Citation: Mattias Derlen; Johan Lindholm, Peek-A-Boo, It's a Case Law System: Comparing the European Court of Justice and the United States Supreme Court from a Network Perspective, 18 German L.J. 647 (2017) Content downloaded/printed from HeinOnline Wed Jan 3 08:29:22 2018 -- Your use of this HeinOnline PDF indicates your acceptance of HeinOnline's Terms and Conditions of the license agreement available at http://heinonline.org/HOL/License -- The search text of this PDF is generated from uncorrected OCR text. -- To obtain permission to use this article beyond the scope of your HeinOnline license, please use: Copyright Information Use QR Code reader to send PDF to your smartphone or tablet device Peek-A-Boo, It's a Case Law System! Comparing the European Court of Justice and the United States Supreme Court from a Network Perspective By Mattias Derlen &Johan Lindholm* A. Introduction /. CJEU and Case Law - Throwing Darts in the Dark? From the moment of its inception the European Union (EU) has included a court that was entrusted to give coherence and integrity to the interpretation and application of the Union's primary and secondary law.' That the Court of Justice of the European Union (CJEU) was to play an important role in settling disputes was clear. But few anticipated how instrumental the Court would become in the development of EU law.2 No one can dispute that the CJEU's judgments constitute an important source of European Union law. When the Court renders a judgment it settles the case at hand but also sets a precedent for how subsequent cases are to be resolved.' Collectively this precedent constitutes case law, sometimes even "settled" or "established" case law, which can serve * Professors of Law at the Department of Law at Ume6 University, Sweden. Email: johan.lindholm@umu.se, mattias.derlen@umu.se. The authors would like to thank James H. Fowler and Sangick Jeon for allowing us to reproduce their figures, the law schools at the University of Mississippi and the Ohio State University, Elizabeth Perry and Professor Russell A. Miller for helpful comments on earlier drafts, Andreas Nordstr6m for helping us create the data set, Vetenskapsridet for funding our research, and Martin Rosvall for introducing us to the amazing world of network analysis. 1Article 31 of the Treaty establishing the European Coal and Steel Community ("La Cour assure le respect du droit dans I'interpr6tation et I'application du pr6sent Trait6 et des reglements d'ex6cution."), now Article 19 of the Treaty on European Union (TEU). 2 See, e.g., Karen J. Alter, Who are the 'Masters of the Treaty'?: European Governments and the European Court of Justice, 52 INT'L ORG. 121 (1998); Joseph H.H. Weiler, The Transformation of Europe, 100 YALE L.J. 2403 (1991); PAUL CRAIG & GRAINNE DE BORCA, EU LAw 63-66 (5th ed. 2011). John J. Barcel6, Precedent in European Community Law, in INTERPRETING PRECEDENT 407, 417(D. Neil McCormick et al. eds., 1997). Barcel6 notes that the Court never explicitly refers to its previous judgments as "precedents". Id. However, the Court has acknowledged that the General Court's judgments can "constitute a precedent for future cases". Case C-197/09 RX-ll, M v EMEA, EU:C:2009:804, para. 62; Case C-334/12 RX-ll, Jaramillo et al. v EIB, EU:C:2013:134, para. 50. German Law Journal as a legal basis for deciding subsequent cases, occasionally extensively or even exclusively.4 For example, this is true concerning the principle of state liability, for which the Court now openly refers to its judgment in Francovich as the source of law.s The importance of such settled case law is demonstrated by how simply arguments going against such case law are rejected by the Court, sometimes by the well-known put-down "suffice it to say".6 It is clear that the CJEU's case law constitutes one of the primary sources of European Union law. But that is where the certainty ends. There are many questions about CJEU case law that legal scholars have not yet answered - or about which scholars have not found common ground. This article addresses the fundamental questions of how the CJEU' establishes and uses precedent.' The CJEU has frequently been criticized for lacking a clear method for establishing and using precedent. At least five specific (and interrelated) criticisms have been levelled at the Court: First, the Court normally only cites previous judgments in support of its arguments. Judgments pointing in other directions are typically ignored.' This approach is still described as a step forward as compared to the traditional approach of the Court, where passages from previous cases were repeated without giving any source."o Second, the Court does not 4See, e.g., Case C-409/06, Winner Wetten GmbH v Burgermeisterin der Stadt Bergheim, EU:C:2010:503, paras. 36, 39, 53, 33 and 58. 1 See, e.g., Case C-176/12, Association de m6diation sociale v Union locale des syndicats CGT et al., EU:C:2014:2, para. 50 ("a party injured as a result of domestic law not being in conformity with European Union law can none the less rely on the judgment in Joined Cases C-6/90 and C-9/90 Francovich and Others [1991] ECR 1-5357"). 6 See the overview in MITCHEL LASSER, JUDICIAL DELIBERATIONS 107-12 (2004). 7 The judicial system of the European Union consists of a two-part structure, with the CJEU at the top and the General Court below it. In this article, we only discuss the CJEU, as it is the most important court of the Union. EU legal development by way of case law primarily takes place in the CJEU. This court decides on average 600-700 cases per year. See Annual Report of the Court of Justice of the European Union, Luxembourg 2015, 93. Arnull has published a useful general introduction to the Court of Justice. See ANTHONY ARNULL, THE EUROPEAN UNION AND ITS COURT OF JUSTICE (2nd ed. 2006). The judicial structure previously also included a Civil Service Tribunal, ruling on disputes between the European Union and its staff, but this organ was dissolved in 2016 and its jurisdiction was transferred to the General Court. For further discussion see, e.g., Alberto Alemanno & Laurent Pech, Thinking Justice outside the Docket: A Critical Assessment of the Reform of the EU's Court System, 54 COMMON MKT. L. REV. 129 (2017). 8 In an earlier article, we explore under what circumstances CJEU case law is an important source of law. See Mattias Derl6n & Johan Lindholm, Characteristics ofPrecedent: European Court of.Justice Case Law in Three Dimensions, 16 GERMAN L.J. 1073 (2015). See, e.g., Anthony Arnull, Owning Up to Fallibility: Precedent and the Court of.Justice, 30 COMMON MKT. L. REV. 247, 252-53 (1993); L. NEVILLE BROWN & TOM KENNEDY, THE COURT OF JUSTICE OF THE EUROPEAN COMMUNITIES 369-70 (5th ed. 2000); ANTHONY ARNULL ETAL., WYATT & DASHWOOD's EUROPEAN UNION LAW 409 (5th ed., 2006). 10 See, e.g., Arnull, supra note 9, at 252. 