset includes information on all legal procedures associated with each decision, including applications for measures of inquiry.

Second, the Court decides whether or not there should be an oral hearing. The Court can decide not to hold a hearing if it considers, after reading the written pleadings or observations lodged during the written part of the procedure, that it has sufficient information to give a ruling (RoP, Art. 76). A member state that is party to the proceedings has the right to request an oral hearing (RoP, Art. 76; Statute, Art. 23) and that a case is heard in the Grand Chamber (Statute, Art. 16). The decisions dataset indicates any hearings that were held (and the dates) before issuing the decision.

Third, the Court decides, on the basis of the preliminary report, whether to dispense with the AG opinion. If the Judge-Rapporteur, after having consulted the AG, considers that the case raises no new points of law, they may propose that the Court issue a decision without an opinion from the AG in order to speed up the judicial process (Statute, Art. 20). The decisions dataset includes all AG opinions and indicates the proceeding that each is associated with.

Fourth, when the Judge-Rapporteur presents the preliminary report, the Court decides which formation of the Court will hear the case. This is a significant decision, affecting both which judges get to decide a case, and signaling how salient the Court thinks the case is. The Court can decide to assign a case to a three-judge chamber, a five-judge chamber, or a 15-judge chamber (the Grand Chamber), or in cases of “exceptional importance,” the full Court (i.e., all 27 judges) (Statute, Art. 16).⁸ We distinguish between chambers (standing groups of judges) and panels (the set of judges who actually participate in a proceeding) because more judges can be assigned to a chamber than actually participate in a case and judges sometimes substitute for each other. When the Court decides to assign a case to a three-judge or five-judge chamber, the chamber that hears the case is the one that includes the Judge-Rapporteur (judges can be attached to more than one chamber at the same time), so by designating the Judge-Rapporteur, the President has influence over the composition of the panel of judges that participate in the case.

The decisions dataset indicates which formation of the Court (e.g., First Chamber, Second Chamber, Grand Chamber, etc.) delivered each judgment and who the Judge-Rapporteur was. The assignments dataset indicates which judges were on the panel that delivered each judgment. To complement this data on judge assignments, the judges dataset includes information on the professional experience of each judge (i.e., prior to joining the CJEU, whether the judge had experience as a judge, academic, civil servant, practicing lawyer, or politician), which will help scholars study how judge-specific characteristics influence the Court’s behavior.

The chamber system is critical to empirical research about judicial bargaining at the Court. Unlike other courts that political scientists are interested in, like the US Supreme Court or the European Court of Human Rights (ECtHR), the CJEU does not publish voting data and judges do not write dissents or publicly talk about the internal deliberations of the Court (Fjelstul 2022). Thus, we cannot use any of the methods for scaling the preferences of judges that have been developed in the American context (e.g., Martin and Quinn 2002; Clark and Lauderdale 2010; Lauderdale and Clark 2012, 2014, 2016).

Malecki (2012) develops a Bayesian item response theory (IRT) model that leverages the constant rotation of judges in chambers to estimate the preferences of judges in preliminary rulings, but this approach relies on additional hand-coded data (on third-party interventions) for identification that makes the model unsuitable for more recent years, where such data is not available. Developing methodologies for estimating the preferences of judges in the context of the chamber system is critical for opening the door to better empirical research on judicial decision-making at the CJEU. The database provides scholars with the data they will need to develop and evaluate such methodologies.

After the Court assigns a proceeding to a chamber, the oral procedure begins. During the oral procedure, the Court hears from agents, lawyers, witnesses, and experts. It concludes with the submission of the AG opinion (Statute, Art. 20; RoP, Art. 82), if applicable, at which time deliberations can begin between the judges. The Judge-Rapporteur is responsible for producing the first draft of the judgment, which is distributed to and commented on by the other judges (with the assistance of the legal secretaries in their cabinets). Oral deliberations between the judges are conducted in French, without any interpreters present. Cheruvu (2019) shows that the Court’s use of French as the working language matters, with Francophone judges working faster than non-Francophone judges. The president of the chamber decides when to end the deliberations, at which point the judges vote, although anecdotally, judges usually try to work by consensus. The final judgment is written by the Judge-Rapporteur, even if they are in the minority (interviews). The decisions dataset includes all judgments of the Court.

