TY - JOUR
AU - De,, Oindrila
AB - Abstract The concept of a cartel ringleader has specific legal meaning but has received relatively little attention in the economic literature of cartels. This article draws on a sample of 89 European cartels to identify how often ringleaders exist, who they are and what they do. It argues that ringleaders are more likely where the traditional ‘cartel problems’ are likely to be most acute, i.e. with larger numbers of members exhibiting substantial size asymmetries. This is confirmed for price‐fixing and bid‐rigging cartels and is especially pronounced where the ringleader displays ‘aggressive’ as opposed to merely ‘organisational’ behaviour. The concept of a cartel ringleader has a very specific legal relevance for the fining policies of competition authorities (CA) but surprisingly it has received relatively little attention in the economics literature. The purpose of this article is to argue that the ringleader can be interpreted as an organisational ‘solution’ to the classic cartel problems, in those circumstances where those problems are likely to be most acute, i.e. for cartels which comprise relatively large numbers of asymmetric members. Starting from a close reading of the European Commission’s decision documents for cartels detected over the years 1990–2008, it first identifies how frequently they exist (in about 20% of cases), who they are and what are their specific roles and responsibilities.1 Using a logit model, it then finds that they are indeed significantly more likely in those cartels with larger numbers of members who are asymmetrically sized, especially for price‐fixing and bid‐rigging cartels. We argue that this helps explain one of the main empirical puzzles concerning cartels: the conventional theoretical wisdom is that collusion is more likely in markets in which firms are few and symmetric, but many real‐world cartels actually involve relatively large numbers of firms, who often exhibit considerable size asymmetries. That collusion is most likely with fewness and symmetry of firms is a conventional wisdom. In policy terms it forms part of standard structural screening methods for the potential presence of cartels and the checklist of necessary conditions for mergers to have coordinated effects. It is also firmly embedded in the academic Industrial Organisation literature. This dates from at least the early Structure‐Conduct‐Performance paradigm which often merely asserted that collusion is more likely in more concentrated markets with a small number of symmetrical firms. This was subsequently formalised, if qualified, by game theoretic understandings (Ivaldi et al., 2003): coordination among potential cartel members is easier, in terms of a shared common focal point with only a small number of similar firms; and incentive compatibility constraints are less likely to be satisfied as firm numbers and asymmetries increase.2 However, evidence from the empirical literature on cartels offers only limited support for these propositions. Levenstein and Suslow (2006, pp. 85–6) suggest that ‘The empirical evidence on the relationship between industry concentration and cartel stability is mixed. All else equal, concentration undoubtedly aids cartel stability… But organizational responses, such as industry associations, can overcome the challenges posed in forming a cartel in an un‐concentrated industry, and cartels can, by increasing profitability, allow marginal firms to survive and so decrease concentration’. In a number of different cross‐section studies3 the sample mean number of cartel members varies between 7 and 29, and concentration is often only moderately high (the average 4‐firm concentration ratio ranges from 43% to 75%). As far as we know, there is no previous quantitative cross‐section study of size asymmetries within cartels, but Grout and Sonderegger (2005) report considerable heterogeneity in market shares of members in their case studies of EC cartels. De (2010), for roughly the same sample as this article, also reports a relatively large number of members in many cartels (on average 6.85 for 98 cartels) and often significant levels of asymmetry among firms. Part of the explanation for this apparent disconnect between theory and empirics might lie in sample selection: all the evidence necessarily relates to detected cartels; this leaves open the possibility that it is the more stable undetected cartels which may typically be the smaller number symmetrical cases. Equally, however, it is often overlooked that most repeated game theory is better suited as a depiction of tacit, rather than explicit, collusion. Given the explicit communication and formal agreements that characterise real‐world cartels, constraints on their formation and stability may be less binding, even with larger numbers and asymmetric firms. The empirical case study literature on prosecuted cartels often emphasises that it is the organisational characteristics of cartels which impact most on their success.4 This literature reveals that cartels often use sophisticated mechanisms to coordinate, monitor and enforce agreements. However, it rarely, if ever, discusses the potential role of the ringleader as one of these organisational characteristics. In fact, the term ‘ringleader’ rarely appears in any of the seminal theoretical and empirical surveys of the cartel literature. Historically, perhaps the most relevant literature is that on price leadership in oligopoly. Following Markham (1951) this identified three types: dominant firm, barometric and collusive. In the dominant firm case, with a dominant Stackleberg leader and competitive fringe, the leader is usually the most efficient firm. In the barometric case, the leader is sometimes merely a ‘bell‐weather’ – the firm which most closely conforms to the representative firm in terms of costs and products; in other cases (Cooper, 1997), it may be the best informed firm. Collusive leadership has since attracted relatively little attention in the theoretical literature. Markham himself clearly saw this as an alternative to a full‐fledged cartel, where public pre‐announcements are used as a coordinating device instead of explicit agreement. Subsequently, Rotemberg and Saloner (1990) showed how, with asymmetric information, collusive leadership might emerge where the leader is the firm with perfect information on demand. In other less formal treatments, the role of leader is delegated almost arbitrarily as a matter of convenience, or shared around with the leader changing from time to time (Scherer and Ross, 1990, pp. 166–70). More recently, two papers provide results given asymmetries. Mouraviev and Rey (2011) show that with asymmetric costs price leadership can sometimes enhance the profitability of collusion. Here, the least efficient firm must be the leader because, without leadership (i.e. simultaneous pricing) it is this firm’s incentive constraint which is binding. On the other hand, in Ishibashi (2008), firms with larger capacity are best suited for the role. However, these models are still best interpreted, as was Markham’s original work, as descriptions of tacit collusion – what is possible without formal agreement. The concept of a ‘leader’ features more in the general economics literature on organised crime, where it is often argued that leaders should be punished more harshly than other members of a conspiracy (Silving, 1967; Robinson, 1987) – not because of the gravity of their offence but because they have committed multiple offences, by also organising and/or instigating the offence (Elder, 2010). More specifically in the context of cartel law, the ringleader has begun to attract increasing attention, especially the possibility that ringleaders be treated asymmetrically in leniency programmes. In part, this has been fuelled by a transatlantic policy difference – ringleaders are eligible for fine reduction and even immunity under the EC’s leniency programme but not in the US.5 Leslie (2006) suggests two conflicting effects of ringleader exclusion from leniency: on the one hand, excluding ringleaders from the programme will reduce the distrust between the leader and other members as the leader has no incentive to break the agreement; on the other hand, increasing the penalty for ringleaders should increase deterrence. In a more formal model, Herre and Rasch (2009) find that if the probability of detection is low, then a non‐discriminatory leniency policy should be preferred as additional information provided by the ringleader will facilitate prosecution; however, with high probability, ringleader exclusion creates an asymmetry among firms, which makes collusion more difficult to sustain. There are few empirical studies of cartel