648 Vol. 18 No. 03 The CJEU & SCOTUS - A Network Perspective explain why a particular judgment is cited, rather than another. As noted by Barcel6, the CJEU "does not discuss the facts of the prior case or the ratio decidendi to demonstrate that the holding is truly in point"." In fact, the common approach of the Court in this regard is described as "selective and superficial".1 2 Third, it is claimed that the Court practices "faux infallibility" by rarely openly departing from previous judgments." Express overruling has happened only in a few, well-known cases,14 and for example not a single time in the 52 Grand Chamber judgments from 2010 studied by Jacob.s Fourth, and similar to the issue of overruling, it is argued that the Court rarely distinguishes related cases, making the precedential value of old judgments uncertain." Even scholars who argue that the Court does engage in distinguishing make it clear that the practice of the Court is problematic. The approach of the CJEU to distinguishing includes manipulating judgments to avoid following previous case law.17 Finally, and most problematically, the Court is accused of simply ignoring the meaning of previous judgments in order to be able to reach a desired conclusion. This can happen as an implicit overruling, deviating from a previous judgment without even discussing it, " or as a form of pretend continuity, where a previous judgment is used as an authority for a particular conclusion, despite that obviously not being the case. 19 These criticisms target practically all aspects of the Court's interaction with previous judgments, giving an overall impression of a bumbling Court, uncomfortable and inexperienced in working with case law, a pale shadow of courts in common law countries. Our conclusion contradicts these assessments. The CJEU is a court with civil law roots but a case law future. Taking the constitutional nature of the CJEU into account, and moving away 11 John J. Barcel6, Precedent in European Community Law, in INTERPRETING PRECEDENTS- A COMPARATIVE STUDY 407, 416 (Neil D. MacCormick & Robert S. Summers eds., 1997). 12 Takis Tridimas, Precedent and the Court of.Justice, in PHILOSOPHICAL FOUNDATIONS OF EUROPEAN UNION LAW 307, 314 (Julie Dickson & Pavlos Eleftheriadis eds., 2012). 1" MARC JACOB, PRECEDENTS AND CASE-BASED REASONING IN THE EUROPEAN COURT OF JUSTICE - UNFINISHED BUSINESS 159 (2014). 14 Tridimas, supra note 12, at 316-20; BROWN & KENNEDY, supra note 9, at 370. 15 JACOB, supra note 13, at 160. 16 See, e.g., Tridimas, supra note 12, at 313-16; BROWN & KENNEDY, supra note 9, at 369-75; Ulrich Everling, Zur Begrandung der Urteile des Gerichtshofs der Europiischen Gemeinschaften, 1994 EUROPARECHT 127, 136-39 (1994). 17 JACOB, supra note 13, at 130-45. 18 See, e.g., Barcel6, supra note 11, at 416; Stefano Civitarese, A European Convergence Towards a Store Decisis Model?, REVISTA DIGITAL DE DERECHO ADMINISTRATIVO, No. 14, Julio - Diciembre de 2015, 173, 182. 19 See, e.g., Tridimas, supra note 12, at 315; Arnull, supra note 9, at 253. 2017 649 German Law Journal from the more traditional understanding of precedent, we view the CJEU's approach as an acceptable exercise of judicial authority in a case law system. 11. The Claim - CJEU is a Constitutional, Precedent-Driven Court We agree that there are grounds for criticizing the CJEU's approach vis-a-vis its own case law. For example, the reasoning of many judgments lacks transparency and consistency and the language is often cryptic and overly succinct.20 But the Court's method is not as deficient as some commentators argue. Our claim, which this article will support, is that the CJEU is a precedent-driven21 constitutional court comparable to the Supreme Court of the United States (SCOTUS) and with a comparable approach to precedent. We further argue that this case law approach is acceptable given the nature of the CJEU. Thus, we use the SCOTUS as a yardstick for measuring whether the CJEU takes case law seriously.22 From this we conclude that the existing debate regarding how the CJEU establishes and relates to precedent would benefit from a broader (internal and external) comparative perspective. On one hand, criticisms against the CJEU's approach are normally based on a limited number of CJEU decisions.2 3 On the other hand, criticisms of the CJEU's approach to case law seldom reflect on the way precedent is deployed in other legal systems. Regarding the former of these points, it is problematic to evaluate a sprawling, extensive, and continuously expanding system of case law-comprising thousands of individual decisionson the basis of a limited number of well-known judgments.24 This shortcoming in typical evaluations of the CJEU's practice is not surprising given the limitations of traditional legal 20See also Mattias Derl6n, Multilingual Interpretation of C/EU Case Law: Rule and Reality, 39 EUR. L. REV. 295, 296- 99 (2014). 21 The concept of precedent is discussed infra Part B. 22 This choice for comparison is explained and defended immediately infra. 23 This does not prevent scholars from making general claims. See, e.g., TREVOR C. HARTLEY, THE FOUNDATIONS OF EUROPEAN COMMUNITY LAW 72 (6th ed. 2007) (claiming that the frequency of citations in CJEU judgments has increased). 24 See Atieh Mirshahvalad et al., Significant Communities in Large Sparse Networks, 7 PLoS ONE (2012); Mattias Derl6n et al., Coherence Out of Chaos: Mapping European Union Law by Running Randomly Through the Maze of C/EU Case Law, 2013 EUROPARATTSLIG TIDSKRIFT 517; Mattias Derl6n & Johan Lindholm, Goodbye van Gend en Loos, Hello Bosman?: Using Network Analysis to Measure the Importance of Individual C/EU Judgments, 20 EUR. L.J. 667 (2014). 650 Vol. 18 No. 03 The CJEU & SCOTUS - A Network Perspective methods. 25 We echo Posner's and Fallon's battle cry, calling for empirical analysis of precedent in the US context.26 Only by combining quantitative and qualitative methods can we fruitfully discuss and understand the use of case law as a source of European Union law. For these reasons, this comparison of the CJEU's and USSC's attitudes to precedent takes an empirical approach relying on network analysis. 27 Fowler has pioneered this approach in the American context in two landmark articles.28 We draw inspiration for our study of the CJEU from existing studies of SCOTUS case law that use network analysis, including Fowler's. We replicate those studies and apply them to a network consisting of the CJEU's judgments, 29 and enrich our analysis by contrasting and comparing our results from those achieved with respect to the SCOTUS. The discussion below reveals that there are fundamental similarities between how the SCOTUS and the CJEU establish and use precedent. This, in turn, suggests that the latter court in many ways approaches its precedent in the same way as the former and should be judged accordingly. Three main arguments supporting this claim are presented below. The first argument, advanced in Part D, is that similarities in the basic features of the networks of CJEU and SCOTUS judgments negate claims that the CJEU establishes and uses precedent without any form of method. The second argument, presented in Part E, is that previous judgments are an indispensable source of law for both the CJEU and the SCOTUS and that this constitutes clear evidence of a system of precedent."o The third and final argument, defended in Part F, is that the CJEU's citation approach has four main components- (i) stages of development, (ii) issue shifting, (iii) a general approach to existing case law and a different approach to important cases, and (iv) overruling and avoiding precedent - and that its approach stands up quite well in a comparison with the SCOTUS's. 25 All studies mentioned in Part A.I are examples of such qualitative studies, encompassing a limited number of judgments from the Court of Justice. Even Jacob, who conduct a quantitative study, only discuss 52 judgments. See JACOB, supra note 13, at 87. 26 Richard A. Posner, An Economic Analysis of the Use of Citations in the Law, 2 AM. L. & EcoN. REV. 381, 402 and n. 30 (2000); Richard H. Fallon, Jr., Constitutional Precedent Viewed Through the Lens of Hartion Positivist Jurisprudence, 86 N.C. L. REV. 1107, 1115 (2008). 