Andreoli-Versbach, Patrick und Franck, Jens-Uwe: Actions Speak Louder than Words: Econometric Evidence to Target Tacit Collusion in Oligopolistic Markets Munich Discussion Paper No. 2013-8 Department of Economics University of Munich Volkswirtschaftliche Fakultät Ludwig-Maximilians-Universität München Online at https://doi.org/10.5282/ubm/epub.16179 Actions Speak Louder than Words: Econometric Evidence to Target Tacit Collusion in Oligopolistic Markets Patrick Andreoli-Versbach* and Jens-Uwe Franck†‡ July 2013 Abstract Tacit collusion reduces welfare comparably to explicit collusion but remains mostly unaddressed by antitrust enforcement which greatly depends on evidence of explicit communication. We pro- pose to target specific elements of firms’ behavior that facilitate tacit collusion by providing quantitative evidence that links these actions to an anticompetitive market outcome. We apply our approach to incidents on the Italian gasoline market where the market leader unilaterally an- nounced its commitment to a policy of sticky pricing and large price changes which facilitated price alignment and coordination of price changes. Antitrust policy has to distinguish such active promotion of a collusive strategy from passive (best response) alignment. Our results imply the necessity of stronger legal instruments which target unilateral conduct that aims at bringing about collusion. Keywords: antitrust law, tacit collusion, oligopolistic competition, gasoline market JEL classification: K21, K42, L13, L41, L71 * International Max Planck Research School for Competition and Innovation, Munich Center for Innovation and Entrepreneur- ship Research and Ludwig-Maximilians-Universität München, Marstallplatz 1, 80539 Munich, Germany, Phone: +49 (0)160 3081 775, Fax: +49 (0)89 242 46 503, Email: [email protected]. † Contact author: Ludwig-Maximilians-Universität München, Law Faculty, Ludwigstrasse 29/III, 80539 Munich, Germany, Phone: +49 (0)89 2180 2905, Fax: +49 (0)89 2180 2904, Email: [email protected]. ‡ We thank David Abrams, Thomas Ackermann, Carlo Bardini, Tomaso Duso, Fabio Massimo Esposito, Valentina Giuliani, Andrea Güster, Dietmar Harhoff, Justus Haucap, Michael Meurer, Hans Müller, Alessandro Noce, Martin Pesendorfer, Fabio Pinna, Stephen Ryan, Pierluigi Sabbatini, Monika Schnitzer, John Sutton, Simcoe Timothy, Andreas Hahne, Daniel Petzold, Niels Petersen, Heike Schweitzer, Ansgar Wohlschlegel, and Joachim Winter. Thanks for insightful comments and suggestions are due to participants at the Mannheim Law & Economics Forum (2013), the EALE Conference in Stockholm (2012), the LawEcon Workshop at the University of Bonn and the Max Planck Institute for Research on Collective Goods (2012), the EARIE in Rome (2012), the EDGE conference in Bocconi (2011), the CISS workshop in Turkey (2011), the Max Planck Insti- tute conference in Wildbad Kreuth (2011), the ZEW conference on "Quantitative assessment of Antitrust Analysis" in Mann- heim (2010), and PhD workshops at the Ludwig-Maximilians-Universität München. We have also benefited from discussions with and comments from practitioners at three economic consulting firms in London and Oxford: Compass Lexecon, NERA and Oxera. Finally, we are indebted to seminar participants at the Italian Antitrust Authority in Rome and the German Federal Cartel Office in Bonn for their valuable suggestions. Patrick Andreoli-Versbach gratefully acknowledges financial support from the Deutsche Forschungsgemeinschaft through GRK 801 and the generous financial support from the International Max Planck Research School for Competition and Innovation. Please note that all data and codes we used to generate the results of this paper will be available online for the purpose of replication or further research as soon as the paper has been published. 2 1 Introduction In most markets firms quickly realize that they can earn supracompetitive profits by coordinating their market conduct. In response, antitrust policy seeks to foster “effective competition” by tar- geting collusive activities. The current legal framework to accomplish this goal has mainly evolved around communication as a means to reach a collusive agreement. In contrast, purely tacit collusion remains largely unaddressed by antitrust law though it may bring about the same negative welfare effects. We argue that a crucial step forward in targeting tacit collusion could be taken through the foren- sic use of econometric evidence which may reveal collusive strategies. Theoretical and empirical findings on collusive behavior provide a basis for deriving clear test hypotheses to distinguish (lawful) oligopolistic interdependence from tacit collusion. Thus econometric