Scholars have coded the outcomes of preliminary rulings in the form of positions on legal issues through 2008 (Carrubba and Gabel 2015; Larsson and Naurin 2016), but there is no research-ready data on the outcome (i.e., the disposition) of all of the other legal procedures, including direct actions, which constitute the majority of proceedings. The Court can rule that an application is successful, unfounded, or inadmissible.⁹ These categories are not mutually exclusive. For example, an application could be partially successful and partially unfounded. When a proceeding involves multiple legal procedures (e.g., an action for annulment and an action for damages), the Court might rule that the applicant is successful with respect to one legal procedure but not others. Data on judicial outcomes in direct actions is important for measuring the preferences of judges (e.g., Fjelstul 2022). The procedures dataset includes information on the disposition with respect to each legal procedure (except preliminary rulings, where there is no disposition). This data, combined with data on chamber composition and the characteristics of judges, will allow scholars to empirically test theories of judicial bargaining.

In sum, the CJEU Database Platform includes data on all of the major aspects of the Court’s internal processes and its key decision-makers. It also covers the universe of CJEU cases and decisions (1952–present). This will help researchers design and conduct better empirical tests to answer major questions in the study of EU judicial politics.

Accessing the data

We provide a number of ways for researchers to access data from the CJEU Database Platform. We have developed a web application that researchers can use to explore and download the data, read the codebook and other documentation, and learn about The IUROPA Project and our plans for expanding the CJEU Database.¹⁰ See the Supporting Information for details about the web application.

We also provide a REST API that researchers can use to query our database directly. For researchers who use R, we have developed an R package that provides an easy-to-use interface to the API. Researchers can use the R package to search the data and documentation and to download the data directly into R.¹¹ See the Supporting Information for a vignette that illustrates how to use the package and a manual for the package.

As we mention above, the CJEU Database Platform is the foundation of a larger CJEU Database. As other scholars contribute data to the CJEU Database, we will make the data available through the web application, our REST API, and our R package, making it easy for researchers to access a wide variety of CJEU-related data all in one place.

Researchers can easily merge data they have downloaded from the CJEU Database Platform with other data, including data on legal issues in references for a preliminary ruling from Carrubba and Gabel (2015) and Larsson and Naurin (2016) and network data from Fjelstul (2019), which tracks connections between case law, the treaties, and legislation. The CJEU Database Platform will also complement other related data-collection projects, both machine-coded and hand-coded, currently in progress.

Empirical applications

While debates about the power and autonomy of the CJEU in relation to the EU member states have been heated over the years, most of the existing research has been based on assumptions and empirical data that treats the Court as a unitary actor with a pro-EU attitude (Pollack 2013; Larsson and Naurin 2019). The CJEU Database Platform will help scholars develop and test theories that break with the unitary actor assumption and make progress in areas that have not yet received attention. This includes important questions relating to agenda setting, procedural politics and bargaining within the Court, the selection of judges to the Court and the significance of judges’ backgrounds for the process and outcome. In the conclusion, we suggest several significant themes that future research may engage in using the CJEU Database Platform.

In this section, we show the usefulness of the data by providing three brief illustrative examples related to some of these under-researched themes. These analyses establish new stylized facts about the CJEU that researchers can use as the basis for theory development. First, how has the Court’s use of discretionary procedural steps, like AG opinions, evolved over time, and what are the implications for the independence and legitimacy of the judicial process? Second, given the dramatic increase in the size of the Court, how has the professional experience of judges changed over time? Third, how frequent are conflicts of interest (situations where a judge participates in a case involving their own member state), and what can we infer about the independence of the Court? The CJEU Database Platform provides the data we need to address these important questions.

Discretion in the judicial process

The opinion of the AG is a critical step in the judicial process at the Court of Justice, but it is also a discretionary step, and one that takes a significant amount of time. The CJEU has a large backlog of cases, and cases can take years to resolve. Reducing the use of AG opinions is likely to speed up the process, but it could also reduce the political independence of the CJEU and decrease the quality of CJEU judgments. AGs are more likely to be insulated from political pressure than judges (Carrubba and Gabel 2015). By providing the Court with a well-researched, independent analysis of novel legal issues, AGs may enhance the independence of the judicial process and the quality of CJEU judgments. How has the Court’s use of AG opinions changed over time? In this section, we use data from our decisions dataset to investigate this question. We find that the Court is reserving AG opinions for important cases, which may negatively impact decision-making.

