Per curiam decisions and the judge-rapporteur

A number of civil law collegial courts issue their decisions per curiam without identifying an individual judge’s position on a case. Without the ability for judges to opine separately, a panel must produce a single judgment that accurately reflects its judges’ viewpoints. An institutional feature that may allow a judge to disproportionately influence the outcome is their service as the JR. The JR is responsible for writing the court’s judgment in a case. A court assigns a JR immediately after it receives a case and before any discussion among the court’s judges. They collect information regarding the case and draft a preliminary judgment for deliberations with the panel. As a result, the JR plays an agenda-setting function and their opinion “has a greater weight in the eyes of the other judges […] [T]he rapporteur holds a near monopoly over knowledge of facts and other materials concerning the case, including the competing arguments, so the other judges may be left at an informational disadvantage” (Kelemen Reference Kelemen2017, 43). The other panel judges, therefore, must exert effort to alter the JR’s opinion, which may be especially costly if they have a substantial workload. As Kelemen (Reference Kelemen2017, 42-43) describes, “The pressure of time often prompts judges to defer to the rapporteur’s opinion. In these circumstances, the other members of the panel are less likely to conduct their own research. So, in most cases the rapporteur’s opinion becomes the opinion of the court, and s/he is also the author of the judgment.” A judge that wants to convince their panel to support their appointer’s position on a case, thus, should be in a more advantaged position to do so when they are serving as the JR.

Application: Court of Justice of the European Union

To test my hypotheses, I examine the European Union’s highest court, the Court of Justice. I choose to analyze the CJEU for a few reasons. First, the CJEU is a powerful court, and EU member states have a vested interest in affecting its decision-making. Second, the CJEU is a per curiam court in which each member state appoints one judge for a six-year renewable term. Having both per curiam decision-making and renewable terms is empirically useful, as most civil law courts do not have both features. Third, although the EU treaties state that judges “shall be appointed by common accord of the governments of the Member States,” in practice, member states have historically seldom opposed other member states’ appointments to the CJEU.Footnote ² Fourth, previous research provides a method to measure member states’ preferences over case outcomes on the CJEU by leveraging observations (amici curiae briefs) that member states send to the Court (e.g., Carrubba and Gabel Reference Carrubba and Gabel2015). Fifth, the Court’s President assigns a JR to each case that arrives at the CJEU, with the JR having similar responsibilities to those at other civil law courts (e.g., Zhang, Liu and Garoupa Reference Zhang, Liu and Garoupa2018). Lastly, few articles focus on how individual judges on the CJEU affect the Court’s decision-making as a whole (e.g., Hermansen Reference Hermansen2020; Wijtvliet and Dyevre Reference Wijtvliet and Dyevre2021).

The CJEU’s primary task is to adjudicate disputes regarding the application of, and compliance with, EU law. Disputes arrive at the CJEU through a variety of mechanisms and parties. The CJEU’s wide jurisdiction and ability to exercise judicial review has substantial implications for member states and can affect, for instance, their gains from trade. Relatively unexplored in the literature, however, is how individual judges affect the CJEU’s decision-making. If a judge is ideologically aligned with their member state government or – since they are subject to six-year renewable terms – prefers to keep their seat on the CJEU, they may have incentives to please their (re)appointing government. To reduce the likelihood of this conflict of interest, CJEU judges do not hear infringement cases brought against their member state or preliminary reference cases originating from their member state. Nonetheless, since court rulings may affect the interpretation and application of law across the EU, member states may have a vested interest in cases in which they are not a party.

One means by which member states express their interest in a case is through submitting observations directly to the CJEU. All EU member states and institutions have the ability to submit observations on any pending case before the CJEU. Member states use these observations to express their beliefs on how the CJEU should rule in a particular case (e.g., Dederke and Naurin Reference Dederke and Naurin2018). I argue an observation can be particularly influential if the JR is ideologically aligned with their member state’s government. Similar to other civil law courts, the CJEU’s President assigns the JR before any deliberations. As Zhang, Liu and Garoupa (Reference Zhang, Liu and Garoupa2018, 146) explain, “even though conceptually each judge in the chamber is entitled to the same voting power, in reality their influence may vary depending on [whether they are the JR] in any given case. This group dynamic may, therefore, influence judicial voting.” Similarly, I expect that the JR’s influence should affect case outcomes.