analyses may pro- vide quantitative evidence that firms strategically use specific elements of market conduct to (tac- itly) collude. Antitrust remedies should in turn take up such instances of market behavior to tack- le tacit collusion. The paramount significance of evidence of explicit communication entails fundamental problems for the fight against cartels.1 Communication is not a necessary condition to collude. At the heart of collusion lies the incentive of firms to cooperate rather than to compete.2 In oligopolies firms can exercise their unilateral market power to facilitate anticompetitive coordination without en- gaging in communication. As firms weigh up the costs and benefits of explicit collusion, antitrust law’s focus on communication incentivizes them to concentrate on tacit means of collusion. Le- gal instruments to counter collusion, the effectiveness of which depends on evidence of explicit communication, are least effective in concentrated industries,3 i.e. precisely in those industries where the cartelization rate is presumably the highest and communication is least needed to sus- tain collusion.4 Any economic approach to support the enforcement of antitrust law5 is chal- 1 Throughout this paper we use the term “cartel” to describe any kind of welfare-decreasing form of collusion, be it an explicit or a tacit one, and irrespective of whether or not we consider it an infringement of antitrust law. 2 Much of the theoretical discussion on tacit collusion is based on the supergame approach. The best known result describing firms’ incentives to collude is the “Folk Theorem” which states that for sufficiently low discount rates almost any price may be sustained as the equilibrium outcome of a repeated game. While the “Folk Theorem” provides fairly general conditions under which tacit collusion may be sustained as an equilibrium, it says nothing about how firms behave in reality. The strategies used in the “Folk Theorem” are chosen because of their analytical ease and not because they describe firms’ collusive behaviour. See Fudenberg and Tirole (1991) for a discussion of the “Folk Theorem”. 3 While economic theory shows that concentration facilitates collusion, and thus predicts a positive relation between cartelization rate and market concentration, empirical evidence seems to contradict this result (Levenstein and Suslow, 2006). This gap be- tween the number of cartels predicted from a theoretical perspective and the number of cartels that appear in the empirical analy- sis may plausibly be explained by a sample-selection bias. Only cartels which, first, have been detected and which, secondly, were regarded as illegal by antitrust authorities or courts are contained in the sample. 4 Fonseca and Normann (2012) use a laboratory experiment to investigate the role of communication in sustaining collusion. They show that highly concentrated industries collude irrespective of communication. 3 lenged by a legal significance of evidence of communication. Economists can use observable variables such as prices, and their knowledge of the strategies employed by firms to infer collu- sion6 but have no instruments to prove whether firms collude with or without communication. From an incentive-based perspective, (illegal) communication appears to be of relative unim- portance: While non-enforceable communication might facilitate coordination on a particular collusive equilibrium,7 “talk is cheap” in the absence of effective enforcement mechanisms.8 It is, however, not out of economic naivety that antitrust law concentrates so much on evidence of communication in its struggle against collusion. Firstly, this reflects skepticism about whether instances of tacit collusion may be distinguished from oligopolistic competition with a degree of precision that suffices for forensic purposes. This concern may be associated with the so-called “indistinguishability problem” as put forward by Phlips (1996). He suggested that game theoretic arguments combined with the unavailability of some key data can make an economic based proof of collusion very difficult as something that looks like collusion might stem from a multiplicity of (indistinguishable) equilibria.9 Hence, the application of any legal instrument that addresses tacit collusion faces the challenge to prevent an unacceptable high number of false positives. Secondly, for purposes of antitrust enforcement it does not suffice to show that an observable market out- come emerged as the result of a collusive strategy. Antitrust remedies may not straightforwardly tackle firms because they charge “collusive”, i.e. supracompetitive, prices but must address specif- ic elements of firms’ market conduct which may be characterized as collusive. Without taking into account these issues, antitrust enforcement that tackles tacit collusion risks either unduly restricting market operators’ leeway to compete or to ultimately amounting to an instrument of price control. In the following, we outline an approach that addresses both these concerns, and hence provides the basis for an expansion of the law’s ambition to tackle tacit collusion. Oligopolistic interde- pendence as such and oligopolistic collusion are conceptually distinct. Tacit collusion arises from 5 See for comprehensive analyses of the use of economics to support cartel enforcement Werden (2004) and Kaplow (2011a). 