27 The use of network analysis is described in more detail infra Part CII. 28 James H. Fowler et al., Network Analysis and the Law: Measuring the Legal Importance of Precedents at the U.S. Supreme Court, 15 POL. ANALYSIS 324 (2007); James H. Fowler & Sangick Jeon, The Authority of Supreme Court Precedent, 30 Soc. NETWORKS 16 (2008). 29 The dataset used in this study, described further infra, was compiled by us as part of a larger research project and previously described and analyzed in sources cited supra note 24. 30 See also Derl6n & Lindholm, supra note 8. 2017 651 German Law Journal The CJEU is still influenced by its civil law roots. But this article empirically demonstrates that it exhibits key features of a case law based legal system. The CJEU has a systematic approach for deciding when to cite case law and determining which case law to cite. The CJEU regards case law as an indispensable source of law. And, as measured by a number of key metrics, the CJEU uses approaches to case law and precedent that are similar to those used by the SCOTUS. Consequently, while the CJEU deserves much of the criticism it receives, our findings suggest that its approach to precedent is not as poor as some would claim. B. Setting the Scene - Moving Beyond the Traditional View of Precedent The importance of precedent31 is continuously emphasized, including descriptions of precedent as the "life blood of legal systems." 3 2 Yet, despite all the spilled ink, the discussion of the nature and meaning of precedent has remained surprisingly stagnant. It is traditionally claimed that the common law and civil law traditions approach the doctrine of precedent in fundamentally different ways. This claim is primarily based on differences in the binding effect of previous judgments and, in particular, the absence of store decisis in the civil law tradition.33 Differing from the common law tradition, where court judgments are seen as a way to develop the law from below, 34 the civil law tradition does not regard earlier decisions as absolutely binding. Instead, previous judgements merely serve as interpretations of statutory law.35 The binding/non-binding dichotomy is too simplified to capture the attitude towards case law in the civil law and common law traditions. Indeed, the discussion about whether precedent constitutes a binding source of law in civil law has overshadowed the practical importance of case law.3 ' Even ifjudgments are not formally binding, the authority 1 As will be demonstrated by the discussion below, we here use precedent in a wide sense, encompassing all use of previous judgments, not limited to binding precedent. 32 C.K. ALLEN, LAW IN THE MAKING 243 (7th ed. 1964). As famously summarized by David & Brierly: "The place given to judicial decisions as a source of law distinguishes the laws of the Romano-Germanic family from the Common law." RENE DAVID & JOHN E.C. BRIERLEY, MAJOR LEGAL SYSTEMS IN THE WORLD TODAY 133 (3rd ed. 1985). 34 MATHIAS SIEMS, COMPARATIVE LAW 46 (2014). 1 Eric Tjong Tijn Tai & Karlijn Teuben, European Precedent Law, 16 EUR. REV. PRiv. L.827, 832 (2008). An alternative theory explains the difference between civil law and common law as the distinction between jurisprudence constante and stare decisis, emphasizing that civil law courts are expected to have regard to previous decisions when there is a high level of consistency in case law (settled case law or jurisprudence constante). See Vincy Fon & Francesco Parisi, Judicial precedents in civil law systems: A dynamic analysis, 26 INT'L REV. L. EcoN. 519 (2006). However, we remain unconvinced by this argument. The existence of a line of cases, rather than a single decision, matters in common law as well as civil law. See, e.g., William M. Landes & Richard A. Posner, Legal Precedent: A Theoretical and EmpiricalAnalysis, 19 J. L. EcoN. 249, 250 (1976). 36 Stefan Vogenauer, Sources of Law and Legal Method, in THE OXFORD HANDBOOK OF COMPARATIVE LAw 869, 894-95 (Mathias Reimann & Reinhard Zimmermann eds., 2006). 652 Vol. 18 No. 03 The CJEU & SCOTUS - A Network Perspective of decisions of higher courts can be a defacto very strong influence on decisions of lower courts.37 This is sometimes described as "precedent in the broad sense" or "persuasive precedent," as opposed to precedent in the strict sense.3 ' The differences between the common law and civil law traditions are more formal than practical.3 1 Concentrating on the issue of binding sources of law is not productive. To move on we have to acknowledge that there are different kinds of precedent, even within the common law paradigm. Two related aspects are of particular importance: constitutional precedent and self-precedent. As to the former a distinction is made between constitutional, statutory, and common law precedent. Out of the three, statutory precedent enjoys a "super-strong presumption of correctness," common law precedent occupies a middle position, and constitutional precedent is given a weaker presumption of correctness.40 The idea behind the weaker protection for constitutional precedent is the difficulty the legislator faces if it wants to intervene to object to case law. There are only very limited, external checks on the judicial interpretation of the constitution. In this situation a strong form of precedent would create an undesirable lock-in effect. Consequently, the SCOTUS needs to be able to correct the path of the law.41 This has been confirmed by the SCOTUS itself.4 2 Self-precedent is distinguished from vertical precedent (prior decisions of a higher court) and horizontal precedent (prior decisions issued by a peer court). Self-precedent is a prior decision issued by the same judge or the same court.43 The distinction is illuminating because the view of precedent changes as these relationships change. Gasc6n observes that, while 17 Tai & Teuben, supra note 35, at 833. 8See, e.g., Richard Bronaugh, Persuasive Precedent, in PRECEDENT IN LAW 217, 217 (Laurence Goldstein ed., 1987); Tai & Teuben, supra note 35, at 828. 1 Vogenauer, supra note 36, at 894-95. 40 William N. Eskridge, Overruling Statutory Precedents, 76 GEO. L.J. 1361, 1362 (1988); Earl Maltz, The Nature of Precedent, 66 N.C. L. REV. 367, 388 (1988). 41 Oona A. Hathaway, Path Dependence in the Law: The Course and Pattern of Legal Change in a Common Law System, 86 IOWA L. REV. 601, 656 (2001). See RUPERT CROSS & J. W. HARRIS, PRECEDENT IN ENGLISH LAW 20 (4th ed. 1991) (finding it unsurprising that the USSC has taken a less rigorous attitude towards precedent given the difficulties in changing the US constitution). 42 See, e.g., Burnet v. Coronado Oil & Gas Co. 285 U.S. 393, 407 (1932); Planned Parenthood v. Casey, 505 U.S. 833 (1992). 43 Marina Gasc6n, Rationality and (Self) Precedent: Brief Considerations Concerning the Grounding and Implications of the Rule of Self Precedent, in ON THE PHILOSOPHYOF PRECEDENT 35, 36 (Thomas Bustamante & Carlos Barnal Pulido eds., 2012). The more traditional approach is to only employ two categories: vertical and horizontal stare decisis. See, e.g., Amy Coney Barrett, Precedent and Jurisprudential Disagreement, 91 TEX. L. REV. 1711, 1712-13 (2013); Frederick Schauer, Has Precedent Ever Really Mattered in the Supreme Court?, 24 GA. ST. U. L. REV. 381, 385 (2007). Yet, following Gasc6n, we find it valuable to distinguish between different courts on the same hierarchical level and the same court, as only the latter is concerned by this study. 