Data & empirical evidence

Data. Carrubba and Gabel (Reference Carrubba and Gabel2015) conducted the first large scale systematic data collection effort to assess member state observations’ impact on the CJEU’s decision-making. Covering every judgment the Court issued from 1960 to 1994 (3,719 cases), they use these data to test a series of hypotheses regarding the Court’s sensitivity to threats of noncompliance and legislative override. The unit of analysis within these data is within-case legal issues (5,434 legal issues across the 3,719 cases).Footnote ³ For example, if a case had two legal issues and France sent an observation to the CJEU regarding the first legal issue and not the second, this within-case variation is reflected. The dependent variable of interest in these data is a binary variable (CJEU Ruling for Plaintiff) that takes the value of 0 when the CJEU ruled against the plaintiff’s position in the legal issue and the value of 1 when the Court rules in favor of the plaintiff’s position. Although I do not have data on each individual judge’s vote, as the CJEU’s decisions are per curiam, if my theory about the influence of the JR is correct, the panel outcome should be correlated with the JR’s preferences.

Additionally, for each legal issue, these data include whether an actor (member state or European Commission) submitted a brief in favor of the plaintiff or defendant. Combining these data with data from Cheruvu (Reference Cheruvu2019) and Fjelstul (Reference Fjelstul2019) on panel composition, I create a variable indicating whether the JR of a case received an observation from their member state favoring the plaintiff (+1) or the defendant (–1) in a case, taking the value of 0 otherwise (JR Net Observations). For example, the panel consisting of Judges Kakouris (JR, Greece), Mancini (Italy), and Hirsch (Germany) heard the case Elida Gibbs Ltd v Commissioners of Customs and Excise (CELEX number 61994CJ0317), which only concerned a single legal issue. Greece submitted an observation in favor of the defendant. Since Greece submitted an observation in favor of the defendant to a panel that contained the judge it appointed to the CJEU and their judge is the JR, JR Net Observations takes a value of $ -1 $ .

I also include a number of theoretically relevant controls in my analyses. First, I include a control for the net total of all observations member states submitted for a given legal issue favoring the plaintiff or the defendant (Net Observations) excluding observations already included in the statistical model through the JR Net Observations variable. In the aforementioned case Elida Gibbs Ltd v Commissioners of Customs and Excise, Germany, Italy, and the United Kingdom each submitted observations in favor of the defendant. Net Observations, thus, takes the value of $ -3 $ . By including this control in my models, I am accounting for the possibility that panels are ruling in favor of the plaintiff (defendant) to preclude the possibility of legislative override (e.g., Carrubba and Gabel Reference Carrubba and Gabel2015; Larsson and Naurin Reference Larsson and Naurin2016) and are not only catering to the preferences of the member state that appointed its JR. If panels are mainly concerned with the probability of legislative override, then including this control should also eliminate the effect of JR Net Observations.

Second, I control for the type of case the panel heard (infringement, preliminary reference, annulment, failure to act, staff case). This distinction may be relevant since the CJEU, for example, tends to overwhelmingly rule in favor of the Commission in infringement cases. As a result, the probability a member state observation affects a case’s outcome may differ depending on the type of case. Relatedly, I control for whether the Commission is a plaintiff or defendant in a case and whether the Commission submitted an observation favoring the plaintiff or defendant in the case, as research finds that the Commission has a substantial influence on the CJEU’s decisions (e.g., Stone Sweet and Brunell Reference Stone Sweet and Brunell1998). Furthermore, I control for whether a government is a litigant in a case (e.g., Carrubba and Gabel Reference Carrubba and Gabel2015). Additionally, I control for the number of judges hearing a case (Chamber Size), as existing scholarship provides evidence that larger chambers hear more important cases (e.g., Kelemen Reference Kelemen2012). Lastly, I control for the advocate-general’s (AG’s) position for each legal issue using a binary variable that takes the value of 1 when the AG favors the plaintiff. Scholars find that the AG’s opinions are strongly correlated with the CJEU’s decision-making (e.g., Larsson and Naurin Reference Larsson and Naurin2016). Table 1 provides descriptive statistics for variables included in the models.