ringleaders. However, two otherwise largely theoretical studies report some descriptive facts which are relevant for our purposes. Ganslandt et al. (2012) present two observations based on a sample of 43 EC cartel cases: (i) in 10 (23%), the EC found evidence of ringleaders; (ii) the size of the second largest firm was on average 70% of the size of the largest firm. From this they conclude ‘These observations raise the question of why so many cartels are asymmetric and have ringleaders’ (Ganslandt et al., 2012, p. 766). However, they do not establish any statistical association because they do not report how many of the ringleaders actually occurred in the cartels which were asymmetric. On the basis of this therefore rather casual empiricism, they construct a theoretical model which predicts that a certain amount of asymmetry within cartels may actually facilitate collusion. However, it is based on two restrictive assumptions: cartels can only be formed if there is a ringleader; there are indivisible fixed costs of collusion which are not easily shared and must be borne by the leader.6 In these circumstances cartels will only exist where there are moderate asymmetries within the cartel: on the one hand, symmetry means that there will be no one firm with the incentive to cover the indivisible cost, on the other hand, too much asymmetry will give the smallest firm a strong incentive to deviate. Bos and Wandschneider (2011) motivate their model of the effects of ringleader exclusion on collusive prices, with three stylised facts, also based on EC cartel cases: cartels often have more than one ringleader, ringleaders are often the leading member(s), and their roles vary across different cartels. Each of these mirrors similar findings by De (2010, ch. 5), which are now presented in the first part of our Section 1 below.7 The main objective of this article is, however, different. Rather than assuming the existence of a ringleader, as do Ganslandt et al. (2012), or to examine the implications for price of ringleader exclusion from leniency, our purpose is to identify the conditions under which ringleaders are actually observed and to test whether they are more common in cartels with many, and asymmetric, members, i.e. whether they can be interpreted as an organisational solution to the classic cartel problems. The structure of the article is as follows. The next Section draws on the previous literature to derive four propositions about the existence and nature of ringleaders. Section 1 describes the sample of EU cartels and derives some relevant stylised facts. Building on these, Section 2 conducts logit and multinomial logit analysis of the probability that a cartel will have a ringleader of different types – in terms of the number of members and their asymmetries. Section 3 summarises and concludes. 1. The Ringleader’s Role in Solving the Cartel Problems This Section recalls from the literature what are usually described as the ‘cartel problems’, and why they are most likely to be acute in cases with a large number of potentially asymmetric cartel members. It asks how far these problems can be alleviated by a ringleader and suggests four empirically verifiable propositions. The difficulties inherent in forming and maintaining a cartel, referred to seminally by Osborne (1976) as the ‘cartel problems’, are to: locate the contract surface; choose a point on that surface; to detect; and to deter potential cheating. These problems are likely to be more pronounced with a large number of members. With many firms in the cartel, a price cut will yield a large increase in market share for the deviator, which is likely to outweigh the losses in subsequent punishment periods. Equally, the larger the number of firms, the smaller is each firm’s share of collusive profit and this reduces the incentive to collude. For a cartel using output quotas, a large number of firms can be problematic if the cartel is not all inclusive; Selten (1973) suggests that the maximum number is six. As it may be more attractive not to join the cartel but to free ride on the cartel price, the cartel will be stable if and only if its members do not find it profitable to exit (internal stability) and the players outside the cartel do not find it profitable to join the cartel (external stability).8 It is also usually recognised in the theoretical literature that asymmetry among firms will generally hinder collusion. The coordination problem becomes harder when firms have divergent preferences and a natural focal point is hard to find. A more equal distribution of assets relaxes the incentive constraints for both large (more efficient) and small (less efficient) firms, and thereby facilitates collusion (Motta, 2004). The majority of papers on asymmetry focus on cost asymmetry9 but with some exceptions which consider asymmetry in capacity, or the discount factor, or product differentiation.10 Unless the least cost firm produces all the output (which implies illegal side payments which are difficult to conceal), Pareto efficient allocation requires that the cartel price will be even higher than the monopoly price of the most efficient firm, and the incentive for it to deviate is high. Similarly, asymmetric capacity also reduces the sustainability of collusion. However, here, the literature differs on which firms have the highest incentive to deviate. Compte et al. (2002) show that with market shares proportional to capacity, if the largest firm is too large, it is the capacity of this firm which plays an important role as it has the highest incentive to cheat. On the other hand, Vasconcelos (2005) finds that with an unequal distribution of capacity, the small firms gain more from deviation from the collusive equilibrium whereas the large firms gain more from deviating from the punishment phase. However, Bos and Harrington (2010) show that, where cartels are not all inclusive, the capacity of the medium‐sized firms will play the most important role. Similarly, Kuhn and Rimler (2006) show that although a decrease in product varieties facilitates collusion, consolidation of varieties (the same number of varieties in the hands of smaller number of firms) also facilitates collusion. For firms with different discount factors, the same collusive price is sustainable as long as the average discount factor is sufficiently high (Harrington, 1989). But output quotas need to be distributed in accordance with firms’ discount factors. There is also a theoretical literature on how cartels might handle enforcement of asymmetric arrangements – mostly in the presence of private information about costs (Athey and Bagwell, 2001, 2008) or price (Harrington and Skrzypacz, 2007) – using market share favours to the high cost firm to encourage truthful reporting of costs, or a compensation mechanism on the basis of reported sales respectively. Empirical studies also tend to focus on the organisational mechanisms which cartels use to cope with these problems. However, none of these studies explicitly identifies ringleaders as one of them. Therefore, we next explore the potential role of ringleaders by applying a widely used typology of various cartel organisational activities: (i) formation, instigation and approaching potential members, (ii) organisation and monitoring, (iii) the nature of the agreement, and (iv) coercion and enforcement. Although the CAs do not always tightly or formally define the term cartel ‘ringleader’, the European Commission (2006)11 now distinguishes two broad types of activity in its decisions – ‘instigation’ and ‘leadership’. An instigator is involved in the initial birth and/or subsequent enlargement of the cartel (which coincides with (i) above), and ‘leadership’ which implies a proactive and leading role, entailing coercion and leadership in internal punishment mechanisms (which coincides with (ii) and (iv)). We will argue that the nature of the agreement (iii) to a large extent determines whether or not a ringleader is required for the organisation and monitoring purposes, and to some extent, for the purpose of enforcement. 