How does a tax case get from a national court to a CJEU judgment, what role does the Advocate General actually play, and how are the cases actually decided?
Paweł Mikuła talks with Professor David Hummel – for ten years référendaire in the cabinet of Advocate General Juliane Kokott at the CJEU – about how cases are allocated to judges and Advocates General, why he refused to read parties’ observations before writing his own solution, why an opinion is not primarily written to convince the judges, and why so many CJEU judgments read as if a paragraph is missing.
A must-listen for VAT professionals, EU tax practitioners, and anyone interested in how case law is actually produced at the Court of Justice.
DOWNLOAD FREE DATABASE
All CJEU rulings in VAT cases
All judgments from 1970
in one searchable place
Sign up to get instant access!
The administrator of your data will be Vatvocate, and your data will be processed in accordance with the principles set out in the privacy policy.
This article summarizes a discussion with Professor David Hummel, former référendaire in the cabinet of Advocate General Juliane Kokott at the Court of Justice of the European Union, and does not constitute a sourced academic paper. The analysis presented reflects the conversation’s key insights rather than comprehensive research.
Inside the Court of Justice: A Decade as Référendaire in Advocate General Kokott’s Cabinet
Introduction
This article summarizes a podcast conversation between Paweł Mikuła, partner at Halcyon | Tax. Customs. Legal., and Professor David Hummel, who served as référendaire in the cabinet of Advocate General Juliane Kokott at the Court of Justice of the European Union from 2016 until 2026. Since 1 November 2026, Professor Hummel holds a full chair in tax law and public law at Friedrich-Alexander University in Erlangen-Nurnberg.
The conversation offers a view of how tax cases — and VAT cases in particular — are processed at the Court: how they are allocated, how opinions are drafted, what makes an opinion influential, and why so many judgments read the way they do. It also touches on the working culture of an Advocate General’s cabinet, the limits imposed by language and translation, and the role of national legal traditions in supranational adjudication.
The Structure of an Advocate General’s Cabinet
Each member of the Court — both judges and Advocates General — is supported by four référendaires. The basic structure has not changed in the last decade. Members are largely free to staff their cabinets as they see fit, subject to one condition: référendaires must be qualified lawyers.
The smart approach, Hummel observes, is to staff a cabinet according to the work the member intends to do. Juliane Kokott has long taken an interest in tax law, competition and banking cases, and environmental law. Her cabinet was structured accordingly: a tax specialist (Hummel himself), a specialist in environmental law, and two référendaires working on competition and cartel cases.
The transition into this kind of role is not always smooth. Hummel arrived from an academic background and initially drafted in the way an academic would — raising questions, exploring multiple solutions, leaving decisions open. He was told politely that this was nice, but a little more directly, please. The work of a référendaire, he suggests, is close to the work of a professor in the sense that both involve solving a problem in a way that is useful to others. But the orientation is different: a cabinet draft must produce a workable answer to a specific question in a specific case, not an exploration.
Case Allocation: How Files Reach the Right Hands
The Court has a deliberate and somewhat unusual system for allocating cases. When a reference arrives, an internal department first identifies the legal issues involved. The President of the Court then decides which judge will be the reporting judge. Separately, the First Advocate General — one of the eleven Advocates General serving in that role — chooses which of his or her colleagues will be responsible.
There is one firm internal rule: a German Advocate General and a German judge will never both be on a case from Germany. Beyond that, allocation is discretionary.
Hummel defends this system, even though it sits uneasily with German legal sensibilities (where allocation is normally pre-determined). The reason is structural. Each Member State nominates one judge, and those judges are by necessity generalists — typically public lawyers, EU lawyers or international lawyers. None of them is necessarily a specialist in tax law. If allocation were random, asylum cases might land with someone who knows nothing about asylum, while tax cases might land with a civil lawyer. Discretionary allocation allows the President and the First Advocate General to match cases to the „hobbies” of individual members. In practice, Kokott’s cabinet received more tax cases, more environmental cases, and fewer asylum cases than others — which Hummel views as a sensible use of expertise.
The volume gives a sense of scale. The Court receives roughly 40 tax cases per year. Within that, between 10 and 20 are clear references on tax law. A very productive référendaire can write up to 10 opinions per year; a well-functioning cabinet might produce 30 to 35 in total. Hummel’s annual share of tax opinions was in the range of 10.