Table 1.

Descriptive Statistics for Variables in Models

Empirical approach. I estimate a series of linear probability models to test hypotheses 1 and 2. All results are also robust to a logit specification. To test hypothesis 2, I create a variable Turnover that takes the value of 1 if the party of the prime minister in power at the time of a case is different from the party of the prime minister that appointed the JR. For robustness, I present results for alternative measurements of this variable in the Appendix, such as a change in the prime minister (Table A3) – even of the same political party – (Garoupa, Gili and Gómez Pomar Reference Garoupa, Gili and Pomar2021), change in left-right ideology (Table A4), and change in EU ideology (Table A5) sourcing from ParlGov. Formally, the OLS specifications for my analyses, where $ i $ indexes the legal issue, are:

(1) $$ CJEU\ Agrees\ with\;{Plaintiff}_i={\beta}_1\cdot JR\hskip0.5em Net\hskip0.5em Observations+\delta \mathbf{X}+\psi +{\unicode{x025B}}_{ip} $$

(2) $$ {\displaystyle \begin{array}{c}CJEU\ Agrees\ with\hskip0.5em {Plaintiff}_i={\beta}_1\cdot JR\hskip0.5em Net\hskip0.5em Observations+{\beta}_2\cdot Turnover\\ {}\hskip23.12em +{\beta}_3\cdot JR\hskip0.5em Net\hskip0.5em Observations\cdot Turnover\\ {}\hskip7.7em +\delta \mathbf{X}+\psi +{\unicode{x025B}}_{ip}\end{array}} $$

with $ \delta {\mathbf{X}}_i $ a vector of the aforementioned control variables, $ {\psi}_p $ JR fixed-effects, and $ {\varepsilon}_{ipt} $ standard errors clustered by JR. The JR fixed-effects control for JR-specific factors that may affect their decision-making on a case. For equation 1, a positive $ {\beta}_1 $ would be evidence in support of my hypothesis 1, indicating that when a member state sends an observation to the JR they appointed favoring the position of the plaintiff, the probability that the CJEU rules in favor of the plaintiff increases.

Results. Table 2 presents the results of the analysis testing hypothesis 1. Conforming with expectations for hypothesis 1, an observation from the JR’s appointer favoring the plaintiff increases the probability the JR’s panel rules in favor of the plaintiff. Model 1 presents the results for a bivariate regression with JR Net Observations as the independent variable. Each additional observation favoring the plaintiff increases the probability the CJEU rules in favor of the plaintiff by 20.5 percentage points ( $ p<0.01 $ ). Model 2 shows that this result is robust to the inclusion of JR fixed-effects ( $ \beta =0.207 $ , $ p<0.01 $ ). Model 3 includes AG for Plaintiff as a control, model 4 includes controls for the Commission’s involvement in a case, model 5 includes all controls, and model 6 includes all controls and JR fixed-effects. Across all model specifications JR Net Observations is positive and statistically significant, with an observation favoring the plaintiff increasing the probability the CJEU rules in favor of the plaintiff by 7.4 percentage points in model 6. For robustness, I run the aforementioned specifications as a logit in the Appendix and obtain similar results (Table A1).

Table 2.

Hypothesis 1 Results

Note: Standard errors clustered by JR are in parentheses *p $ < $ 0.1; ^**p $ < $ 0.05; ^***p $ < $ 0.01.