1.1. Formation, Organising and Monitoring Any conspiracy, whatever the context, is born from an individual/group’s realisation that it may be practicable and profitable; thus any cartel requires somebody to take the initial initiative – to at least instigate discussions. In that sense, probably every cartel requires one or more ringleaders. On the other hand, where there are only a few conspirators, all members of the group may be in on the conspiracy almost from the start and there is no need for any single instigating leader. This suggests that a fruitful way of expressing the question is to ask ‘under what circumstances, will a conspiratorial group choose to delegate or recognise (sometimes reluctantly perhaps) that one of its members should take on this role? This can then be viewed as an alternative to a default which is that all members share the responsibility, in which case there is no ringleader. This framing of the question has the policy corollary that all firms should be fined equally, unless one or more of them can be singled out for disproportionately larger penalties. Clearly, the advantages of delegation should increase rapidly with N, the number of members. With multilateral agreements and exchanges of information etc. (i.e. bilateral contacts), the number of links will increase quadratically with N but only linearly if all members need deal only with a single leader. Similarly, organising meetings and monitoring are largely administrative functions which might also be delegated to any member of the group or shared around. Again the logic of delegating to a ringleader is obvious as the size of the group increases. On the other hand, there is evidence that activities like convening and conducting meetings, exchanging information on prices or sales for monitoring purposes are sometimes conducted through the offices of an appropriate trade association, either alongside or instead of a ringleader (e.g. Lysine).12 But equally there are many examples where a cartel’s designated ringleaders played the same role (e.g. Carbonless paper).13 1.2. Nature of the Agreement Case study evidence14 reveals heterogeneity between cartels concerning the focus of their conspiracy, i.e. the nature of the agreement and what they coordinate on, and it is probable that the need for a ringleader will vary depending on the nature of the agreement. In broad terms, agreements relate to one or more of price‐fixing, bid rigging or market sharing, where market sharing can entail one or more of territorial allocation, customer allocation and/or setting market share quotas. The vast majority of cartels include an element of price‐fixing (indeed, in Harrington (2006, p. 6), all cartels do), but market share agreements also occur in many cartels. As such, coordination on price is a very common task for cartels, with the implication that a price leader will often be required. In fact, price coordination may require sophisticated mechanisms on the part of cartel members where the product is complex and differentiated. As discussed earlier, the traditional Industrial Organisation literature identified three types of price leadership: dominant firm (à la Stackelberg), barometric and collusive. Insofar as these also occur within cartels, the term ringleader has an obvious meaning for Stackleberg and the leader would typically have a much larger market share, perhaps deriving from a significant cost or capacity advantage. On the other hand, for collusive or barometric leadership, the literature offers mixed results: in some models, the leader will be the largest firm, in others the smallest, and in yet others the ‘representative’ firm. Another possible outcome is that there could be a number of ringleaders, with the role shared around, with perhaps all firms involved at some point in the activity, in which case there may be no unique ringleader who can be singled out by the CA for a heavier penalty. However, other types of agreement, once in place, should require relatively trivial monitoring with little likelihood of punishment necessary. This is likely particularly for market sharing of the territorial or customer allocation forms which will be more or less self‐enforcing. The geographical scope of the cartel might also be important. The case study literature often differentiates national from international cartels, or, in the context of the European Union, global as opposed to intra‐European cartels. As the agreement becomes more global, this may also increase the benefits from having a leader, or even multiple leaders, each representing, monitoring and coordinating the members from its own region. 1.3. Aggressive Leadership, Coercion and Enforcement Price agreements and market share quotas are intrinsically more prone to cheating and therefore monitoring and punishment – credibly threatened and/or enacted. Reconciling differential incentives may often require that one firm cedes more than others and this requires that there is a firm that is capable/willing to make concessions and to have the capability and credibility to threaten punishment on the one hand, or to compensate and even make side payments on the other hand. Depending on the structure of the cartel (in terms of market shares, capacity, differential costs), this might point directly to the leading firm(s). Coercion and enforcement activities entail a much more aggressive role for the ringleader – much more than just a facilitator. In summary, the precise roles of the ringleader, where necessary, will vary from case to case. Sometimes these may be largely facilitating, i.e. approaching new members, monitoring and convening and conducting meetings but sometimes they may be more aggressive – dictating price, coercing and, where necessary, leading in the compensation schemes and punishment activities of the cartel. From the above discussion, we propose four testable propositions. 1. Proposition Cartels may have no ringleader, or no unique ringleader. The default is that all firms take the responsibility of leadership collectively, in which case all are equally culpable. Alternatively, leadership may be shared by a subset. A ringleader may be less necessary where activities can be conducted through a trade association but more necessary where members are globally dispersed. 2. Proposition The likelihood of a ringleader will vary with the nature of the agreement. A ringleader may be required where the agreement involves coordinated pricing and/or bid rigging but unnecessary where there is only market sharing which is confined to allocation of territories/customers, both of which require less monitoring and enforcement. 3. Proposition Ringleaders are more likely in cartels with more members who are more asymmetric. With larger numbers and asymmetries, the traditional cartel problems are likely to be most acute and a leader may be a necessary organisational response. 4. Proposition ‘Aggressive’ ringleaders are likely to enjoy some size‐related advantage over other members. Following the above, but also Ganslandt et al. (2012), to the extent that some ringleader activities involve an indivisible fixed cost and are not merely organisational (e.g. buyouts, coercion and aggressive punishment strategies) these may require a ‘deep pocket’ and or size‐related (e.g. cost or capacity) advantage and the largest member is most likely to be able to cover this cost while retaining the incentive to collude. 2. The Sample and Descriptive Findings These propositions are tested for a sample of 89 prosecuted EU cartels15 – horizontal agreements infringing Article 101(1) or Article 65(1) of the ECSC treaty – over the period 1990–2008. The primary source of information is the European Commission’s (EC) final prohibition decisions published in the Official Journal (L series), summary decisions (C series) and press releases published on the DG Competition website.16 This time period includes years both before and after the EC introduced its first leniency notice in 1996 (European Commission, 1996), which was subsequently revised in 2002 and 2006. Transparent rules specifying increased fines for ringleaders also came into force with the introduction of the leniency notice. Initially, ringleaders were excluded from applying for leniency but, in 2002, this was changed so that they were also allowed to apply for reduced fines, but not complete immunity. Thus, our chosen start year pre‐dates the formal legal recognition of ringleaders; this is because it is clear from the case documents that the possibility of ringleaders was recognised even in the early 1990s (the first ringleader was identified in 1994).17 The end year was the most recent one available at the time of the research, bearing in mind the time lag between the investigation and availability of the report in the public domain. From a close reading of the case documents, a number of descriptive findings can be drawn. Some of these are also documented by Bos and Wandschneider (2011), whose study was conducted quite independently from ours. 2.1. Frequency of Ringleaders In 19 of these 89 cartels (21%), the EC identified a ringleader or ringleaders (Table 1).18 It should be underlined that these are the cases in which the EC felt sufficiently confident to name a ringleader explicitly. In a further eight cases, it reported allegations of ring‐leadership from other cartel members, but was not satisfied that it could establish sufficiently hard corroborating evidence (for example, industrial thread). Arguably then, the incidence of ringleaders may be higher than recorded here. Issues of sample selection are discussed in subsection 2.4. Table 1 Incidence of Cartels and Ringleaders Over Time . 