The Decision to Write an Opinion
Once a case is allocated, the reporting judge prepares a preliminary internal report. This report is significant: it lays out the facts, the legal problems, the relevant legal bases, and ends with proposals on the composition of the chamber (three or five judges), whether there should be an opinion, and whether there should be an oral hearing. The Advocate General’s cabinet must agree to these proposals before the report is confirmed at the general meeting of the Court.
The internal default is straightforward in principle. Cases of the highest importance go to the Grand Chamber and always receive an opinion. Difficult cases with significant practical impact go to a chamber of five and normally receive an opinion. Easy, narrow, technical questions go to a chamber of three, normally without an opinion.
In practice, however, Hummel observes that the last decade has seen many opinions in three-judge cases — with the overwhelming majority written in Kokott’s cabinet. The reason is that tax cases were often proposed for chambers of three without an opinion on the basis that they were „merely technical.” Many in the cabinet — and Kokott herself — disagreed. The compromise reached with reporting judges was typically to keep the case in a three-judge chamber, but with an opinion. In five-judge chambers, judges often welcomed the prospect of an opinion as a source of guidance.
The most demanding situation arose when the cabinet considered the report’s proposed solution to be wrong. In those cases, an opinion was written specifically to persuade the chamber not to follow the report. This, Hummel notes, was not unusual in tax law.
The Workflow of Drafting an Opinion
The preliminary ruling procedure, as Hummel emphasises, is in the end a question of law. The facts come from the referring court; the legal question is set out in the reference itself. The quality of written observations submitted by the parties — and even by the Commission — was, in his experience, frequently disappointing.
His personal working method reflects this. He would not read the parties’ observations until he had already worked out his own solution. The starting point was the facts as the Court had them, the legal bases (mostly identifiable from the reference), and his own reasoning. Only after reaching a conclusion would he read the observations, primarily to check whether anyone had raised an argument he had missed. In most cases, he says, no such argument appeared; sometimes a strange Commission argument had to be acknowledged and dismissed.
Once a draft existed, it went to colleagues in the cabinet for review. This step Hummel regards as genuinely valuable. None of the other référendaires was a tax specialist — they came from environmental law and competition law. Convincing a non-specialist that a tax solution is the right one is, he argues, the best test possible: judges in the chamber are themselves generalists, and if a draft cannot persuade colleagues without a tax background, it will not persuade the judges either. The process could be uncomfortable, but it imposed a discipline of clarity.
If disagreements with colleagues remained, the question went to Kokott herself, with Hummel’s recommendation and the alternative. On most disputed points the issues were minor.
The Influence of Parties on the Court
Hummel does believe parties can influence outcomes, but only if their observations are well-conceived. The most common mistake was the assumption that the Court itself is specialised. Lawyers would arrive with highly detailed, technical arguments that no generalist judge could absorb. The successful approach, in his view, requires the opposite: stepping back and explaining the case in straightforward terms, illustrating with examples, and showing the practical consequences of each possible decision.
To make the point: a party arguing that a transaction is a supply of goods rather than a service should explain what follows from that — what other provisions of the directive apply, what the result will be in concrete terms. Lawyers tend to assume that anyone reading the brief will already see those consequences. For a generalist court, that assumption is unrealistic.
He illustrates the importance of consistency with an anecdote. The Court had two similar cases — one from Slovakia, one from Spain. The Commission’s Legal Service prepares observations, and the arguments submitted by the Commission in the two cases were entirely inconsistent with each other. No one within the Service had noticed.
The Function of Opinions: A Library of Arguments
Asked whether opinions are sufficiently followed by the Court, Hummel offers a more nuanced view than the question expects.
The likelihood that an opinion is followed depends, in his view, primarily on its quality. A good opinion is neutral, logical and follows a clear thread. It identifies the problem, sets out the possible solutions, explains why the proposed solution is the best one, and demonstrates an understanding of the consequences. He notes that some opinions in the case law fail this test — they wander, do not present alternatives, or do not see where their own proposal leads.
