Skip to main content

Conflict and Cooperation in the EU Heterarchical Legal System

Periodic Reporting for period 3 - EUTHORITY (Conflict and Cooperation in the EU Heterarchical Legal System)

Reporting period: 2018-09-01 to 2020-02-29

The EUTHORITY Project investigates patterns of conflict and cooperation between domestic and supranational courts in the European Union legal system. The generic question it seeks to answer is why and when national judges choose to cooperate and why when they choose to resist the authority of supranational courts.

Organizations and supranational legal regimes like the EU have very limited enforcement capacities. For that reason, the effectiveness of their laws depends on the support of domestic actors such as domestic courts. Domestic judges, though, are not on the EU payroll. The EU cannot promote, demote or fire them. Nor do EU courts have the power to reverse their doctrinal determinations. This heterarchical setting creates opportunities for domestic judges to resist the authority of EU law and to negotiate the terms under which they cooperate with EU judges. As intermittent defiance coexists with patterns of active compliance and pockets of covert disobedience, the absence of a judicial hierarchy can affect the reach of supranational rulemaking and result in fragmented and unequal judicial enforcement. While such a decentralised legal system may be favoured by those who wish to limit the power of ECJ judges, it may also produce inequalities in the application of EU law across courts and EU member states. These disparities have direct, practical implications for EU citizens.

The overall aim of the EUTHORITY Project is to document and identify the factors that determine the position of domestic peak courts--i.e. supreme and constitutional courts--towards EU law and legal integration. Towards that end, the investigators model strategic dynamics within the EU multi-level judiciary and undertake to collect and analyse data on the organizational structure, attitudes and decisions of peak courts across all EU-28 member states.
Over the first 30 months of the Project, the research team has developed its theoretical framework, collected large amounts of data on preliminary references, ECJ opinions and domestic decisions on the relationship between EU law and national law, completed an EU-wide expert survey and tested numerous hypotheses using state-of-the-art data science technique.

Theoretical efforts have helped sharpen our understanding of judicial motivation with respect to both conflict and cooperation. Formal models of judicial defiance have been elaborated which aid in identifying the conditions under which national courts are likely to defy the authority of international adjudicators as well as the capacity of domestic judicial actors to extract concessions from supranational judges through “judicial dialogue” (modelled as cheap-talk signalling in an iterated game). Formal modelling has also helped in advancing our understanding of the motives driving participation in the preliminary ruling mechanism. In particular, we have developed a model analysing how domestic judges may respond to formal dismissals given their policy motivation, reputational concerns and learning skills. Varying these parameters, we showed that dismissals can produce a chilling or a learning effect. Some of these results have already been published, while others are in the review process.

The team’s data collection efforts have reached major milestones. In some instances, e.g. preliminary references and ECJ judgments, we collected more data than we originally planned to. Data-collection has been facilitated by the use of data-harvesting techniques—a methodology that the team now masters well. Work on the Domestic Judicial Response (DJR) dataset has made great strides. After developing a comprehensive coding protocol, we created a survey interface for coders, defined a workflow, identified the relevant universe of judicial opinions, digitalised them where needed and recruited and trained coders. We have now coded more than 250 opinions on more than 50 variables. We have also constructed a new dataset on preliminary references, which compiles information on the geographic coordinates, level and type of the referring courts in addition to several other indicators. The final dataset has more than 9000 observations. We scrapped the entire universe of ECJ judgments from the CURIA and EUR-Lex websites. After cleaning and pre-processing this data, we applied text-mining methods to construct topic models of the case law and get an overview of the Court’s agenda. The team has also completed the first wave of its EU law expert survey—the EU Law Barometer. The team has expended considerable efforts and resources on this work package. The development of the questionnaire required careful theorising and validation exercises. A pilot survey was conducted in Belgium to test a beta version of the survey instrument. After iterated revisions, the instrument was then translated in 21 languages—a task that demanded rigorous drafting and proofreading by a multi-lingual team. The team proceeded to collect the names and contact details of 9000 EU law experts across all EU member states. The team liaised with national bars and EU law associations to increase the response rate and created a website (www.eulawbarometer.com) to advertise the survey. In the end, 587 EU law experts have rated the attitudes and practices of their peak courts regarding EU law. For her PhD project, Monika Glavina has conducted a survey of Croatian and Slovenian judges, with item questions covering familiarity with EU, workload, resources, referral propensity, and so on. 450 judges have responded to the survey, representing respectively 16 and 15 percent of the Croatian and Slovenian judiciary. She has also conducted qualitative interviews of judges in the two countries, gathering a wealth of information on obstacles and incentives to the application of EU law. Meanwhile progress has been made on the JUDICIARY dataset. The coding scheme has been through several iterations and, after specifying a workflow, the team has begun to collect data.