2017 653 German Law Journal important differences exist between the common law and civil law traditions as regards vertical precedent, the same does not hold true for self-precedent.44 Self-precedent does not require the court to always follow its previous judgments. It only requires the judge to justify departures from his or her previous rulings. 45 The underlying idea of self-precedent is not legal certainty or stability but rationality and the absence of arbitrariness. 46 Komdrek has developed the fundamental plurality of precedent into his theory of reasoning with previous decisions, identifying the legislative model as distinct from the traditional case-bound model. 47 In the legislative model judgments are drafted and interpreted as if they were legislative texts and the authority of the court is derived from its position in the judicial hierarchy. Thus, the wording of the judgment is closely scrutinized in search of a rulelike pronouncement by a higher court.48 The use of the legislative model of reasoning with previous decisions is not limited to the civil law tradition. Both the SCOTUS and the CJEU occupy positions that enable the use of the legislative model. The former court selects the cases to decide, delivers a small number of judgments each year, and is widely regarded as a political institution.49 The latter court was clearly envisioned as a superior authority on the interpretation of EU law. Neither the SCOTUS nor the CJEU are specialized constitutional courts like the ones found in many legal orders that have centralized judicial review,so like for example the German Bundsverfassungsgericht, but both are constitutional courts in the sense that they by merit of their elevated positions in their respect systems perform constitutional functions in a way and to an extent that distinguish them from lower courts of their respective legal systems.s" 44 Gasc6n, supra note 43, at 37. 45 Id. at 43. 46 Id. at 37-38. 47 Jan Komdrek, Reasoning with Previous Decisions: Beyond the Doctrine of Precedent, 61 AM. J. COMP. L. 149 (2013). 48 Id. at 162-63. 49 Id. at 165. See Earl M. Maltz, The Function of Supreme Court Opinions, 37 HOUSTON L. REV. 1395, 1420 (2000) (underscoring that "[b]y virtue of its position, the [SCOTUS] necessarily provides general legal rules that bind other actors in the system"). 50 See, e.g., Alec Stone Sweet, Constitutional Courts, in THE OXFORD HANDBOOK OF COMPARATIVE LAW 816, 817-18 (Michel Rosenfeld & Andris Saj6 eds., 2013). 1 Komdrek, supra note 47, at 165-66. But see David A. 0. Edward, Richterrecht in community law, in RICHTERRECHT UND RECHTSFORTBILDUNG IN DER EUROPAISCHEN RECHTSGEMEINSCHAFT 75, 80 (Reiner Schulze & Ulrike Seif eds., 2003) (concluding that a CJEU judgment "is not legislation, is not intended to be legislation and it should not be interpreted as if it were"). 654 Vol. 18 No. 03 The CJEU & SCOTUS - A Network Perspective In conclusion, we have to adjust our expectations when discussing how the CJEU establishes and uses precedent. We cannot expect the Luxembourg court to treat its own case law in the same way as a lower common law court would treat case law from a higher court. The CJEU is a constitutional court and it has a particular assignment in the legal system. It is reasonable that these courts will build upon their own previous cases, but the format is constitutional self-precedent. C. Methodological Questions 1. Finding a Yardstick - The SCOTUS as an Object of Comparison This article examines how the CJEU establishes and uses precedent. Our study draws inspiration from a comparison with the SCOTUS. This is a comparative study, with the comparison serving as an instrument to gain a better understanding of the CJEU. In any comparison both similarities and differences are examined in order to gain the most fruitful resultS.5 2 The differences between the two courts are obvious and include basic structural differences as well as differences in argumentation and style. One noticeable distinction is that the SCOTUS has the power to grant or deny certiori, while the CJEU has no similar docket control mechanism. The CJEU court can expedite the process for questions already answered.53 But it cannot close areas of case law to further discussion and rely on the existing decisions, as the SCOTUS can.54 There are also clear differences in style between the argumentative reasoning of the SCOTUS and the official, authoritative voice of the CJEU. Though, as pointed out by Lasser, the latter aspect is mitigated by the voice of the Advocate General.ss Still, the two courts also have significant similarities, making a comparison viable. It is always problematic to compare the CJEU with any other court, given the peculiarities of the Luxembourg court. While it is unhelpful to fall back on the sui generis description, 6 the CJEU certainly occupies an unusual position, being neither a national court nor a traditional, international court. The Luxembourg court has, on its own initiative and using the basic treaties of the Union, taken on the role of a constitutional court, developing EU law into an 52 Gerhard Dannemann, Comparative Law: Study of Similarities or Differences?, in THE OXFORD HANDBOOK OF COMPARATIVE LAW 383 (Mathias Reimann & Reinhard Zimmermann eds., 2006). See generally MORTEN BROBERG & NILS FENGER, PRELIMINARY REFERENCES TO THE EUROPEAN COURT OF JUSTICE 400-03 (2010). 54 MICHAEL J. GERHARDT, THE POWER OF PRECEDENT 45 (2008). 1 LASSER, supra note 6, at 236-38. 16 ROBERT SCHUTZE, EUROPEAN CONSTITUTIONAL LAW 67-68 (2012). 2017 655 German Law Journal effective legal system. Thus, it is reasonable to treat the CJEU as a constitutional court,57 making a comparison with another constitutional court such as the SCOTUS both practical and fruitful. In the words of Tridimas, the SCOTUS is an interesting comparison, as it "exercises constitutional jurisdiction in a pluralist judicial system applying an abstract founding law"." Finally, we have chosen the SCOTUS for comparison with the CJEU because of the established position of precedent in the American legal system in general and in the SCOTUS in particular. Store decisis, the binding force of precedent, has been said to be "the defining feature of American courts,"" including the SCOTUS. The justices regard the principle as "the heart of the rule of law.""o Naturally, it could be claimed that the SCOTUS has lost its way and deviated from its common law roots. For example, it is frequently pointed out that the SCOTUS takes a less rigorous attitude towards precedent as compared to English courts."1 But it is difficult to claim that precedent is not an important part of the American legal system. 11. The Broader Perspective -A Short Introduction to Network Analysis This study compares the approaches to precedent in the CJEU and the SCOTUS from a network perspective. The main network analysis concepts employed in the article are nodes, links, centrality, and authority.62 The first step when performing network analysis is to 17 Takis Tridimas, The Court of Justice and JudicialActivism, 21 EUR. L. REV. 199 (1996). 8 Tridimas, supra note 12, at 324. We also believe that a comparison between the two courts is interesting as both are arguably driven by ideas. While this concept cannot be fully explored here, the essential idea is the following: The CJEU is not linked to any country but rather to an abstract idea of Europe, see e.g., Ditlev Tamm, The History of the Court of Justice of the European Union Since its Origin, in THE COURT OF JUSTICE AND THE CONSTRUCTION OF EUROPE: ANALYSES AND PERSPECTIVES ON SIXTY YEARS OF CASE-LAW 9 (2013). This very characteristic trait of being linked to an idea rather than the history of an individual nation is absent in most other constitutional courts, but arguably not in the SCOTUS. The American nation is itself built on ideas, as evidenced in the Declaration of Independence and the Constitution, and these ideas still affect the US legal system, see e.g., KONRAD ZWEIGERT & HEIN KOTZ, INTRODUCTION TO COMPARATIVE LAW 239 (3rd ed. 1998). * Timothy R. Johnson et al., The Origin and Development of Store Decisis at the U.S. Supreme Court, in NEW DIRECTIONS IN JUDICIAL POLITICS 167, 167 (Kevin T. McGuire ed., 2012). 