6 One of the best known examples of economic detection of collusion is provided by the work of Christie and Schultz (1994). They detected collusion between Nasdaq market makers by comparing their bid-ask spread to the equivalent spread on the New York Stock Exchange. Christie and Schultz’ (1994) work had an impressive impact as it led to regulatory investigations by the Securities and Exchange Commission (SEC) and class action lawsuits that were settled for over $1 billion. 7 Genovese and Mullin (2001) provide narrative evidence of the role of communication for collusion in the Sugar Institute Case. They find that one key missing aspect in formal theories of collusion is the role for rich communication within the collusive agreement. 8 To use the words of Thomas Hobbes (1651/1959, chap. 14, p. 71), author of the Leviathan, “[…] the bonds of words are too weak to bridle mens ambition, avarice, anger, and other Passions, without the fear of some coercive Power […].” 9 For example, Peltzman (2000) confirms the long standing observation that prices respond faster to cost increases than to cost decreases in 77 consumer and 165 producer goods. This finding has been generally associated with collusion even though other reasons such as inventories and “menu costs” might lead to the same (indistinguishable) outcome. 4 decisions endogenous to the market by one or several firms which aim at reducing or eliminating competition. In contrast, oligopolistic interdependence stems from best response to market con- ditions (including other firms’ behavior) which favor non-competitive performance. Thus, while the market outcome might appear to be “indistinguishable,” the specific strategies that lead to the outcome differ significantly. The gist of our approach to identify collusive behavior lies in an identification of patterns of behavior used by firms to bring about or facilitate (tacit) collusion.10 Yet antitrust law must not simply infer the existence of a punishable (tacit) agreement from the insight that a certain market outcome is the result of a collusive strategy. Rather, it is essential to distinguish the active promotion of a collusive strategy by one firm from the passive (best re- sponse) alignment of competing firms. Consequently, antitrust enforcement should not concep- tualize such instances of collusive leader-follower behavior as an illegal coordination which would – with regard to the “followers” – result in punishing oligopolistic interdependence. Rather, anti- trust law should capture such instances of “unilateral collusion” only through considering as ille- gal the unilateral conduct that actively promotes the implementation of a collusive strategy. To effectively fight tacit collusion it appears therefore to be necessary to strengthen legal instruments that target the unilateral conduct that firms strategically employ to promote collusion. To illustrate our behavioral approach to tackling tacit collusion and to demonstrate the capacity of econometric evidence we refer to incidents on the Italian gasoline market. In Andreoli- Versbach and Franck (2013), hereafter AVF, we provide quantitative evidence of the means, i.e. specific pricing strategies, and the effects, i.e. higher prices, caused by the unilateral public an- nouncement of ENI, the market leader. On 6th October 2004 ENI announced a new pricing pol- icy which consisted of infrequent price variations (sticky pricing) and large price changes. Using daily firm level prices of gasoline in Italy and average weekly EU prices over the time period from January 2003 to May 2005, AVF show the effect of the new pricing policy. ENI increased the time lag between price changes from 6 to 16 days and increased the mean price change from 1% to 5.8%. After the policy change ENI did not change its price for 57 days irrespective of cost changes. Initially ENI’s competitors kept their short-run cost-based pricing and thus increased their prices following (lagged) cost increases.11 Once competitors started to align to ENI in mid- November 2004 a different pricing pattern emerged: sticky-leadership pricing. Each large price 10 In this respect, our approach is conceptually in line with Hay (2000, p. 128) who argues that “if there is to be a category of unlawful tacit collusion which is to be distinguished from classic oligopoly, the difference must lie […] on the specific elements of behavior that brought about that state of mind”. 