We focus our analysis on the contemporary period, since the formation of the General Court (1989–2021). The appropriate unit of analysis is a proceeding, so we aggregate our decision-level data by proceeding. We consider direct actions and preliminary rulings. We use our decisions dataset to code whether each proceeding was heard by a small (three judges), medium (five judges), or large (more than five judges) panel, and whether there is an AG opinion associated with the proceeding.

Figure 1 shows how the Court’s use of AG opinions has changed over time. During the 1990s, the Court requested an AG opinion in nearly all preliminary rulings and direct actions. Then, between 2003 and 2005, the Court’s reliance on AG opinions decreased markedly. During this period, the Court “made judicious use” of discretionary procedural steps, including AG opinions.¹² As a consequence of these procedural reforms, by 2010, the percentage of proceedings with an AG opinion had fallen to around 60 percent of preliminary rulings and around 40 percent for direct actions. This is one of the most significant procedural developments in the Court’s history. In the 2010s, this percentage began to increase again, but not to the same level as before.

Figure 1. This figure shows the percentage of proceedings with an AG opinion over time.

Figure 2 uses an alluvial plot to show, since 1989, the number of direct actions and preliminary rulings at the Court of Justice that have been assigned to small, medium, and large panels and the number of those proceedings in which there was an AG opinion. There is an AG opinion in 68 percent of proceedings. Most preliminary rulings (59 percent) go to medium panels. The larger the chamber, the higher the proportion of proceedings in which there is an AG opinion. In large chambers, the overwhelming majority of proceedings (93 percent) have an AG opinion. In medium chambers, 76 percent do, but in small chambers, only 39 percent do. Since the Court tends to reserve large chambers for important cases (Kelemen 2012), these empirical patterns suggest that the Court is more likely to request an AG opinion in important cases.

Figure 2. This figure shows the number of direct actions and preliminary rulings that have been assigned to small, medium, and large chambers and the number of those proceedings in which there was an AG opinion.

In sum, the Court of Justice has changed its approach to discretionary procedural steps, like AG opinions. Future research can use our data to develop and test theories of conflict expansion and contraction (Schattschneider 1960). By requesting an AG opinion or holding a hearing (another discretionary step), the Court effectively broadens the scope of participation and increases transparency and visibility into the case. Previous research has shown that courts may use public hearings and other public relations efforts as tools to increase public visibility, thereby addressing potential concerns about noncompliance (Vanberg 2005; Staton 2010; Krehbiel 2016). The EU is a particularly interesting case to study due to the relatively low degree of public attention to CJEU decisions and the particular compliance problems that international courts face.

The professional experience of judges

The number of judges at the CJEU has increased dramatically over the last two decades. Between 2004 and 2013, 13 member states joined the EU, nearly doubling the number of judges. On top of this, after the dissolution of the Civil Service Tribunal in 2016, the number of judges at the General Court has gradually doubled from one per member state to two per member state. In 2004, there were 15 judges at the Court of Justice and another 15 at the General Court. Currently, there are 27 judges at the Court of Justice and 54 at the General Court.

This rapid increase in the number of judges has raised questions about whether there would be enough qualified candidates, particularly from the Central and Eastern European member states (Dumbrovsky, Petkova, and Van Der Sluis 2014). The member states even created a merit commission (the Article 255 panel) to evaluate the professional qualifications of candidates to ensure they meet a satisfactory standard. How have the professional backgrounds of judges changed over this period? We find that the Court has generally professionalized over time, in the sense of judges having prior experience as a judge, and that judges from Central and Eastern European member states are relatively more likely to have prior professional experience as judges.

To assess how the professional backgrounds of judges have changed over time, we use data from our assignments and judges datasets. We use assignments as our unit of observation. Thus, there is one observation per judge per decision. This controls for the fact that some judges participate in more proceedings than others. First, we code whether each judge on the panel has prior experience working as a judge, an academic, a civil servant, a practicing lawyer, and/or a politician. Second, we calculate the percentage of assignments in which the judge has each type of experience by year, based on the year of the decision. The result is a measure that captures how common each type of experience is that we can use to make cross-sectional and time-series comparisons.

Figure 3 plots this data using a heat map. This plot shows how common each type of experience is and how the experience of judges has changed over time for both the Court of Justice and the General Court. We can also compare the Court of Justice to the General Court. The first thing to notice is that experience as an academic and as a civil servant is generally more common at the Court of Justice and at the General Court than experience as a judge. Experience as a politician, which may be perceived as implicating the Court’s independence and legitimacy, is the least common type of experience.