1990–3 . 1994–6 . 1997–9 . 2000–2 . 2003–5 . 2006–8 . Total 1990–2008 . All cartels 11 10 8 21 19 20 89 RL cases 0 2 3 8 2 4 19 . 1990–3 . 1994–6 . 1997–9 . 2000–2 . 2003–5 . 2006–8 . Total 1990–2008 . All cartels 11 10 8 21 19 20 89 RL cases 0 2 3 8 2 4 19 Open in new tab Table 1 Incidence of Cartels and Ringleaders Over Time . 1990–3 . 1994–6 . 1997–9 . 2000–2 . 2003–5 . 2006–8 . Total 1990–2008 . All cartels 11 10 8 21 19 20 89 RL cases 0 2 3 8 2 4 19 . 1990–3 . 1994–6 . 1997–9 . 2000–2 . 2003–5 . 2006–8 . Total 1990–2008 . All cartels 11 10 8 21 19 20 89 RL cases 0 2 3 8 2 4 19 Open in new tab Over the period, the proportion of cartels having ringleaders increased steadily up to 2001 before declining somewhat thereafter (see Figure 1 and Table 1, which depict the cumulative probability that ringleaders were identified). Fig. 1. Open in new tabDownload slide Cumulative Probability of Ringleader Fig. 1. Open in new tabDownload slide Cumulative Probability of Ringleader 2.2. What Ringleaders Do Table 2 summarises the ringleaders’ activities that the Commission observed and used to justify its decisions.19 In most cases, the ringleader undertook more than one of these activities – on average more than three. Table 2 Ringleader Activities Types of activity . Frequencies . Organisational Instigation/approaching new members 14 Administrative of which: 17 Convening/conducting meetings 16 Monitoring/secretariat 12 Acting as intermediary 7 Aggressive Dominant price leadership 9 Coercion/threatening/enforcing 5 Classification of cartels Organisational ringleader(s) only 8 Aggressive ringleader(s) 11 No ringleader 70 Types of activity . Frequencies . Organisational Instigation/approaching new members 14 Administrative of which: 17 Convening/conducting meetings 16 Monitoring/secretariat 12 Acting as intermediary 7 Aggressive Dominant price leadership 9 Coercion/threatening/enforcing 5 Classification of cartels Organisational ringleader(s) only 8 Aggressive ringleader(s) 11 No ringleader 70 Notes Most ringleaders engage in more than one activity. A cartel is classified as organisational only, if its ringleaders engage only in organisational activities; most aggressive ringleaders are also organisational. Open in new tab Table 2 Ringleader Activities Types of activity . Frequencies . Organisational Instigation/approaching new members 14 Administrative of which: 17 Convening/conducting meetings 16 Monitoring/secretariat 12 Acting as intermediary 7 Aggressive Dominant price leadership 9 Coercion/threatening/enforcing 5 Classification of cartels Organisational ringleader(s) only 8 Aggressive ringleader(s) 11 No ringleader 70 Types of activity . Frequencies . Organisational Instigation/approaching new members 14 Administrative of which: 17 Convening/conducting meetings 16 Monitoring/secretariat 12 Acting as intermediary 7 Aggressive Dominant price leadership 9 Coercion/threatening/enforcing 5 Classification of cartels Organisational ringleader(s) only 8 Aggressive ringleader(s) 11 No ringleader 70 Notes Most ringleaders engage in more than one activity. A cartel is classified as organisational only, if its ringleaders engage only in organisational activities; most aggressive ringleaders are also organisational. Open in new tab Using the distinction drawn in the previous Section, these are classified as organisational or aggressive. In the large majority (14 out of 19), the ringleader(s) is found to have instigated the agreement,20 although in no case is instigation the sole activity of the leader, and in 17, the ringleaders are involved in some form of administrative functions. Thus, in nearly all cases, ringleader(s) are found to have instigated and been responsible for the administrative functions of convening and conducting meetings, and of monitoring. The more aggressive functions, such as dominant price leadership and coercion/threats/enforcement are less frequent (9 and 5 respectively) but, as implied by the above, are usually conducted alongside the above organisational activities. It follows then that most aggressive ringleaders also engage in organisational activities but the reverse is not so. Examples of aggressive cartels include lysine (in which Ajinamoto and ADM together engaged in all six activities); and carbonless paper and sorbates, in which ringleaders were actively aggressive in both in price leadership and coercion/enforcement. On the other hand, in paraffin wax, sodium gluconate and gas‐insulated switchgear, the ringleaders were completely administrative in nature, concentrating only on convening/conducting meetings, playing the role of intermediary and monitoring the agreements. The EC decision documents also provide some indirect evidence of what ringleaders do to enforce agreement, by using, inter alia, compensation schemes and price wars (Table 3). Table 3 Types of Enforcement: Ringleader Versus Non‐ringleader Cartels . All cartels . Ringleader cartels . Non‐ringleader cartels . Compensation schemes 32 (36) 13 (68) 19 (27) Price wars 18 (20) 7 (37) 11 (16) Total 89 19 70 . All cartels . Ringleader cartels . Non‐ringleader cartels . Compensation schemes 32 (36) 13 (68) 19 (27) Price wars 18 (20) 7 (37) 11 (16) Total 89 19 70 Notes Percentages of column totals shown in parentheses. Open in new tab Table 3 Types of Enforcement: Ringleader Versus Non‐ringleader Cartels . All cartels . Ringleader cartels . Non‐ringleader cartels . Compensation schemes 32 (36) 13 (68) 19 (27) Price wars 18 (20) 7 (37) 11 (16) Total 89 19 70 . All cartels . Ringleader cartels . Non‐ringleader cartels . Compensation schemes 32 (36) 13 (68) 19 (27) Price wars 18 (20) 7 (37) 11 (16) Total 89 19 70 Notes Percentages of column totals shown in parentheses. Open in new tab Compensation can take various forms, but broadly speaking refers to schemes whereby firms that sell more than their allocated quota compensate the other parties ex post. Price wars can sometimes be the result of concerted behaviour by a set of punishing firms or sometimes more of individual retaliation to a specific violation of the agreement. Both are significantly more likely to occur where a ringleader exists – more than two thirds of ringleaders organise some form of compensation scheme, while only a quarter of non‐ringleader cartels have such schemes; and price wars are twice as likely in cartels with a ringleader than in those without. 2.3. Who Ringleaders Are Perhaps surprisingly, in 10 of the 19 cartels, there was more than one ringleader (Table 4).21 In an extreme example, Carton board, there were seven – each a market leader in its own national market within the EU.22 More often than not (in 6 of the 10) these are global cartels23 and often different leaders represented different continents. In gas‐insulated switchgear, the Japanese members rotated leadership among themselves; among the European members, initially Siemens was the leader but Areva/Alstom took over leadership when Siemens temporarily left the cartel. Table 4 Size Ranks of Ringleader(s), and by Type Rank(s) . Total . Aggressive . Organisational . No. 1 sole 7 5 2 Nos. 1 and 2 6 6 0 No. 2 sole 1 1 Sole outsider 1 1 Others (multiple RL) 4 4 Total 19 11 8 Rank(s) . Total . Aggressive . Organisational . No. 1 sole 7 5 2 Nos. 1 and 2 6 6 0 No. 2 sole 1 1 Sole outsider 1 1 Others (multiple RL) 4 4 Total 19 11 8 Notes Two each with two ringleaders (Nos. 1 & 5 and 2 & 4), one with five (Nos. 1, 2, 3, 4 & 5), and one with seven (Nos. 1, 2, 3, 4, 5, 6 & 10). Open in new tab Table 4 Size Ranks of Ringleader(s), and by Type Rank(s) . Total . Aggressive . Organisational . No. 1 sole 7 5 2 Nos. 1 and 2 6 6 0 No. 2 sole 1 1 Sole outsider 1 1 Others (multiple RL) 4 4 Total 19 11 8 Rank(s) . Total . Aggressive . Organisational . No. 1 sole 7 5 2 Nos. 1 and 2 6 6 0 No. 2 sole 1 1 Sole outsider 1 1 Others (multiple RL) 4 4 Total 19 11 8 Notes Two each with two ringleaders (Nos. 1 & 5 and 2 & 4), one with five (Nos. 1, 2, 3, 4 & 5), and one with seven (Nos. 1, 2, 3, 4, 5, 6 & 10). Open in new tab In most (16 of the 19) of the cartels the largest firm in the industry was either the sole or one of the ringleaders.24 The three exceptions are the following: Spanish tobacco processors, in which the ringleader was, strictly speaking, an outsider to the industry, being a distributor, rather than producer, of tobacco; citric acid, in which the two ringleaders, ADM (No. 2) and Roche (No. 4), were the market leaders in two other closely adjacent cartels (Lysine and Vitamins respectively); and sodium gluconate in which the identified ringleader, Jungbunzlauer, was the second largest firm in the market. Table 4 provides a clearer focus on the significant roles of No. 1 firms: in all 11 ‘aggressive’ cases, the No. 1 ranked firm was a ringleader – either alone or together with the No. 2 firm. This implies that if a ringleader is to exert an aggressive role, it must also be able to wield some power within the cartel. In contrast, in the eight organisational only cases, it is much more common to find multiple‐leadership and not always involving a dominant leading firm. This broad dichotomy sheds light on Proposition 3 above and has direct implications for the specification of the econometric model below. 