11 Firms respond to cost shocks with some lags. While current costs decreased immediately after ENI’s policy, lagged costs in- creased and thus competitors increased their prices. See Figure I for a plot of daily prices and costs, i.e. Platts Cif. Med., around ENI’s new price policy announcement (first vertical line). 5 variation was matched by competitors and ENI endogenously emerged as the price leader in the market and coordinated price changes. While the first effect of the policy was to change the price interdependence in the Italian gasoline market this newly emerged tacit coordination had an addi- tional effect: a significant price increase. Using several estimation techniques AVF show that Ital- ian prices rose compared to EU prices after the new sticky leadership pricing emerged. Thus, the econometric analysis used to characterize pre and post policy pricing behavior and evaluate the effect of the new market conduct on the price level might provide solid “statistical” evidence that ENI’s unilateral commitment to a policy of sticky pricing has to be characterized as collusive. Against the background of these incidents on the Italian gasoline market we suggest that an im- plementation of sticky pricing along with large price changes should be prohibited under market conditions such as highly asymmetric market shares and high concentration where it may be ex- pected that price leadership will emerge as a price coordination mechanism and, thus, where such a pricing strategy will bring about collusion. Such an expansion of the legal tools to counter car- tels seems especially relevant for oligopolies where the structural market features favor collusion and at the same time communication might be less needed because of price and cost transparen- cy. The structure of the paper is as follows: Section 2 discusses the status quo of cartel enforcement which focuses on firms’ communication and the law’s difficulties with tackling tacit collusion. In section 3 we outline incidents on the Italian gasoline market as an illustration for how our ap- proach might be applied for purposes of antitrust enforcement. Section 4 describes the way to integrate quantitative evidence of collusion with antitrust law. Section 5 concludes. 2 On Collusion as a Legal Concept, its Limits in the Absence of Evidence of Collusive Communication, and the Reasons therefor Collusion allows competing firms to charge supra-competitive prices and entails negative welfare effects. Meta-studies on cartel overcharges show that the median cartel-price increase ranges be- tween 20 and 30 percent (Bolotova, 2009, and Connor, 2007). This is why antitrust law aims at inhibiting collusion and why the horizontal coordination of prices and quantities is considered a per-se violation of Section 1 Sherman Act or Article 101 Treaty on the Functioning of the Euro- pean Union (TFEU), respectively. Successful collusion requires inter alia an underlying – tacit or explicit – consensus on the terms of the cooperation. Thus, in order to counter collusion, it seems a logical step to regard such underlying understanding as illegal. However, the economic conception of a collusive agreement diverges significantly from the cor- responding legal concepts of “conspiracy” according to Section 1 Sherman Act or “agreement” 6 and “concerted practice” according to Article 101(1) TFEU.12 While the former focuses on firms’ incentives to engage in collusion and their strategies for sustaining a collusive equilibrium, the latter centers around the means to reach an understanding between firms. This divergent perspec- tive on collusion becomes apparent with regard to instances of tacit collusion, i.e. under circum- stances where no direct evidence of consensus between competing firms is available, such as written records or insider testimony. Though, as a matter of principle, both under the Sherman Act and the TFEU circumstantial evidence may suffice to demonstrate the existence of a “con- spiracy”13 or an “agreement”14 respectively, there are doctrinal limits in this regard if it comes to (supposedly) tacit collusion between competitors. In the words of the U.S. Supreme Court, “conspiracy” requires “that [the defendants] had a con- scious commitment to a common scheme designed to achieve an unlawful objective.”15 