Figure 3. This figure visualizes the professional experience of judges over time for the Court of Justice (Panel A) and the General Court (Panel B). The color of each cell indicates the percentage of assignments in which the judge has each type of experience.

In general, the Court of Justice has professionalized over time. Early in its history, experience as a politician was more common than it is today. Experience as a judge has become more common at the Court of Justice since the 1990s, although it has become slightly less common recently. Experience as a civil servant has decreased since the 1990s. Comparing the Court of Justice to the General Court, the Court of Justice is more professionalized in the sense that experience as a judge has been more common. On the other hand, experience as a politician is also more common at the Court of Justice.

There is considerable variation in the experience of judges across member states. To make direct comparisons across member states, we calculate, for each member state, the percentage of assignments (2004–2021) involving a judge from that member state in which the judge has each type of experience. We include assignments at the Court of Justice and the General Court with respect to judgments. We omit experience as a politician because there is little variation across member states.

Figure 4 maps this data. There are substantial differences across member states. Consider Germany and France. In 94 percent of French assignments, the judge has prior experience as a judge. In 85 percent, the judge has experience as an academic, and in 100 percent, the judge has experience as a civil servant. For Germany, those figures are 7 percent, 74 percent, and 32 percent. A few patterns stand out. The common law countries Ireland and the United Kingdom clearly favor practicing lawyers. In all Irish and British assignments, the judge has experience as a lawyer, but experience as a lawyer is less common on the continent. The percentage of assignments for Central and Eastern European member states in which the judge has prior experience as a judge is considerably higher than for their Western European counterparts, with the exception of France.

Figure 4. This figure shows the percentage of assignments across the Court of Justice and the General Court by member state in which the judge has prior experience as a judge (Panel A), as an academic (Panel B), as a civil servant (Panel C), and as a practicing lawyer (Panel D).

In sum, as the Court has expanded in size, it has continued to professionalize. Judges from Central and Eastern European member states, contrary to the expectations of some observers, are more likely to have experience as judges than their Western European counterparts. There are many ways that researchers can advance the literature on CJEU judges using our data. First, they can use our data to study how the professional backgrounds of judges shape the trajectory of their career at the CJEU and the content of their decisions. For example, it could be that judges who have prior experience as a judge, especially at a constitutional court, are more productive and more influential among their colleagues. In addition, the professional experience of judges could shape their citation practices, influencing the development of precedent. Second, they can use our data to study how the characteristics of judges influence their chances of being reappointed by their member state for additional terms.

Norms and conflicts of interest

All EU member states appoint one judge each to the Court of Justice and two judges each to the General Court for six-year renewable terms. The appointment and reappointment of a judge is primarily in the hands of the government of their member state, even though the nominations are now screened by a merit commission (the Article 255 panel) and the formal decision is taken by the Council of the EU.

The fact that the judges depend on their home governments for reappointment has prompted the establishment of institutional features that limit the exposure of judges to political pressure from their government. The most important of these is the secrecy around deliberations and votes. In addition, in compliance cases, the President of the Court refrains from assigning a Judge-Rapporteur from the member state in question. In the last few decades, the same holds for preliminary rulings (Hermansen 2020). Does this norm of avoiding conflicts of interest extend to the other judges on a panel? Our assignments dataset allows us to examine this question for the first time. We find that this type of conflict of interest occurs frequently, but less frequently than it used to.

At some international courts — like the European Court of Human Rights (ECtHR) and the Inter-American Court of Human Rights (IACtHR) — it is standard practice for judges to participate in disputes involving their own member states (e.g., Naurin and Stiansen 2020; Stiansen and Voeten 2020; Stiansen, Naurin, and Bøyum 2020). This provides the court with local knowledge, but also creates potential conflicts of interest. In full-representation courts, where all member states appoint an equal number of judges, it also guarantees that member states have at least some representation in cases that directly concern them. When such rules are formalized in either the statutes or the rules of procedure of the court, it means that member states have decided on judges’ participation upfront. It may even be a condition for their delegation of sovereignty. Although the CJEU also features an equal number of judges per member state, the chamber system means that there is no guarantee that a member state will have representation on a panel.