2.4. Ringleaders by Cartel Type (Nature of the Agreement) According to Proposition 2, the ‘need’ for a ringleader may differ between different types of cartel agreement, being least pronounced where the collusion merely requires forbearance by members, so that each respects its rivals’ initial endowments of home territories and/or captive customers, and where cheating is therefore easier to detect and there is no necessity to agree a common price. Table 5 reports the incidence of different types of agreements. Confirming Harrington’s (2006) earlier finding, nearly all cartels fix price and/or rig bidding but the various forms of market sharing are also common. In addition, as is clear from the first column, agreements typically involve a combination of price and market sharing. Table 5 Types of Agreement: Ringleader Versus Non‐ringleader Cartels Nature of agreement . All cartels . Cartels with ringleaders . Probability of ringleader . Price agreements Price‐Fixing 70 18 0.26 Bid Rigging 18 7 0.39 Market sharing agreements Territorial 16 5 0.38 Customer 29 5 0.17 Quota 26 10 0.38 Without price‐fixing/bid rigging 13 0 0 Total 89 19 0.21 Nature of agreement . All cartels . Cartels with ringleaders . Probability of ringleader . Price agreements Price‐Fixing 70 18 0.26 Bid Rigging 18 7 0.39 Market sharing agreements Territorial 16 5 0.38 Customer 29 5 0.17 Quota 26 10 0.38 Without price‐fixing/bid rigging 13 0 0 Total 89 19 0.21 Open in new tab Table 5 Types of Agreement: Ringleader Versus Non‐ringleader Cartels Nature of agreement . All cartels . Cartels with ringleaders . Probability of ringleader . Price agreements Price‐Fixing 70 18 0.26 Bid Rigging 18 7 0.39 Market sharing agreements Territorial 16 5 0.38 Customer 29 5 0.17 Quota 26 10 0.38 Without price‐fixing/bid rigging 13 0 0 Total 89 19 0.21 Nature of agreement . All cartels . Cartels with ringleaders . Probability of ringleader . Price agreements Price‐Fixing 70 18 0.26 Bid Rigging 18 7 0.39 Market sharing agreements Territorial 16 5 0.38 Customer 29 5 0.17 Quota 26 10 0.38 Without price‐fixing/bid rigging 13 0 0 Total 89 19 0.21 Open in new tab The second column shows the incidence of ringleaders by type of agreement and, at first sight, this reveals no obvious tendency for ringleaders to be more prevalent for certain types of agreement than others. However, this is misleading precisely because cartels typically employ combinations of agreement types, thus these are estimates of unconditional probabilities. Rather, the key appropriate statistic is reported in the last row of the Table, which isolates those cartels which confine their agreements exclusively to market sharing (usually territorial or customer allocations), without the need to agree on price. There are 13 such cartels, in none of which is there a ringleader. This allows us to draw a very strong conclusion, for this sample at least: conditional on the cartel having no agreement on price‐fixing or bid rigging, it will never involve a ringleader. In summary, these simple descriptive statistics provide strong supporting evidence for our Propositions 1, 2 and 4 for this sample of EC cases. Ringleaders only occur in a minority of cartels; there is often more than one ringleader; ringleaders are never found in cartels where collusion can be effected without involving price; and the ringleader is usually the largest cartel member, especially where its functions are ‘aggressive’. 3. Are Ringleaders More Likely in Larger Number Asymmetric Cartels? This section turns to the main Proposition 3 of the article, concerning size and asymmetry within the cartel, while allowing for the possibility that it may be sensitive to the type of ringleader and nature of the agreement. 3.1. Model and Specification (i) Initially, the model is tested using a simple logit form: pr{RLi}=f(Ni,Ai,Zi),(1) where pr{RLi} is the probability that cartel i has a ringleader or ringleaders, measured using a binary variable with 1 denoting existence of a ringleader. N is number of cartel members over its duration.25A is the asymmetry of market shares of members. In the results reported here this is measured by the range of market shares, i.e. the difference in market shares between the largest and smallest members (range).26 This is our preferred measure as it reflects the sizes of those firms (either the largest or the smallest) which previous theoretical literature most commonly highlight as crucial to the cartel stability in terms of incentive compatibility. Zi are other characteristics of the market and cartel. From the above discussion, we identify three such characteristics: Trade Assoc is a dummy variable indicating the presence of a trade association actively involved in the cartel. The expectation is that this reduces the need for a ringleader. Time is a time trend, included in quadratic form, bearing in mind Figure 1. This variable controls for the possibility that firms may be more or less willing to fulfil the role of ringleader over a time period when there were policy changes in the fining policy of the CA. Also, the inclusion of this trend may help to control for certain types of selection bias (see below). Nature of the cartel agreement. The previous Section has already confirmed that the nature of the agreement does matter: ringleaders never occur where collusion does not involve price agreements. However, this cannot be included as a regressor because it is, by definition, a perfect (negative) predictor. As an alternative treatment, the equation will be re‐estimated, omitting those cartels without price agreements. (ii) In addition, to test whether the marginal effects of firm numbers and asymmetries differ by ringleader type, the equation is then re‐estimated as a multinomial logit, where the cartel chooses between three alternatives, no ringleader, a merely organisational ringleader, or an aggressive ringleader: pr{ki}=pr{U(k)>U(l)}for all otherl≠k, where U(k)=f(Ni,Ai,Zi).(2) 3.2. Sample Size To measure Range (or any other measure of asymmetry), we require usable information on the market shares of individual members. While the EC does not routinely report this in its decision documents, it is possible to infer individual market shares for most cartels, if sometimes only approximately.27 As far as is known, this is the first study to quantify the sizes of individual firms in EC cartels. However, this proved impossible for 25 of the 89 cartels: in 14 cases this was because the cartel comprised an ‘association of members’, very often these associations include very large, but unreported, numbers; five cases involved shipping for which no individual firm data are reported; and in six other cartels, no reliably useful data were reported or could be inferred. Below we discuss whether these exclusions are likely to aggravate problems of sample selection. 