Reasona- bly, this may not be inferred from conscious parallelism alone.16 Rather a plaintiff has to produce additional evidence to prove that an observed parallel market conduct may not be considered the result of oligopolistic interdependence, but indeed forms part of a collusive strategy. Such so- called “plus factors” may encompass first, elements of industry structure that indicate that an industry is conducive to collaboration, second, conduct that appears irrational or inefficient ab- sent collusion, and third, additional factors such as industry performance (e.g. stabile market shares over time, supra-competitive pricing) or facilitating practices (e.g. exchange of infor- mation).17 While the U.S. Supreme Court has stated that plaintiffs can only survive summary judgment by presenting circumstantial evidence “that tends to exclude the possibility that the alleged conspirators acted independently,”18 the case law so far does not provide a taxonomy of plus factors which would allow us to determine which elements of evidence are required to infer an agreement. Thus, Gavil et al. (2008) concluded that “[…] decisions analyzing plus factors gen- erally have failed to establish a clear boundary between tacit agreements – to which Section 1 12 This conceptual divergence may also give rise to terminological misunderstandings between economists and lawyers. Throughout this paper we will indicate when we use terms such as “collusion” or “agreement” in their technical economic or legal meaning. 13 American Tobacco Co. v. United States, 328 U.S. 781, 809 (1946) (“No formal agreement is necessary to constitute an unlawful conspiracy”); Norfolk Monument Co. v. Woodlawn Memorial Gardens, Inc., 394 U.S. 700, 704 (1969) (“business behavior is admissible circumstantial evidence from which the fact finder may infer agreement”). 14 CFI, 26.10.2000, Case T-41/96 Bayer v Commission [2000] ECR II-3383 para. 69; confirmed on appeal by the ECJ, 6.1.2004, Joined Cases C-2/01 P and C-3/01 P Bundesverband der Arzneimittel-Importeure and Commission v Bayer [2004] ECR I-23, para. 97. 15 Monsanto Co. v. Spray-Rite Servs. Corp., 465 U.S. 752, 768 (1984). While Monsanto involved a vertical collaboration, the Court soon after adopted the same reasoning also in a horizontal case, see Matsushita Electronics Industries Co. v. Zenith Radio Corp., 475 U.S. 574, 588 (1986). 16 See e.g., Theatre Enters., Inc. v. Paramount Film Distrib. Corp., 346 U.S. 537, 541 (1954). 17 See for an overview Gavil et al. (2008), pp. 310-311. 18 Matsushita Electronics Industries Co. v. Zenith Radio Corp., 475 U.S. 574, 588 (1986) (quoting Monsanto Co. v. Spray-Rite Servs. Corp., 465 U.S. 752, 764 (1984)). 7 applies – and parallel pricing stemming from oligopolistic interdependence […]. This condition makes judgments about future litigation outcomes unpredictable.”19 While the European Court of Justice (ECJ) considers it generally conceivable that consent to an agreement may be inferred from circumstantial evidence,20 the Court is reluctant to infer an “agreement” between competitors from their market conduct alone, notwithstanding the pres- ence of certain “plus factors.” Given the current state of the jurisprudence, it appears that in the absence of direct evidence of collusion the Court does not presume the existence of an “agree- ment” even if one has proved that observed parallel market conduct was an expression of (tacit) collusion rather than of oligopolistic interdependence as such. This has been reaffirmed by a de- cision on the doctrine of “collective dominance” under Article 102 TFEU where the ECJ implic- itly approved that tacit collusion per se may not fall under Article 101(1) TFEU: “[u]nless they can form a shared tacit understanding of the terms of the coordination, competitors might resort to practices that are prohibited by Article [101 TFEU] in order to be able to adopt a common policy on the market.”21 However, where tacit collusion has been induced by facilitating practices such as, for example, an exchange of information, it may come under Article 101(1) TFEU as an illegal “concerted practice”. In this regard, the ECJ drew a line: On the one hand, by assigning market operators the legal leeway to “adapt themselves intelligently to the […] conduct of their competi- tors” the Court signaled that mere passive alignment would not be treated as an illegal form of co- ordination. On the other hand, the Court submitted that a strategy that actively aims at aligning competitors’ market conduct may fall under Article 101(1) TFEU.22 Thus, to implement this standard it is essential to identify elements of behavior that promote (tacit) collusion. This insight into legal concepts