The control variables I included in these models also provide notable results. As expected, across all model specifications, when the Commission submits an observation favoring the plaintiff (defendant), it is positively (negatively) and significantly correlated with the probability the CJEU rules in favor of the plaintiff. Likewise, when the Commission is the plaintiff in a case, it is positively correlated with the probability the CJEU rules in favor of the plaintiff – as the vast majority of cases in which the Commission is a plaintiff are infringement cases in which they strategically decide which cases to bring the CJEU (e.g., Cheruvu Reference Cheruvu2022). Furthermore, the AG’s opinion is highly correlated with the CJEU’s decision-making in accordance with the existing literature, as an AG opinion in favor of the plaintiff correlates with a 59 percentage point increase in the probability the CJEU rules in favor of the plaintiff. Lastly, Net Observations has a positive and statistically significant coefficient, indicating that for each additional observation member states submitted to the CJEU in favor of the plaintiff for a given legal issue, the Court is about 2.7 percentage points more likely to support the plaintiff in a legal issue, complementing the primary result from Carrubba, Gabel and Hankla’s (Reference Carrubba, Gabel and Hankla2008) analysis.

Table 3 provides the results for the analysis testing hypothesis 2. Model 1 presents the results of the simple interaction, while model 2 includes JR fixed-effects. Both specifications have a negative and statistically significant coefficient on the interaction term. Nonetheless, the statistical significance disappears and magnitude is significantly reduced with the inclusion of the AG’s position (model 3), Commission’s position (model 4), and case characteristics (model 5). The fully specified model also does not demonstrate statistical significance (model 6).Footnote ⁴ To substantively interpret the interaction models, I plot the predicted outcome for each value of JR Net Observations conditional on turnover in Figure 1. The left pane of the figure visualizes model 1, while the right pane visualizes model 5. In both panes, for the no turnover condition, both nonzero values of JR Net Observations are significantly distinguishable from zero. In the turnover condition, however, the confidence intervals overlap zero. Focusing on the right pane, the point estimates for JR Net Observations are remarkably similar irrespective of the value of turnover – explaining the lack of statistical significance on the interaction term in model 5.

Figure 1.

Left Pane Based on Table 3 Model 1 and Right Pane Based on Table 3 Model 5. This figure provides predicted outcomes for all values of JR Net Observations conditional on Turnover with 95% confidence intervals clustered by JR.

Table 3.

Hypothesis 2 Results

Note: Standard errors clustered by JR are in parentheses *p $ < $ 0.1; ^**p $ < $ 0.05; ^***p $ < $ 0.01.

One possibility for the discrepancy in results is that little meaningful ideological variation exists between governments following turnover during the time frame of these data, reducing the impact of the binary turnover variable on case outcomes after accounting for relevant covariates. In other words, even though a JR’s appointer turned over, it may be true that on issues before the CJEU, the new government has similar preferences over case outcomes to the previous one. To address this explanation, I rerun the models using ParlGov’s eu_anti_pro variable in Table A5 in the Appendix,Footnote ⁵ taking the absolute value of the difference between the ideological score of a JR’s appointing government and the new government to create a new variable titled Change in EU Ideology (higher values indicate a larger ideological difference, while 0 indicates no turnover).

Figure 2 – based on Table A5 model 5, which includes controls – plots the predicted outcome for each value of JR Net Observations conditional on the change in EU ideology between governments. Although the coefficient on the interaction term is statistically significant ( $ \beta =-0.038 $ , $ p<0.1 $ ) and the slopes for the predicted outcomes for each value of JR Net Observations are different from one another, this relationship seems driven by turnover between the United Kingdom (UK)’s Labour and Conservative parties (Change in EU Ideology $ =4.668 $ , or approximately 5.617 standard deviations), as the highest value tick mark on the rug plot demonstrates. Indeed, within these data, the UK accounts for approximately 18% of all member state observations and 11% of data points in which JR Net Observations $ =1 $ or $ -1 $ and Turnover $ =1 $ . Nonetheless, taking all of the results together, these empirical findings suggest that the correlation between an appointer’s preferences as expressed through observations and outcomes is driven by ideology and not by retention incentives. It may also be the case, however, that the relative stability of national interests is more important than an appointer’s ideological bent for contextualizing a judge’s decision-making at international courts such as the CJEU. Under either interpretation, a judge’s home member state’s interests are relevant to their decision-making.

Figure 2.

Based on Appendix Table A5 Model 5. This figure provides predicted outcomes for all values of JR Net Observations conditional on Change in EU Ideology, with 95% confidence intervals clustered by JR.