From a judge’s perspective, there may be a tradeoff between their private convictions about the merits of a case and their fear of politically motivated retaliation from their member state. If a judge participates in too many rulings that go against their member state, their government may choose not to reappoint them for another term. The opposite may also be true. A judge could be rewarded with an additional term for ruling in favor of their member state. Judges at the CJEU can, to some extent, opt out of difficult situations. The rules allow judges to refrain from participating in specific cases by informing the President (Statute, Art. 18). However, judges cannot freely opt into deliberations.

As previously explained, the assignment of judges to proceedings follows from the appointment of the Judge-Rapporteur. Recent research has shown that the President selects Judge-Rapporteurs to avoid conflicts of interests, especially when cases are of political interest to member states (Hermansen 2020). The President may also avoid certain Judge-Rapporteurs due to the panel composition that would ensue. However, consistently relying on such a criterion would drastically restrict their choice set.

To assess the extent to which these types of conflicts of interest occur, we use data on the Court of Justice from our assignments, judges, parties, and procedures datasets. In direct actions, there is a conflict of interest if a judge was appointed by the member state involved in the proceeding. In preliminary rulings, there is a conflict of interest if the judge was appointed by the member state in which the referring court is located.

We find that many judgments involve a conflict of interest, but there is not a substantial difference between preliminary rulings and direct actions. At the member-state level, we find a conflict of interest in approximately 6 percent of assignments in preliminary rulings and 7 percent in direct actions. Accounting for the chamber structure, this entails a conflict of interest in approximately 33 percent of preliminary rulings and approximately 38 percent of direct actions. Panel A of Figure 5 shows the change over time. As the size of the Court has expanded — and the reliance on chambers has increased — conflicts of interest have become less common. Since the 2004 enlargement, the percentage of judgments with a conflict of interest has remained stable at around 20 percent.

Figure 5. Panel A shows the percentage of judgments at the Court of Justice with a conflict over interest over time. Panels B and C show the percentage of judgments with a conflict of interest by member state for direct actions and preliminary rulings.

Panels B and C of Figure 5 show this percentage by member state. In order to make direct comparisons, we only include assignments since 2004. The percentage of decisions with a conflict of interest varies substantially across member states, between 0 and 50 percent for direct actions, and between 0 and 25 percent for preliminary rulings. Without investigating the extent to which judges are randomly assigned, we cannot say whether the chamber system exacerbates or mitigates conflicts of interest.

In sum, contrary to the norm that applies to the Judge-Rapporteur, we find no evidence that the nationality of non-rapporteur judges precludes them from participating in cases involving their own member state. We see at least three lines of future research relating to judges’ impartiality. The first relates to the effect of the chamber system. Researchers can contrast changes in the internal rules with changes in the rate of conflicts of interests. The second relates to selection into panels. Since panel participation is a joint decision between the President and individual judges, researchers can identify the conditions under which judges are more likely to participate in deliberations when their participation would create a conflict of interest. The third relates to the dispositional outcome of direct actions: whether conflicts of interest translate into higher win-rates for member states. Researchers can evaluate the extent to which member states’ win-rates depend on their relative bargaining power and the professional experience of their judges.

Conclusion

The CJEU has been central to European integration and the judicialization of politics in Europe for decades (Pollack 2003; Stone Sweet 2004; Alter 2010; Kelemen 2011). It continues to be a major and controversial actor in European politics, evidenced by the debates around the Court’s role in Brexit, the challenges to the supremacy of EU law by the German and Polish constitutional courts, and the Court’s critical role in the EU’s response to the ongoing rule of law crisis in Poland and Hungary. It is also a court of special interest to both comparative judicial politics and international relations. Situated in a grey zone between international and domestic law and politics, research on the CJEU has unique potential for contributing to the integration of two unnecessarily separate research traditions (Staton and Moore 2011).

This article presents The CJEU Database Platform: Decisions and Decision-Makers, which includes data on the universe of CJEU cases, decisions, and judges. The CJEU Database Platform is the first easily accessible database on Court of Justice, the General Court, and the Civil Service Tribunal. Cross-referencing data from the Registry, InfoCuria, and EUR-Lex, the CJEU Database Platform provides political scientists and legal scholars with an extensive collection of reliable, research-ready data that is compatible with other recent data-collection projects on the CJEU (e.g., Derlén and Lindholm 2014; Carrubba and Gabel 2015; Larsson and Naurin 2016; Fjelstul 2019; Ovádek 2021). It also provides a foundation for the broader CJEU Database currently being developed as part of The IUROPA Project.