60 HAROLD J. SPAETH & JEFFREY A. SEGAL, MAJORITY RULE OR MINORITY WILL 6 (1999). 61 See, e.g., CROSS & HARRIS, supra note 41, at 19-20; NEIL DUXBURY, THE NATURE AND AUTHORITY OF PRECEDENT 123 (2008); SAUL BRENNER & HAROLD SPAETH, STARE INDECISIS 1 (1995); Richard P. Caldarone, Precedent in Operation: a comparison of the judicial House of Lords and the US Supreme Court, 2004 PUB. L. 759, 787. The idea is captured elegantly by Moore & Oglebay: "[Stare decisis] does not command unquestioning obedience to the past. It is a friendly and frequently persuasive link with what has gone before." See James Wm. Moore & Robert Stephen Oglebay, The Supreme Court, Stare Decisis and the Law of the Case, 21 TEX. L. REV. 514, 552 (1943). 62 See also Mirshahvalad et al., supra note 24; Derl6n et al., supra note 24. 656 Vol. 18 No. 03 The CJEU & SCOTUS - A Network Perspective arrange the judgments into a network, the raw network. The raw network consists of judgments (nodes) and citations between judgments (links)." As citations can only go backwards in time the network is directed, making every link both an out-link (a link from a node) and an in-link (a link to a node).64 The network is constructed solely from actual citations - the links. This enables us to claim that the network and the resulting patterns are natural, in the sense that they correlate with the perception of the judges of the respective court. The relative importance of a node (i.e. a judgment) is referred to as its centrality in the network. There are several ways of calculating centrality, the most obvious being simply counting the number of links to a node, in this context that would involve determining the number of citations a particular decision generates. The number of links to and from a node is also referred to as the node's degree. Thus, when speaking of a node's in-degree centrality we are referring to the number of citations to a particular judgment."s In-degree centrality is a straightforward centrality measurement, but it can be misleading when applied to case law. Counting the number of times a decision has been cited is not necessarily the best way of measuring its importance. All citations are not equal; being cited by a case, which is in itself important, should count for more than a citation from an unimportant case. Similarly, the fact that a decision is rarely cited does not conclusively prove that it is relatively unimportant, as a decision can be the basis of other cases and be at the core of an important area. At the same time, it would be unacceptable to conclude that only old, foundational cases are important and that more recent cases are irrelevant. A balance between the two extremes must be struck. In this regard, measuring cases' importance by using in-degree centrality will invariably favor older cases over newer cases, simply due to the fact that the former has had more time than the latter to accumulate citations. A better approach is therefore to use a non-local centrality measurement,66 also known as a feedback centrality measurement, thus called because rather than assuming that every 63 See generally Ulrik Brandes & Thomas Erlebach, Fundamentals, in NETWORK ANALYSIS - METHODOLOGICAL FOUNDATIONS 7 (Ulrik Brandes & Thomas Erlebach eds., 2005). 64 The existence of directed links gives added value to the network. See Elizabeth A. Leicht & Mark E. J. Newman, Community Structure in Directed Networks, 100 PHYS. REV. LETTERS 118703 (2008). It is noteworthy that while citations only can go backwards in time, ideas flow through the network in the opposite direction, from older to newer cases. 65 See Jon M. Kleinberg, Authoritative Sources in a Hyperlinked Environment, 64 . OF THE ACM 604 (1998). Regarding its application in a legal context, see, e.g., Daniel M. Katz & Derek K. Stafford, Hustle and Flow: A Social Network Analysis of the American Federal Judiciary, 71 OHIO ST. L.J. 457, esp. 491-44 (2010). 66 Other examples of such measurements, besides the ones discussed below, are Eigenvector and Katz centrality. 2017 657 German Law Journal citation is equally valuable, a decision's centrality is based on the characteristics of cases that refer to it.67 Using such measurements, an important node is one that is linked to other important nodes.68 In the context of a case law network, this conforms to the legally intuitive view that an important decision is a decision cited by other important decisions. In order to facilitate comparisons we follow Fowler in using authority score. 6 ' Authority is one aspect of the HITS algorithm that was developed by Jon Kleinberg.70 This algorithm provides two centrality measurements for each node: hub score and authority score." Authority score is a measurement of the amount of knowledge held by a node and, in a case law network, authority score therefore becomes a measurement of a decision's importance in a traditional sense, calculated by citations from the hubs of the network, discussed below. In case law networks, nodes with a high authority score (authorities) are decisions that are important because they say something vital about the content or development of the law. In other words, they are influential cases.72 Hubs (nodes with high hub score) are nodes that know how to find information on a given topic in the network. In case law networks, hubs are decisions that cite important decisions. In other words, they are well-grounded decisions.13 According to Kleinberg, authorities and 67 See Dirk Koschatzki et al., Centrality Indices, in NETWORK ANALYSIS - METHODOLOGICAL FOUNDATIONS 16, 53 (Ulrik Brandes & Thomas Erlebach eds., 2005). 68 See Renaud Lambiotte & Martin Rosvall, Ranking and clustering of nodes in networks with smart teleportation, 1 (2012), available at http://arxiv.org/pdf/1112.5252.pdf (May 12, 2017); Koschatzki et al., supra note 67, at 53. 69 See Fowler & Jeon, supra note 28. We generally favor using the PageRank algorithm that serves as the basis for how Google ranks webpages. See, e.g., Mattias Derl6n & Johan Lindholm, The Court of Justice and the Ankara Agreement: Exploring the EmpiricalApproach, 2012 EUROPARATTSLIG TIDSKRIFT 462 (2012); Derl6n & Lindholm, supra note 24. Very simplified, PageRank allows a "Random Walker" to explore the structure of the network by randomly following citations and occasionally teleporting to a random link in the network. PageRank, which is expressed as a percentage value, represents the relative probability that the Random Walker will find itself in a certain place and represents, as applied to a case law network, a decision's popularity. See Sergey Brin & Lawrence Page, The anatomy of a large-scale hypertextual Web search engine, 30 COMPUTER NETWORKS & ISDN SYSTEMS 107 (1998); Lawrence Page et al., The PageRank Citation Ranking: Bringing Order to the Web (January 29, 1998), available at http://ilpubs.stanford.edu:8090/422/1/1999-66.pdf (March 15, 2015). 70 Kleinberg, supra note 65. HITS stands for Hyperlink-Induced Topic Search. As indicated by the title of his article, Kleinberg developed the algorithm for use on the World Wide Web, but it has been used in the context of case law networks. See Fowler et al., supra note 28, at 330-32. 