But the more important point is conceptual. The function of an Advocate General’s opinion, he argues, is not to convince the judges in the case at hand. The function is to inform the public — to lay out the problem, the arguments, and the reasoning so that all lawyers can form their own view. On this conception, an opinion that the Court does not follow has not failed; it has simply done a different part of its job.
Indeed, Hummel suggests that some of the best opinions are precisely those that are not followed at the time. They become part of a library of arguments. A case decided one way today may, fifteen or twenty years later, be revisited in light of the consequences that the original decision has produced. At that point, an unused opinion may resurface and provide the route to a different result. This is, in his view, one of the most important contributions opinions make to the long-term development of EU law.
The Quality of Court Judgments: Length, Language, and Compromise
A significant part of the conversation concerns the quality and readability of Court judgments. Both Mikuła and Hummel express dissatisfaction with the form of much of the case law — and Hummel offers several specific explanations.
The first is language. The Court’s working language is French, and Hummel argues that French is not an ideal language for precise legal drafting. He describes it as poetic but not exact, with a sentence structure different from most other European languages. The five largest Member States may submit drafts in their own languages, and Hummel and Kokott drafted in German. The translation experience was illuminating. A 40-page German draft would yield an English version of roughly 38 pages and a French version of about 45. The German style favoured short, precise sentences; the French translator would compress three German sentences into a single French one. When Hummel asked for his three sentences back, he was told this was impossible in French — until, having reached the point of writing in French himself, he insisted.
The second explanation concerns the use of text blocks. Generalist judges, when drafting a proposed judgment, tend to look for previously translated and formulated paragraphs that can be reused, rather than developing fresh reasoning. The result, in many judgments, is a structure where 30 of 40 paragraphs are recognisable text blocks and only one or two paragraphs contain anything new. While Hummel understands the translation logic behind this practice, he argues that it is unsuitable for genuinely developing case law.
The third explanation, and the one that most directly bears on judgments that read as if they have a missing step, concerns compromise. Judges, deliberating among themselves, do not edit the draft phrase by phrase but move whole paragraphs. When a controversial argument is removed in this way, the surrounding text remains, but the line of reasoning loses its coherence. The judgment ends up containing a statement without the reasoning that produced it. A statement, Hummel observes, is not an argument: it may be true or false, but without the supporting reasoning the reader cannot tell which. Where this occurs, the judgment becomes difficult to apply in subsequent cases — and the influence of the case law erodes accordingly.
Tools, Sources, and Time Pressure
When asked which tools and resources he relied on, Hummel offers a direct answer: the first tool is the head. After that, the Court’s own internal sources and earlier case law are extensively used — and reused, given that text blocks already translated into all official languages are valuable in their own right. He notes that the older Curia case-search interface was excellent and remains accessible behind the „advanced search” link on the new site.
National sources played a substantial role in his own work. German tax literature is unusually deep, with a long tradition of substantive commentaries (Otto Schmidt’s database). Most of the VAT problems arriving from Poland or Spain corresponded to issues already addressed in German legal scholarship, and German commentaries provided a starting point. He found French scholarship of limited use for genuinely critical analysis: it tended toward summary rather than evaluation. International sources such as IBFD’s International VAT Monitor were used where relevant.
A national legal background, in this account, inevitably shapes the way work is done at the Court. Hummel views this as natural rather than problematic — provided that the legal solutions ultimately proposed are grounded in EU law and visible to all parties.
The constraint that has changed over time is time itself. An academic can think about a problem for half a year. A référendaire has roughly one month. For a specialist in the relevant field this is workable; for a non-specialist confronting a new area, it is short. Hummel argues that the older case law was shorter in form but reflected longer thought — and that current judgments tend to be longer in form but produced under heavier workload pressure, with predictable effects on quality.
Looking Forward: Tax Cases After 2027
Advocate General Kokott’s mandate runs until 27 October 2027. Whether she will be renewed is unknown. For the intervening period, Hummel’s successor in the cabinet is a former colleague who knows the cabinet well but is not a tax specialist. Hummel expects, accordingly, that fewer tax opinions will emerge from the cabinet in the period leading up to 2027 — though some interesting cases on direct taxation remain. If Kokott is renewed, he expects she will look again for a tax expert; if not, the position the cabinet has held in tax matters at the Court for many years will not necessarily continue in the same form.