The development of theory and empirical research on the decision-making of the CJEU has made significant progress in some areas, including the interaction between the Court and the EU member states, within a principal agent or separation of powers framework (Garrett, Kelemen, and Schulz 1998; Pollack 2003; Carrubba and Gabel 2015; Larsson and Naurin 2016). Nevertheless, major questions concerning how the Court makes decisions remain to be more systematically researched.

Perhaps most importantly, with a few exceptions (Malecki 2012; Frankenreiter 2018; Fjelstul 2022), previous research has treated the CJEU as a unitary actor with homogeneous preferences in favor of more European integration (Larsson and Naurin 2019, 494; Pollack 2013, 1265). It is clear that this assumption is a crude simplification, based mainly on convenience and lack of data on votes and separate opinions. The CJEU Database Platform provides researchers with tools for opening the black box of the CJEU’s decision-making machinery, providing access to detailed information about the members of the Court, the decision-making procedures they are involved in, and the outcome of these procedures. By doing so, it will be possible to make progress on several fronts.

First, we know very little about how the 27 judges and 11 AGs interact to produce the case law of the CJEU. Who governs in the Court? Who sets the agenda? Who weighs more heavily in the deliberations? More research should focus on the role of the Court’s leadership and its organizational hierarchy. This includes the President’s powers with regards to assigning judges to chambers, and assigning cases to Judge-Rapporteurs and chambers (Hermansen 2020). It also includes the role of the General Meeting of the Court, which decides on important procedural questions (such as oral hearings, AG opinions, and chamber size), and the Chamber Presidents, whose agenda setting and mediating powers have become more significant after reforms to the Court’s rules of procedure.

Second, the CJEU Database Platform contains information on the background of judges that may be used to address questions relating to the sources and impact of judges’ attitudes. What difference do nationality, professional background, gender, or the political ideology of appointing governments make for judges’ behavior, the judicial bargaining process, and the outcome? What are the dominant conflict dimensions inside the Court? Some previous research has emphasized socialization within the Court as an important factor impacting on judges’ attitudes towards the role of EU law (Vauchez 2012), but so far very few have taken judge’s professional backgrounds into account (Chalmers 2015). There is also strikingly little research on the role of gender on the bench in a court that saw its first female judge appointed in 1999, 47 years after its inception (Grossman 2016).

Third, while much previous research has assumed the CJEU to be a highly autonomous supranational institution (Stone Sweet 2004, 9), “tucked away in the fairyland Duchy of Luxembourg” (Stein 1981, 1), there are in fact strong direct ties between the judges and the EU governments. Member state governments are the primary actors in the appointment process, and the fact that judges have renewable terms give governments important powers over their future careers (Stiansen 2022). Still, there has been surprisingly little research on how member states use this power to exercise judicial accountability (Kelemen 2012, 50). Until now, the research on judicial independence and accountability in the EU has relied on the Court’s judgments — its collective responses to the legal issues raised in the proceedings and the preferences signaled in the submissions of EU member states (Carrubba and Gabel 2015; Larsson and Naurin 2016). To explore the more specific connection between EU governments and individual judges would be a major step forward. How are member states using their powers to appoint and reappoint judges? To what extent are judges acting as “representatives” of their member states? What was the effect of the new merit commission (the Article 255 panel) that was introduced in 2010 to provide recommendations on judicial nominees? While votes and separate opinions remain secret, and all decisions are taken per curiam, the Court nevertheless publishes plenty of information about its members’ roles and assignments during the case management. The CJEU Database Platform makes this information easily available for researchers.

Finally, another important area where progress is needed, and where the CJEU Database Platform can contribute, concerns how EU law is mobilized by interest groups, businesses, and individuals in Europe. In the preliminary ruling procedure, this includes the role of lawyers and of judicial competition and cooperation between the CJEU and national courts. Recent research has challenged long-standing assumptions about the significance of national judges in the process of “constitutionalization” of the EU treaties through the preliminary ruling procedure (Weiler 1991), emphasizing instead the instrumental role played by private parties and activist lawyers (Pavone 2022). However, so far there has been little systematic large-scale research on legal mobilization of EU law, including who the litigants are that succeed in promoting their interests through the case law of the CJEU. The CJEU Database Platform is a necessary start in this regard, providing the groundwork for scholars to further classify litigants into theoretically relevant categories and to study their presence and achievements in the decisions of the Court.