71 The initial step of using the HITS algorithm is to construct a focused subgraph of the network, for which the algorithm is employed. See Kleinberg, supra note 65, at 608-10. This is necessary in the context of the World Wide Web and other large networks in order to limit computational cost, but not with a network as small as ours. Consequently, we operate the HITS algorithm on the entire CJEU case law network. 72 Fowler et al., supra note 28, at 331. 7 Id. 658 Vol. 18 No. 03 The CJEU & SCOTUS - A Network Perspective nodes are mutually reinforcing: a good authority is a node pointed to by many good hubs, and a good hub is a node that points to many good authorities. 74 This definition is circular, which is why the algorithm must be iterative and take into account the entire network and assume that the sum of the authority-weight and the sum of the hub-weight are equal.7 s D. First Argument - Basic Features ofa Network I. Introduction -A Non-Random Approach In order to constitute a case law network, we can reasonably expect judgments to be connected through citations and that those citations are made on the basis of the issues discussed in each case. At least, in case X concerning issue A we expect the court to cite its previous decision in case Y concerning the same issue and we expect it not to cite case Z concerning issue B. When we extrapolate this reasoning to a whole network of case law, we expect certain patterns to emerge.76 First, we expect decisions to be connected to each other by citation to a relatively high degree, here referred to as the connectedness of the network. If the judgments are very loosely connected to each other, then this suggests that the court is not citing relevant cases, potentially because there are no relevant cases. Second, we expect that citations among cases are not distributed equally or, to use network analysis terminology, the network's degree distribution does not follow a flat or normal distribution. Instead, certain precedents where particularly important points of law were established should receive most of the citations and the great majority ofjudgments should receive relatively few citations.7 1 As explained below, the CJEU's case law network is quite similar to the SCOTUS's in both these regards and distinctly different from a random network. Thus, the first argument in support of our claim is that similarities in these basic features of the CJEU and SCOTUS case law networks negate claims that the CJEU establishes and uses precedent randomly. While 74 Kleinberg, supro note 65, at 611. 71 Kleinberg gives the following example, where p denotes page, x authority weight and y hub weight: "If p points to many pages with large x-values, then it should receive a large y-value; and if p is pointed to by many pages with large y-values, then it should receive a large x-value." Kleinberg, supro note 65, at 611 (italics omitted). In other words, if a judgment cites many influential cases, it should be considered a well-founded case, and if a judgment is cited by many well-founded cases it should be considered an influential case. 76 Naturally, this is only a minimum requirement. In reality a court will often interact with judgments that are similar, for example in order to distinguish them. 7See infro Part DII. 8 See infro part D.III. 2017 659 German Law Journal this hopefully surprises no one, it is a necessary first step, considering the extent of the above-described criticism of the CJEU and its approach towards case law. 11. Connectedness In network analysis, the degree of connectedness between nodes is referred to as the density of the network. A network's density is calculated by comparing the actual number of links to the potential number of links in a complete network, i.e. one where every node (i.e. case) is connected to every other. By dividing the former with the latter one derives a measure between 0 and 1 where 1 is the density of a complete network." While both the CJEU and SCOTUS networks are fairly sparse, they differ significantly from each other. The density of the CJEU's citation network is 0.00096.80 The density of the SCOTUS network studied by Fowler & Jeon is, by comparison, only 0.00048."' Thus, the SCOTUS network's density is only half that of the CJEU's. This may at first appear surprising considering that, on average, a recent SCOTUS judgment contains roughly twice as many citations as a CJEU judgment.8 2 The explanation is very simple: network size. The CJEU case law network analyzed here consists of all 8,879 judgments issued by the Court since its first case in 1954 until the middle of May 2011. The CJEU network is significantly smaller than the 30,288 judgments included in the SCOTUS network studied by Fowler & Jeon, including all majority opinions between 1754 and 2002."3 The fact that the CJEU's case law network is much smaller than the SCOTUS's is in no way surprising. The latter institution is more than four times older than the former.8 4 Still, this fact affect density. If the average number of references remains the same, then density will decrease almost exponentially as the size of the network increases."' Although the average number of references has 79 JOHN SCOTT, SOCIAL NETWORK ANALYSIS 81 (4th ed. 2017). 0 it should be noted that density is quite different than simply measuring average number of citations. 81 Calculated on the basis of Fowler & Jeon, supra note 28, at 18 (220,500 citations between 30,288 cases). 82 Compare Figures 5 and 6 infra. 8 Fowler & Jeon, supra note 28, at 17. The study by Fowler & Jeon includes all judgments in the U.S. Supreme Court Reporter. The latter includes judgments decided by the Supreme Court of Pennsylvania, before the establishment of the SCOTUS. See id at 17, footnote 1. It should be noted that Fowler et al., supra note 28, studies a slightly different data set consisting of all decisions between 1791 and 2005. Id. at 326. 84 In fact, the CJEU network is growing about five times as fast as the USSC network and will surpass it in size around the year 2085 if the current trend continues. 85 SCOTT, supra note 79, at 85-87. See also infra Figure 1. 660 Vol. 18 No. 03 The CJEU & SCOTUS - A Network Perspective historically increased in both networks," there is a very real, practical limit to how many citations can be included in a single judgment. Considering this, it is necessary to use an alternative measurement for network connectedness. One such alternative is inclusiveness, which "is the total number of points minus the number of isolated points."8 7 In this context, an "isolated point" is a case that neither cites nor is cited by at least one other case. Networks can be usefully compared by measuring "the number of connected points expressed as a proportion of the total number of points."" When we use inclusiveness instead of density a very different picture emerges. In the CJEU case law network 89% of all CJEU decisions are connected by citation, inward or outward, to at least one other case. That is, 11% of all cases are unconnected, i.e. do not cite any other case and are not cited by any other case. This finding alone suggests that the CJEU has a method when it cites case law, for if citations were actually distributed randomly the network's inclusiveness would be much higher, nearly complete. The inclusiveness among CJEU decisions is quite similar to that of SCOTUS judgments: the SCOTUS network has an only slightly lower inclusiveness of 84%." Thus, we see that the CJEU not only has a method but that the CJEU's method and the SCOTUS's method produce networks with very similar degrees of inclusiveness. This does not necessarily mean that they are the same or even similar methods, but it indicates that the CJEU approaches previous judgments in a consistent matter. 