3.3. Results Equation I in Table 6 shows the logit results for the full sample, making no distinction between cartel type or ringleader type. Both N and Range are positively significant at least at the 5% level (hereafter, ‘significance’ always refers to the 5% level): ringleaders are more likely, the more members in the cartel and the greater are asymmetries in their sizes. The significant coefficients on the quadratic time trend imply an increasing tendency to ring‐leadership up until 1999/2000, but thereafter the probability declines. This is consistent with an increased willingness by the CA through the 1990s to identify and name ringleaders. The tailing off and then reversal of the trend thereafter could be consistent with a decline in this willingness, increased skills of ringleaders in concealing their activities and/or a deterrent effect on the number of cartels actually requiring a ringleader. Each of these explanations would be consistent with the increased tendency for ringleaders to be penalised more heavily as a consequence of the EC’s leniency programme. The negative coefficient on Trade Assoc is consistent with expectations – ringleaders are less necessary given the presence of a trade association – but this is only very weakly significant (at the 15% level). The equation’s success in correctly predicting outcomes28 is relatively high (81%) for a study such as this. Nevertheless, there are seven cartels where the equation fails to identify the presence of a ringleader and five where it incorrectly predicts a ringleader (referred to hereafter as Type 1 and Type 2 errors respectively). Table 6 Results . Equation I (Full sample) . Equation II (Only price‐fixing &/or bid rigging) . Equation III (Only price‐fixing &/or bid rigging) . . logit . logit . Multinomial logit . . . . Organisational ringleaders . Aggressive ringleaders . . Coef. . P > Z . Coef. . P > Z . Coef. . P > Z . Coef. . P > Z . Time 0.885 0.044 1.713 0.044 1.304 0.107 2.908 0.008 Time squared −0.042 0.025 −0.079 0.024 −0.058 0.081 −0.143 0.002 N 0.448 0.000 0.690 0.001 0.615 0.002 0.804 0.002 Range 8.642 0.044 19.398 0.000 14.713 0.005 29.884 0.000 Trade assoc −1.114 0.152 −2.388 0.054 −1.846 0.132 −3.658 0.011 Constant −9.802 0.007 −18.08 0.008 −15.42 0.017 −27.32 0.002 Observations 64 59 59 Wald chi2(5) 17.11 18.54 36.23 Pseudo R2 0.349 0.547 0.515 Log pseudo likelihood −25.33 −16.79 −24.28 Correct predictions (%) 81.2 86.44 84.75 . Equation I (Full sample) . Equation II (Only price‐fixing &/or bid rigging) . Equation III (Only price‐fixing &/or bid rigging) . . logit . logit . Multinomial logit . . . . Organisational ringleaders . Aggressive ringleaders . . Coef. . P > Z . Coef. . P > Z . Coef. . P > Z . Coef. . P > Z . Time 0.885 0.044 1.713 0.044 1.304 0.107 2.908 0.008 Time squared −0.042 0.025 −0.079 0.024 −0.058 0.081 −0.143 0.002 N 0.448 0.000 0.690 0.001 0.615 0.002 0.804 0.002 Range 8.642 0.044 19.398 0.000 14.713 0.005 29.884 0.000 Trade assoc −1.114 0.152 −2.388 0.054 −1.846 0.132 −3.658 0.011 Constant −9.802 0.007 −18.08 0.008 −15.42 0.017 −27.32 0.002 Observations 64 59 59 Wald chi2(5) 17.11 18.54 36.23 Pseudo R2 0.349 0.547 0.515 Log pseudo likelihood −25.33 −16.79 −24.28 Correct predictions (%) 81.2 86.44 84.75 Open in new tab Table 6 Results . Equation I (Full sample) . Equation II (Only price‐fixing &/or bid rigging) . Equation III (Only price‐fixing &/or bid rigging) . . logit . logit . Multinomial logit . . . . Organisational ringleaders . Aggressive ringleaders . . Coef. . P > Z . Coef. . P > Z . Coef. . P > Z . Coef. . P > Z . Time 0.885 0.044 1.713 0.044 1.304 0.107 2.908 0.008 Time squared −0.042 0.025 −0.079 0.024 −0.058 0.081 −0.143 0.002 N 0.448 0.000 0.690 0.001 0.615 0.002 0.804 0.002 Range 8.642 0.044 19.398 0.000 14.713 0.005 29.884 0.000 Trade assoc −1.114 0.152 −2.388 0.054 −1.846 0.132 −3.658 0.011 Constant −9.802 0.007 −18.08 0.008 −15.42 0.017 −27.32 0.002 Observations 64 59 59 Wald chi2(5) 17.11 18.54 36.23 Pseudo R2 0.349 0.547 0.515 Log pseudo likelihood −25.33 −16.79 −24.28 Correct predictions (%) 81.2 86.44 84.75 . Equation I (Full sample) . Equation II (Only price‐fixing &/or bid rigging) . Equation III (Only price‐fixing &/or bid rigging) . . logit . logit . Multinomial logit . . . . Organisational ringleaders . Aggressive ringleaders . . Coef. . P > Z . Coef. . P > Z . Coef. . P > Z . Coef. . P > Z . Time 0.885 0.044 1.713 0.044 1.304 0.107 2.908 0.008 Time squared −0.042 0.025 −0.079 0.024 −0.058 0.081 −0.143 0.002 N 0.448 0.000 0.690 0.001 0.615 0.002 0.804 0.002 Range 8.642 0.044 19.398 0.000 14.713 0.005 29.884 0.000 Trade assoc −1.114 0.152 −2.388 0.054 −1.846 0.132 −3.658 0.011 Constant −9.802 0.007 −18.08 0.008 −15.42 0.017 −27.32 0.002 Observations 64 59 59 Wald chi2(5) 17.11 18.54 36.23 Pseudo R2 0.349 0.547 0.515 Log pseudo likelihood −25.33 −16.79 −24.28 Correct predictions (%) 81.2 86.44 84.75 Open in new tab Equation II re‐estimates the model but now only for those cartels found to be fixing price or rigging bids. Signs of all coefficients are unchanged, and significance levels are now increased, especially for Range, and Trade Assoc is now significant at the 6% level. The time trend continues to predict a turning point at 2000. The explanatory power of the equation is enhanced: 86% of cases are now correctly predicted, and there are only four Type 1 and four Type 2 errors. Thus, exclusion of non‐price‐fixing/bid‐rigging cartels provides more precise estimates. Equation III estimates the multinomial model, distinguishing between organisational only and aggressive ringleaders, as opposed to no ringleaders. All estimated coefficients have the same signs as before and all coefficients remain significant at the 5% level, except the time trend which slips to 10% significance and 13% for trade association for organisational ringleaders. A Wald test on the difference between the coefficients on Range reveals that this has a significantly greater impact for aggressive ringleaders: greater asymmetries are particularly associated with aggressive ringleaders. None of the other variables has a significantly different impact between the two types of ringleader. The equation generates predictions which are 85% correct29 and there are five Type 1 and two Type 2 errors (and a further two for which the ‘wrong’ type of ringleader is predicted). A Hausman test for independence of irrelevant alternatives confirms that the assumption can be accepted. In summary, the three models robustly confirm Proposition 3 for this sample: ringleaders are more likely in asymmetric, large number cartels and this is most pronounced for aggressive ringleaders. There is weaker evidence that trade associations may sometimes be a viable alternative to ringleaders. Over time, ceteris paribus, the EC was increasingly successful in identifying ringleaders through the 1990s, although this trend was reversed through the 2000s. This time pattern is consistent with an impact for the changed treatment of ringleaders as part of the EC’s leniency programme.30 3.4. Possible Sample Selection Bias In any empirical study designed to explain the decisions of a CA there are almost inevitably potential selection problems. In this case, there are four potential sources. First, there is the generic issue familiar in the previous cartel literature, that the researcher only observes detected cartels and these may not be a random sample of the population as a whole. This possibility cannot be rejected and it qualifies all results as conditional on cartel detection. Second, there are sample attrition problems, similarly not uncommon in studies such as this. As noted above, 25 observations were lost because of lack of data on firm size. However, most of these cartels (notably the associations of members and shipping lines) will have had relatively large numbers of members and, as none had an identified ringleader, their exclusion is likely to have biased downwards the impact of N in the above regressions. Third, the exclusion of the non‐price‐fixing, non‐bid‐rigging cartels from Equations II and III might introduce a bias if they differ significantly from those remaining in the sample, particularly in their structures. t‐tests on the sample means for N and Range reveal weakly (at the 10% level) significant tendencies for cartels without price‐fixing or bid rigging to have fewer members but greater asymmetries. As these two effects pull in opposite directions in terms of their predicted impact on the probability of ringleaders, it is not clear whether the exclusion of these cases will have introduced any significant bias to the estimates in Equations II and III. Finally, the ‘existence’ of a ringleader depends on (to some extent) subjective judgements by the CA on what constitutes penalisable ringleader activities. This might be problematic in the following sense. Suppose the ‘true’ model entails two stages. Stage 1 refers to the probability that a ringleader actually exists (as defined by Equations I or II) and stage 2 refers to the conditional probability that the CA publicly identifies a ringleader, given that one exists. The researcher only observes the net outcomes of the two stages conflated and separate identification of parameters from the first stage is therefore impossible. Given the time profile of ringleader cases observed earlier in Table 1, this raises particular doubts for 2000–2, in which there is an evident spike in the proportionate incidence of ringleaders – over 40% of all identified ringleaders occurred in this 3‐year period. This raises the obvious question of whether there was some unusual event(s) at this time which may have biased our results in some way. Certainly, this coincides with the run‐up to the revision of the leniency notice in 2002. However, only six of the 21 cases in these years involved leniency, and