of coordination reveals ambiguities and restrictions with regard to tacit collusion. It raises the question why the law finds it so difficult to cope with this phenome- non, given that it seems uncontroversial in terms of competition policy that tacit collusion on prices and quantities should be prevented as rigorously as collusion based on explicit consensus. To begin with, the respective judicial definitions of “conspiracy” and “agreement” do not restrict 19 See also Kaplow (2011b), p. 816, who concludes after an extensive analysis of the concept of agreement in antitrust law: “[…] this Article does not come close to demonstrating that it would be good policy to proscribe and highly penalize all instances in which interdependent oligopolistic behavior appears to occur. The design of optimal policy is not dictated by definitions but rather by direct assessment of the consequences of different regulatory approaches.” 20 Accordingly, the Court infers a tacit approval of a collusive initiative from the attendance of a meeting where an anticompetitive agreement was concluded, see ECJ, 28.6.2005, Joined Cases C-189/02 P, C-202/02 P, C-205/02 P to C-208/02 P and C-213/02 P Dansk Rørindustrie A/S and others v Commission [2005] ECR I-5425 para. 143: “That complicity constitutes a passive mode of participation in the infringement which is therefore capable of rendering the undertaking liable in the context of a single agree- ment […].” 21 ECJ, 10.7.2008, Case C-413/06 P Bertelsmann and Sony Corporation of America v Impala [2008] I-4951, para. 123. 22 ECJ, 16.12.1975, Joined Cases 40 to 48, 50, 54 to 56, 111, 113 and 114/73, Suiker Unie and others v Commission [1975] ECR 1663 paras. 173-174; ECJ, 14.7.1981, Case 172/80 Züchner v Bayerische Vereinsbank [1981] ECR 2021, paras. 12-14. 8 these concepts in a way that would exclude collusion which has been sustained tacitly. Whatever the rhetoric of the courts might be when they characterize the requirements of an agreement – typically they refer to a need to show a “meeting of minds,”23 a “joint intention”24 or a “concur- rence of wills”25 –, the respective antitrust law concepts have to be defined strictly instrumentally. Hence it is, first, the underlying policy to contain as far as possible any kind of welfare-reducing collusion and, second, the role a legal intervention and, in particular, a prohibition of agreements between competitors may feasibly play in this regard, that determine which behavior should be regarded as illegal. Part of the law’s problem in coping with tacit collusion lies with the difficulty to distinguish col- lusion from oligopolistic interdependence as the latter may also result in suspiciously parallel market conduct and supra-competitive prices. This problem is addressed by the requirement of “plus factors” which – in addition to parallel pricing – are meant to indicate collusion, such as market conduct which may reasonably only be explained as part of a collusive strategy.26 From this perspective, the problem of distinguishing oligopolistic collusion from oligopolistic competi- tion comes essentially down to a question of error costs: by defining the “critical mass” of plus factors required to infer an illegal coordination, courts strike a balance between the ambition to contain (tacit) collusion and the risk of producing false positives.27 However, in particular the ECJ’s categorical reluctance to infer an agreement in cases of mere tacit collusion suggests that there is more to the law’s difficulties to cope with tacit collusion than the problem of multiple (indistinguishable) equilibria and the issue of reaching an acceptable de- gree of error costs in this regard. Legal standards and remedies that are supposed to influence market conduct in order to guarantee effective competition may not simply prohibit an undesired economic condition such as a collusive equilibrium and punish firms because they charge “collu- sive” prices. Such a policy effectively meant nothing other than price control. This unwelcome consequence is prevented as antitrust standards and remedies relate to individual behavior and define which acts or omissions are required or prohibited. When authorities or private plaintiffs order a firm to bring an infringement to an end or seek to obtain injunctions before a court, it is 23 American Tobacco Co. v. United States, 328 U.S. 781, 810 (1946); Copperweld Corp. v. Independence Tube Corp., 467 U.S. 752, 771 (1984). 24 ECJ, 15.7.1970, Case 41/69 ACF Chemiefarma [1970] ECR 661 para. 112. 25 CFI, 8.7.2008 AC-Treuhand [2008] ECR II-1501, para. 118. 