86 See infra Part E.ill. 87 SCOTT, supra note 79, at 81. 8 Id. 8 Yonatan Lupu & Erik Voeten, Precedent in International Courts: A Network Analysis of Case Citations by the European Court of Human Rights, 42 BRITISH J. POL. SC. 413, 424 n. 54 (2012). 2017 661 German Law Journal Inclusiveness: 3/4 = 0.75 :9- I -- I - I - I - I I I - I - I / - I h Density: 3/6 = 0.5 Inclusiveness: 5/6 = 0.84 Density: 5/15 = 0.34 Figure 1. Density & Inclusiveness ///. Degree Distribution Another way of proving that the CJEU approaches previous judgments in a consistent way is to consider the degree distribution of the CJEU case law network. Degree distribution is the variation in total number of inward and outward citations, or more generally, the number of links per node."o In a random network, where links are placed randomly between nodes, most nodes will have the same number of links." But, as pointed out by Albert & Barabasi, most complex networks are not random and links are not distributed randomly. Instead, complex networks in general and citation networks in particular tend to follow a power law distribution where most nodes will have few links and a small group of nodes will have a great number of links.9 2 Translated to the case law networks studied here, most judgments can be expected to have few inward and outward citations, and citations will instead gather in a small group of judgments with a great number of inward and outward citations. Fowler etal. demonstrate that the network of SCOTUS judgments follows a power law distribution, both regarding inward and outward citations (see figure 3 infra).13 Similarly, we find that the CJEU case law network is a power-law network. Figure 3 below demonstrates that this holds true both 90 Fowler et al., supra note 28, at 332; R6ka Albert & Albert-Liszl6 Barabisi, Statistical Mechanics of Complex Networks, 74 REV. MOD. PHYs. 47, 49 (2002). 91 Albert & Barabasi, supra note 90, at 49. 92 Mark E.J. Newman, Random graphs as models of networks, 99 PNAS 2566 (2002). 9 Fowler et al., supra note 28, at 332; Fowler & Jeon, supra note 28, at 18. See also Lupu & Voeten, supra note 89, at 425-26 (concluding that the same is true for the European Court of Human Rights in Strasbourg). I \1 662 Vol. 18 No. 03 2017 The CJEU & SCOTUS - A Network Perspective 663 regarding inward and outward citations. This demonstrates, first, that the CJEU case law network is not random but follows established patterns and, second, that there is no significant difference between the CJEU and the SCOTUS networks in their basic degree distribution. a 1 1 2 5 10 20 50 100 0 0 0 0 Inward Citations Figure 2. CJEU, Degree Distribution C 0 0 Inward Citatons 0 100 ol 1 2 5 10 20 50 Outward Citations Outward Citaticms Source: Fowler&Jeon 2008) Figure 3. SCOTUS, Degree Distribution 0 0 - E z o Lo 0 - a) 0 E z 0 m 1 German Law Journal IV. Interim Conclusions The CJEU's case law is connected largely in the same way as the constitutional, precedentdriven American court: relatively sparse (i.e. few links as compared to other types of networks) but few cases are entirely disconnected. Furthermore, in both networks, citations clearly follow a power law distribution. This constitutes a first argument in favor of a conscious approach vis-a-vis case law at the CJEU. The CJEU network is not random, but it also displays patterns that indicate some form of method. These similarities between the two courts are only a first step. Next we turn our attention to empirical measurement of the development of the networks over time. E. Second Argument - Case Law as an Indispensable Source of Law I. Introduction - Humble Beginnings No legal system can start out with a strong, established system of precedent and this includes the systems studied here. The explanation is partly practical: a minimum core of judgments is needed before a court can develop a systematic citation practice. There are also cultural explanations why neither court started out with a strong, established system of precedent. For the CJEU the cultural explanation can be found in its civil law heritage. The Luxembourg court was created by six continental civil law nations,9 4 and modeled mainly on the French system." In such a context, with heavy emphasis on legislation and a relatively limited role for the judiciary, there are clear cultural limits on the Court's capacity to act as a lawmaker. The absence of a system of precedent is hardly surprising. The SCOTUS, by comparison, also lacked a strong system of precedent early on; it hesitated to refer to and build upon its previous decisions prior to the nineteenth century." In this regard it essentially followed the pattern of English courts, where the strengthening of precedent, culminating in the 1898 London Tramways case," started during the nineteenth century." Thus, time is a relevant factor when considering how the SCOTUS and the CJEU established and use precedent. In this section, we present existing research on how the SCOTUS's use of 94 Belgium, France, Germany, Italy, Luxembourg, and the Netherlands. 9s Tamm, supro note 58, at 9. 96 Fowler & Jeon, supro note 28, at 18. 97 London Tramways v. London County Council [1898] AC 375. DUXBURY, supro note 61, at 125-26. 664 Vol. 18 No. 03 The CJEU & SCOTUS - A Network Perspective precedent developed over time and replicate these studies for the CJEU. In so doing we find that there are fundamental similarities between the SCOTUS and the CJEU in their gradual development towards a position where case law is now an indispensable source of law. The similarities in the development and current state of their approach constitute clear evidence of the CJEU having a system of precedent where new judgments are well embedded in existing case law. I. The SCOTUS - Development Towards Binding Precedent There are primarily two network analysis studies of how the SCOTUS's use of precedent has developed over time. The first study, conducted by Fowler & Jeon, investigated how the SCOTUS's citation practice has changed over time using primarily two measurements. The first measurement is average number of citations, i.e. average out-degree. The second measurement, which is a slight modification of the first, is the percentage of SCOTUS decisions each year that contain at least one outward citation." Both measurements illustrate a clear development, with average out-degree and percentage of cases citing at least one previous case increasing during the nineteenth century and continuing to expand during the twentieth century. The second study, carried out by Johnson et al., came to the same conclusion after finding a steady increase in average out-degree. Out-degree increased from an average of 1.1 citations during the Court's first fifty years to 18.7 citations in the last fifty years. The authors regard this increase as observable evidence of the institutionalization of the norm of precedent at the SCOTUS. 