only two of these six were ringleader cases: neither of these rates differs significantly from those for the years other than 2000–2. The inclusion of the quadratic time trend in the equation may have statistically ‘controlled’ for any unobserved time‐variant effects. For example, the observed inverted U is consistent with a CA that became increasingly successful in detecting and prosecuting cartels, and increasingly confident in naming ringleaders in the years up to 2002. Thereafter, the stiffening of fines on ringleaders may have made the CA more cautious in naming a ringleader, for fear that more severe punishment would increase the likelihood of appeal. Similarly, ringleaders themselves may have become more adept in concealing their leading roles in periods of more severe punishment. However, while plausible, these hypotheses are inherently non‐substantiable. Nevertheless, whatever the explanation for this trend, we can confirm that its impact is independent of the cartel’s structure (firm numbers and asymmetries). All equations in Table 6 have been re‐estimated omitting the quadratic time effect – if either N or Range was systematically time variant, this should be revealed by instability in their estimated coefficients, as compared to those shown in Table 6. In fact, while the overall fits (pseudo R2) decline, the coefficients on both variables, and their significance levels, remain virtually unchanged from those shown in Table 6.31 In conclusion, it is essential to interpret these results cautiously. They have established that the CA is more likely to identify a ringleader in those cartels with a larger number of members who exhibit greater size asymmetries (measured by range). We cannot exclude the possibility that the CA’s propensity to name a ringleader, given that one exists, may have varied over the time period. However, there is no reason for believing that this has had any bearing on the observed effects of cartel size and asymmetry. 4. Implications and Conclusions The main purpose of this article is to evaluate how far the ringleader can be interpreted as an organisational mechanism enabling cartelists to overcome the ‘cartel problems’ in those instances where they are likely to be most pronounced, i.e. larger asymmetric sets of members. It has generated a series of results, and it is acknowledged that these should be seen in the context of a relatively small sample size, which may be subject to sample selection bias, a simple model, and dependent variables which reflect the subjective decisions of a CA. Subject to these qualifications, there are six main results. All should be interpreted to refer to the stated decisions of the CA on the set of cartels it actually detects. First, ringleaders are identified in about one in five cartels. Second, especially where the ringleader engages in ‘aggressive’ activities, it tends to be the largest member or members of the cartel. Where the activities of the ringleader are confined to a more facilitating organisational nature and the cartel has a global coverage, it is not uncommon to observe a number of ringleaders, some or all of whom are not dominant. Third, ringleaders only occur where the conspiracy involves price‐fixing or bid rigging. Other forms of agreement, notably pure territorial or customer allocation, do not appear to require a ringleader – such agreements are more amenable to self‐monitoring and enforcement and secret deviation is difficult. Fourth, ringleaders are less common where an existing trade association may be able to fulfil similar roles. Fifth, a significant inverse quadratic time trend suggests that the European Commission had an increasing propensity to name ringleaders in the 1990s but this has been reversed since 2000. This is consistent with, but does not prove, a deterrent role from the increasingly asymmetric fines imposed on ringleaders. It is not clear whether this is ‘cosmetic’ – in the sense that ringleaders have become more sophisticated in revealing their leading roles and/or that the CA has become more cautious in naming a ringleader and thereby invoking appeals – or whether policy has had the effect of deterring the subset of cartels in which a ringleader is necessary. Finally, and most important given the article’s main objective, the evidence strongly confirms that ringleaders are statistically more likely to be identified in cartels with a relatively large number of members who are asymmetrically sized. This finding has an important implication for policy makers, as it clearly underlines that collusion may not be confined to small number symmetrical size distributions of firms. Further research on all these findings is required, including replication if possible for other CAs, and also further in‐depth case analysis of individual cartels. Moreover, this article is largely empirical in its intentions. Further theoretical research is required, perhaps employing theories of coalition and network formation. The stylised facts uncovered in this article should provide a useful input into any such future empirical work. Footnotes 1 " The article draws on the database constructed by De (2010) for her PhD thesis. It develops the analysis in her Chapters 5 and 6. 2 " For other literature surveys on the market characteristics that should facilitate collusion, see: Posner (1976, 2001), Scherer and Ross (1990), Motta (2004), Cabral (2005), Feuerstein (2005) and Grout and Sonderegger (2005). 3 " These are surveyed by De (2010), and include Posner (1970), Hay and Kelley (1974), Frass and Greer (1977), Jacquemin et al. (1981), Marquez (1994), Dick (1996), Symeonidis (2003), Grout and Sonderegger (2005), Suslow (2005), Zimmerman and Connor (2005), Combe and Monnier (2010), Levenstein and Suslow (2011). 4 " See, for example, Harrington (2006), De (2010), Levenstein and Suslow (2011) for detailed analysis of the organisational characteristics and their impact on cartel success. 5 " The EC moved from a ‘discriminatory approach’ towards ringleaders to a ‘non‐discriminatory approach’ with the 2002 and 2006 leniency notices. Ringleaders can apply for fine reduction and even for immunity from fines provided ‘the undertaking did not take steps to coerce other undertakings to participate in the infringement’ (European Commission, 2002) ‘Notice on immunity from fines and reduction of fines in cartel cases’, Para A11 (c)). In the US, amnesty can be granted under the corporate leniency programme if there are multiple ringleaders or instigators of the cartel and the first reporting firm is one of them, whereas in the EU, a single ringleader (or instigator) and multiple ringleaders (or instigators) are treated identically. 6 " Although they do not elaborate at length on what these fixed costs might be, they cite the cost of buying out potential entrants and higher ringleader penalties under a leniency regime. 7 " It should be stressed that the Bos and Wandschneider (2011) study was conducted quite independently of De’s (2010) and, therefore, of this article too. 8 " However, Brock and Scheinkman (1985) show that if the number of firms increases, holding capacity constant, with a small number of firms, the ability to collude increases with an additional firm but eventually falls. This result is strikingly different from the theoretical prediction without capacity constraints. 9 " Patinkin (1947), Bain (1948), Bae (1987), Harrington (1991), Verboven (1997) and Rothschild (1999) and Miklós‐Thal (2009). 10 " See Lambson (1994), Compte et al. (2002) and Bos and Harrington (2010) for capacity asymmetry; Harrington (1989) for the discount factor and Kuhn and Rimler (2006) for product differentiation. 11 " European Commission (2006), Case T‐15/02 BASF v Commission Summary of Judgement, March 15th 2006. 12 " Para 100, Official Journal, L series 152/24, 7/06/2001, Case COMP/36.545/F3‐Amino Acids, Decision of 7 June 2000. 13 " Para 148, Official Journal, L series 115/1, 21/04/2004, Case COMP/E‐1/36.212 – Carbonless paper, Decision of 20 December, 2001. 14 " Posner (1970), Hay and Kelley (1974), Frass and Greer (1977), Suslow (2005), Zimmerman and Connor (2005), Bolotova et al. (2006), Harrington (2006), Combe and Monnier (2010) and Levenstein and Suslow (2011). 15 " Detailed description of the data can be found in De (2010). The sample is drawn from the same source as was used in the two above cited studies, but is larger than theirs: Ganslandt et al. (2012) have 43 and Bos and Wandschneider (2011) have 75. 