26 This is presumed if, for example, a certain conduct “is so perilous when not imitated and imitation so uncertain that no reason- able actor would so act, then parallel action does imply some exchange of commitments or at least some comforting assurances connoting a traditional conspiracy”, Areeda and Hovenkamp (2010), §1415c, p. 107 with reference to Blomkest Fertilizer v. Potash Corp., 203 F.3d 1028, 1037 (8th Cir. 2000). 27 See Posner (2001), p. 99: “[…] a damages judgment in a tacit collusion case would promote competition at a tolerable cost in legal uncertainty and judicial supervision.” 9 already the remedy’s behavioral nature that requires a specification of elements of conduct that violate antitrust law. The intended deterrent effect of concurring remedies such as imposing fines or damages likewise depends on whether market operators are in a position to foresee what con- duct they may be sanctioned for, and how they are expected to behave to avoid sanctions. This appears particularly challenging where an undesired economic effect or market condition is the consequence of the interdependent behavior of several market actors.28 But once again: if the elements of behavior that bring about a collusive equilibrium remain unclear, any legal interven- tion may ultimately amount to a price control by antitrust authorities or courts. Furthermore, with regard to criminal and quasi-criminal sanctions it is required by the principle of culpability29 and the need to prove intent30 or negligence,31 respectively, that antitrust enforcement ensures that market operators may anticipate their legal leeway and addresses certain modes of behavior rather than an economic effect or condition. Thus, the key to overcoming the law’s difficulties to counter tacit collusion lies in an approach which identifies specific elements of behavior whose object or effect it is to bring about or facili- tate collusion. Such an approach has a chance for success as market operators that seek to im- plement a collusive strategy need to adjust their market conduct to reach an optimal and stable collusive equilibrium. Even in oligopolistic markets that are characterized by features that facili- tate tacit collusion, prices and other parameters have to be adjusted according to an underlying (tacit) agreement, and the need for such adjustments may lead firms to resort to a certain behav- ior that may be identified as serving a collusive strategy. Empirical and theoretical research32 on how cartels behave provides solid test hypotheses to identify such elements of collusive behavior. 28 Cf., e.g., E.I. Du Pont De Nemours & Co. v. FTC (Ethyl), 729 F.2d 128, 139 (2d Cir. 1984): “In view of this patent uncertainty the [Federal Trade] Commission owes a duty to define the conditions under which conduct claimed to facilitate price uniformity would be unfair so that businesses will have an inkling as to what they can lawfully do […]. The Commission’s decision in the present case does not provide any guidelines; it would require each producer not only to assess the general conduct of the anti- knock business but also that of each of its competitors and the reaction of each to the other, which would be virtually impossi- ble.” 29 Under European law, Article 7(1) ECHR enshrines the principle that offences und punishments are to be strictly defined by law, see on the relevance of this norm as to fines in EU Competition Law ECJ, 28.6.2005, Joined Cases C-189/02 P, C-202/02 P, C-205/02 P to C-208/02 P and C-213/02 P Dansk Rørindustrie A/S and others v Commission [2005] ECR I-5425 para. 202. 30 Cf. 438 U.S. 422, 435 (1978): “We agree with the Court of Appeals that an effect on prices, without more, will not support a criminal conviction under the Sherman Act […]. [A] defendant's state of mind or intent is an element of a criminal antitrust of- fense which must be established by evidence and inferences drawn therefrom, and cannot be taken from the trier of fact through reliance on a legal presumption of wrongful intent from proof of an effect on prices.” As to the required standard of intent the Court concluded id., at 444, “that action undertaken with knowledge of its probable consequences and having the requisite anti- competitive effects can be a sufficient predicate for a finding of criminal liability under the antitrust laws.” 31 See Article 23(2)(a) Council Regulation (EC) No 1/2003 of 16 December 2002 on the implementation of the rules on competi- tion laid down in Articles 81 and 82 of the Treaty, Official Journal L 1, 04.01.2003, p. 1-25. 32 For a meta-study on cartels’ features see Harrington (2006) and Levenstein and Suslow (2006). For a survey on price fixing in particular see Hay and Kelley (1974). For an analysis of the determinants of cartel duration see Levenstein and Suslow (2011).