00 Fowler & Jeon furthermore sought to empirically verify the claim made in legal literature that the principle of store decisis was firmly established in the SCOTUS by the year 1900. There is significant agreement on the correlation between increasing references to precedent and the establishment of the principle of store decisis,ox but different perspectives prevail as to when store decisis was established in the SCOTUS. Fowler & Jeon primarily confirm the 1900 claim on the ground that the increase in the percentage of cases citing at least one previous case levels-off after 1900 at about 90%.102 Fowler & Jeon do not conclude that 90% of cases citing older decisions is a clear threshold for a system of precedent, only that it was true that 90% of SCOTUS cases cited older cases at the time when 9 Fowler & Jeon, supra note 28, at 19. They also discuss average in-degree, but we will return to that measurement below. 100 Johnson et a/., supra note 59, at 172-73. 101 Frank B. Cross et a/., Citations in the U.S. Supreme Court: An Empirical Study of Their Use and Significance, 2010 U. ILL. L. REV. 489, 528-29 (2010). 102 Fowler &Jeon, supra note 28, at 19. 2017 665 German Law Journal traditional legal doctrine claims that store decisis had been firmly established. Thus, 90% should not be regarded as a requirement for store decisis, only an indication, and a lower percentage does not rule out the existence of a system of precedent. Johnson et al. arrive at a seemingly quite different conclusion, claiming that the SCOTUS began to base its decisions on its own precedents by the early 1800s and that the norm of store decisis was firmly entrenched by 1815.103 The quite significant difference between the two studies is explained by a difference of perspective. Fowler & Jeon take an absolute perspective, concentrating on outward citations to previous SCOTUS judgments. Johnson et al. take a relative perspective, concentrating on the SCOTUS's transition from using English common law to its own precedent. This development is very clear, with references to English common law decreasing from 75.8% of all citations in 1791-1800 to only 10.9% during 1806- 1815.104 Fowler & Jeon observe a variation in the outward citation pattern by the SCOTUS. The trend is clearly increasing, sharply during the nineteenth century and more modestly during the twentieth century. A clear dip in this trend can be observed-regarding average out-degree (average number of citations in a case to a previous case) and the percentage of cases citing at least one previous judgment-during the so-called Warren Court (1953-1969).10s According to Fowler & Jeon this quantitatively confirms legal theory, in the sense that the famously activist Warren Court had less need for precedent due to their focus on creating new law.106 This makes some intuitive sense. The Warren Court is famous for revolutionizing many aspects of US constitutional law,' including cases such as Brown v. Board and Miranda v. Arizona.os If the importance of precedent is connected to evolution, its relevance in a revolutionary context is limited.109 10 Johnson et a/., supro note 59, at 169. 104 Id. at 169-72. 105 Fowler &Jeon, supro note 28, at 19; infro Figure 5. 106 Id. 107 Kurland has written an overview of the history and importance of the Warren Court. See P B KURLAND, POLITICS, THE CONSTITUTION, AND THE WARREN COURT (1970). 10s Brown v. Board of Education, 347 U.S. 483 (1954); Miranda v. State of Arizona, 384 U.S. 436 (1966); SPAETH & SEGAL, supro note 60, at 163. 109 Yet, as discussed further below, this is a complicated issue. See SPAETH & SEGAL, supro note 60, at 207 (arguing that the Warren court does not deviate significantly from the view of precedent as compared to other courts, claiming that "the Warren Court neither invented nor perfected preferential decision making."). 666 Vol. 18 No. 03 2017 The CJEU & SCOTUS - A Network Perspective 667 1/. The CJEU - From Zero to a Hundred in Thirty-Five Years The CJEU's use of previous decisions has developed over time. First and most obviously, the total body of CJEU case law has increased steadily over the years. This is of course unavoidable as new decisions are constantly added. But the CJEU's body of case law is growing much faster as it steadily increases the number of cases it settles each year.o It is not only the number of CJEU judgments that has increased over time; the same is true for its citation practice. The CJEU has steadily increased the average number of citations it makes to its own case law: from less than 0.5 until 1977 to 8.4 in 2011 (see figure 4 infra). 400 9000 300 6000 E 0) 200 3000 w - 0 100 - 0 -0 1954 1964 1974 1984 1994 2004 Year Figure4. ClEU, Number of Cases 110 See figure 4 infro. German Law Journal 0 0 CT T 1750 1800 1850 19090 52000 Year Source: Fowler & Jeon (2008) Figure 5. SCOTUS, Avg. Out-degree over Time 10 8 co 00 e~ oo 0 6 00 "o 00000 4 O 00 0 2 O 0 00 0 090b~dd 1954 1964 1974 1984 1994 2004 Year Figure 6. CJEU, Avg. Out-degree over Time It may appear as if average out-degree follows the number of judgments decided each year. But there is an important difference. The increase in out-degree roughly corresponds with the increase in decisions after 1977. Still, it does not explain why it took the CJEU twentythree years to properly start citing its own case law. The obvious fact that citations require something to cite cannot explain this discrepancy. In 1977 the CJEU had 788 judgments to choose from but only made 44 citations. By comparison, in 1980, the number of previous decisions had increased to 1,125 (a 42% increase) but the number of citations even more to 184 (a 318% increase). Thus, claims made by other scholars that the CJEU initially avoided 668 Vol. 18 No. 03 The CJEU & SCOTUS - A Network Perspective citing its own cases appear to be supported by our data."' This cannot be (entirely) attributed to the CJEU being bound by its French heritage or inexperience with case law. After all, it took the SCOTUS almost a century to properly start citing its own case law. The comparison suggests that case law-based systems must reach a "critical mass" before courts will start to cite their own case law. Average out-degree is a straightforward and intuitive measurement for how the CJEU's use of precedent has developed, but it is rather blunt. For example, a high average out-degree might be explained by more cases being available to cite112 and a few judgments with many citations can have a disproportionately large effect on the yearly average. The latter is particularly true for the early years, when the CJEU decided a very limited number of judgments per year.1 Given the limitations of average out-degree we follow Fowler & Jeon in considering what percentage of all judgments decided each year involve citations to at least one previous decision. This is, in our opinion, a good measurement of the importance of case law as a source of law: if the CJEU decides few cases without citing precedent, then that suggests that it is in practice an indispensable source of law. Using this measurement, we find a trend similar to that seen by using average out-degree, as the relative portion of judgments citing at least one precedent has increased from (i) around 10% in the early 1970s, to (ii) around 60% in the 1980s, to (iii) around 90% in the last two decades.114 The trend of three distinct periods of development at the CJEU follows the pattern identified by Fowler & Jeon in the SCOTUS's case law: lack, growth, and finally establishment of case-law use.11s 1 Arnull, supra note 9, at 252; Tridimas, supra note 12, at 309. 112 Fowler &Jeon, supra note 28, at 19. 113 This might explain the relatively high average out-degree in 1955 and 1959. 114 See infra Figure 8. See also Tridimas, supra note 12, at 309 (describing the development in a similar way). 115 Fowler &Jeon, supra note 28, at 19; infra Figure 7. 2017 669 German Law Journal 89 -0 >O