On the data analysis front, progress has been impressive. The team has acquired new technical skills that have greatly enriched the research:
• Bayesian methods: Bayesian statistics has been applied to the analysis of both EU and domestic court decisions to address small sample size or when frequentist models failed to converge.
• Delayed-lagged modelling of resubmission behaviour: we applied a non-linear delayed-lagged response model to investigate how domestic courts react to negative feedback (formal dismissals) from the ECJ.
• Machine learning: EUTHORITY researchers have used machine learning models for feature selection, prediction and data exploration.
• Data harvesting: the team has written computer codes to scrape data and texts from legal databases.
• Text mining: the team has used correspondence analysis, topic models and corpus dissimilarity metrics to map, analyse and compare the content of thousands of ECJ judgments.
• Visualizations: the team has developed maps of referral activity, animations and interactive plots (some of which can be viewed on the Project website: www.euthority.eu) to visualize patterns in the data collected.

Dissemination efforts have resulted in a dozen research papers, four of which have been either published or accepted for publication while the others are in the submission pipeline or in their second or third draft (these papers are available on SSRN). The Project and papers it has spawned have been presented at various conferences and workshops around Europe. The team has also been involved in the organisation of the first Conference on Empirical Legal Studies held in Amsterdam in 2016—the PI sitting on the organizing committee together with two prominent scholars in the field (Prof. Christoph Engel and Prof. Giuseppe Dari-Mattiacci). The Conference attracted more than 180 submissions and comprised 40 panels with contributions from all areas of law. The EUTHORITY has been busy organizing the second edition due to take place in Leuven on 31 May – 1 June 2018. The Conference is a major international event across the disciplines of law, economics and political science and, as such, an important channel to advertise the Project. Besides CELSE 2016 and 2018, the team has organized a seminar series featuring prominent scholars at the intersection of law, economics and political science. The website of the Project (www.euthority.eu) along with the website of the EU Law Barometer survey (www.eulawbarometer.com – the Project’s expert survey) has been another avenue to disseminate the Project’s initiatives, findings and deliverables.
"EUTHORITY has moved research on the EU judiciary beyond the state of the art at several levels. First, it has developed the first formal models of judicial interaction in non-hierarchical setting. Formal work on judicial institutions has generally assumed a hierarchical court structure like the US judiciary. So our modelling efforts contribute to clarify conflict and cooperation dynamics when judges interact in a heterarchical setup like the EU. Second, at the data collection level, we have gathered data that had never been systematically compiled before, such as data on the doctrinal position of domestic courts over European integration (DJR dataset) or the entire set of CJEU judgments, or not in the same comprehensive fashion, e.g. information on the geographic coordinates of referring courts. Finally, the Project has made application of methods hitherto never or rarely employed in this area of research: Bayesian statistics, delayed-lagged models, text-mining (topic modelling, text-scaling) and machine learning. Our expert survey is aslo the first to implement an expert crowd-sourcing design in this field of study.

While research has so far progressed on the basis of analysis of separate ""self-contained"" parts of the Project, subsequent work will gradually integrate these components into more comprehensive studies combining survey data, manually coded indicators, text-mining and machine coded information. We expect the most path-breaking findings of the Project to emerge from this synthesis.
"