16 " http://ec.europa.eu/competition/cartels/cases/cases.html. This has been supplemented, where necessary, by the opinion and judgements of the Court of First Instance and European Court of Justice. The judgements 1997 onwards can be found on the CURIA website (http://curia.europa.eu/jcms/jcms/j_6/) and, before 1997, from the DG Competition website (http://ec.europa.eu/competition/court/index.html). Some additional information was also collected from various relevant merger cases reported on the DG Competition website, http://ec.europa.eu/competition/mergers/cases/. 17 " This is the Carton board case in which the European Court of Justice’s judgement explains that it applied a higher percentage (9%) of turnover to calculate the ringleaders’ fines compared to the other members (7.5%). ( Judgment of the Court C‐248/98 P, 16th November 2000, para 71). 18 " Bos and Wandschneider (2011) report a similar proportion (19%) of ringleader cases in their sample. 19 " See also Bos and Wandschneider (2011, pp. 7–8) for broadly similar findings. 20 " The exceptions are sorbates, Spanish bitumen, gas‐insulated switchgear, paraffin wax and sodium gluconate cartels; in these case presumably, all members instigated jointly. 21 " Bos and Wandschneider (2011) report that this occurs in 10 of their 14 ringleader cases. 22 " We treat this case as a single cartel in the following analysis. 23 " The proportion of global cartels in the whole sample was much lower at 34%. 24 " Bos and Wandschneider (2011) report that the largest firm was one of the ringleaders in 11 of their 14 cases. 25 " This is measured by the average number of members over the cartel’s duration. Typically this is largely time invariant, although there are some cartels with entry in the early years and exit towards the end of the period. 26 " In fact, the results reported below are robust to alternative measures of size asymmetry; including Gini based measures and the coefficient of variation. For further discussion see De (2010, ch. 6). 27 " See De (2010, pp. 109–11). Members’ shares are often reported as ranges, e.g. 10%–20%; in such cases we typically employ the midpoints, subject to moderation where other information is available on an ad hoc basis. 28 " A prediction is deemed successful if the predicted probability of the actual outcome exceeds 50%. 29 " In this multinomial case a prediction is deemed successful if the predicted probability of the actual outcome exceeds the predicted probabilities of each of the two alternatives. 30 " In other experiments we have included a binary variable, Global, representing whether the cartel was global as opposed to being confined to intra‐EU. There are 31 global cartels in the sample. This attracts positive coefficients in all three equations but is only significant at the 10% level in Equations II and III. Thus, there is weak evidence that ringleaders are more necessary in global cartels. With the inclusion of this additional explanatory variable, the coefficients of all other variables are remain virtually unchanged from those shown in the Table and all significance levels are unchanged. 31 " As a further test of robustness, all equations were re‐estimated excluding cases occurring in 2000–2. In spite of the large reduction in the number of degrees of freedom, all coefficients are stable and their significance levels are confirmed in all equations. Indeed, for Range and N, levels of significance are slightly increased in all equations. References Athey , S. and Bagwell , K. ( 2001 ). ‘ Optimal collusion with private information ’, RAND Journal of Economics , vol. 32 ( 3 ), pp. 428 – 65 . Google Scholar Crossref Search ADS WorldCat Athey , S. and Bagwell , K. ( 2008 ). ‘ Collusion with persistent cost shocks ’, Econometrica , vol. 76 ( 3 ), pp. 493 – 540 . Google Scholar Crossref Search ADS WorldCat Bae , H. ( 1987 ). ‘ A price‐setting supergame between two heterogeneous firms ’, European Economic Review , vol. 31 ( 6 ), pp. 1159 – 71 . Google Scholar Crossref Search ADS WorldCat Bain , J.S. ( 1948 ). ‘ Output quotas in imperfect cartels ’, Quarterly Journal of Economics , vol. 62 ( 4 ), pp. 617 – 22 . Google Scholar Crossref Search ADS WorldCat Bolotova , Y. , Connor , J.M. and Miller , D.J. ( 2006 ). ‘ Cartel stability: an empirical analysis ’, available at http://papers.ssrn.com/sol3/papers.cfm?abstract_id=939078 (last accessed: 5 January 2011). Bos , I. and Harrington , J.E. , Jr. ( 2010 ). ‘ Endogenous cartel formation with heterogeneous firms ’, RAND Journal of Economics , vol. 41 (1), pp. 92 – 117 . Google Scholar Crossref Search ADS WorldCat Bos , I. and Wandschneider , F. ( 2011 ). ‘ Cartel ringleaders and the corporate leniency program ’, Working Paper, Centre for Competition Policy . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Brock , W.A. and Scheinkman , J.A. ( 1985 ). ‘ Price setting supergames with capacity constraints ’, Review of Economic Studies , vol. 52 ( 3 ), pp. 371 – 82 . Google Scholar Crossref Search ADS WorldCat Cabral , L.M.B. ( 2005 ). ‘ Collusion theory: where to go next? ’, Journal of Industry, Competition and Trade , vol. 5 ( 3 ), pp. 199 – 206 . Google Scholar Crossref Search ADS WorldCat Combe , E. and Monnier , C. ( 2010 ). ‘ Cartels in Europe: main features ’, Concurrences , vol. 3 ( 30906 ), pp. 23 – 31 . OpenURL Placeholder Text WorldCat Compte , O. , Jenny , F. and Rey , P. ( 2002 ). ‘ Capacity constraints, mergers and collusion ’, European Economic Review , vol. 46 ( 1 ), pp. 1 – 29 . Google Scholar Crossref Search ADS WorldCat Cooper , D. ( 1997 ). ‘ Barometric price leadership ’, International Journal of Industrial Organization , vol. 15 ( 3 ), pp. 301 – 25 . Google Scholar Crossref Search ADS WorldCat De , O. ( 2010 ). ‘ The internal structures and organisation of EC prosecuted cartels and the impact on their performance ’, Doctoral thesis, University of East Anglia . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Dick , A.R. ( 1996 ). ‘ Identifying contracts, combinations and conspiracies in restraint of trade ’, Managerial and Decision Economics , vol. 17 ( 2 ), pp. 203 – 16 . Google Scholar Crossref Search ADS WorldCat Elder , S. ( 2010 ). ‘ Punishing organized crime leaders for the crimes of their subordinates ’, Criminal Law and Philosophy , vol. 4 ( 2 ), pp. 183 – 96 . Google Scholar Crossref Search ADS WorldCat European Commission ( 1996 ). ‘ Commission notice on the non‐imposition or reduction of fines in cartel cases ’, Official Journal of the European Commission , vol. C ( 207 ), pp. 4 – 6 . OpenURL Placeholder Text WorldCat European Commission ( 2002 ). ‘ Commission notice on immunity from fines and reduction of fines in cartel cases ’, Official Journal of the European Commission , vol. C ( 45 ), pp. 3 – 5 OpenURL Placeholder Text WorldCat European Commission ( 2006 ). ‘ Commission notice on immunity from fines and reduction of fines in cartel cases ’, Official Journal of the European Commission , vol. C ( 298 ), pp. 17 – 22 . OpenURL Placeholder Text WorldCat Feuerstein , S. ( 2005 ). ‘ Collusion in industrial economics – a survey ’, Journal of Industry, Competition and Trade , vol. 5 ( 3 ), pp. 163 – 98 . Google Scholar Crossref Search ADS WorldCat Frass , A.G. and Greer , D.F. ( 1977 ). ‘ Market structure and price collusion: an empirical analysis ’, Journal of Industrial Economics , vol. 26 ( 1 ), pp. 21 – 44 . Google Scholar Crossref Search ADS WorldCat Ganslandt , M. , Persson , L. and Vasconcelos , H. ( 2012 ). ‘ Endogenous mergers and collusion in asymmetric market structures ’, Economica , vol. 79 ( 316 ), pp. 766 – 91 . OpenURL Placeholder Text WorldCat Grout , P. and Sonderegger , S. ( 2005 ). ‘ Predicting cartels ’, Economic Discussion Paper, Office of Fair Trading . Harrington , J.E. , Jr. ( 1989 ). ‘ Collusion among asymmetric firms: the case of different discount factors ’, International Journal of Industrial Organization , vol. 7 ( 2 ), pp. 289 – 307 . Google Scholar Crossref Search ADS WorldCat Harrington , J.E. , Jr. ( 1991 ). ‘ The determination of price and output quotas in a heterogeneous cartel ’, International Economic Review , vol. 32 ( 4 ), pp. 767 – 92 . Google Scholar Crossref Search ADS WorldCat Harrington , J.E. , Jr. ( 2006 ). ‘ How do cartels operate? ’, Foundations and Trends in Microeconomics , vol. 2 ( 1 ), pp. 1 – 105 . Google Scholar Crossref Search ADS WorldCat Harrington , J.E. , Jr. and Skrzypacz , A. ( 2007 ). ‘ Collusion under monitoring of sales ’, RAND Journal of Economics , vol. 38 ( 2 ), pp. 314 – 31 . Google Scholar Crossref Search ADS WorldCat Hay , G.A. and Kelley , D. ( 1974 ). ‘ An empirical survey of price fixing conspiracies ’, Journal of Law and Economics , vol. 17 ( 1 ), pp. 13 – 38 . Google Scholar Crossref Search ADS WorldCat Herre , J. and Rasch , A. ( 2009 ). ‘ The deterrence effect of excluding ringleaders from leniency programs ’, Discussion Paper, University of Cologne . Ishibashi , L. ( 2008 ). ‘ Collusive price leadership with capacity constraints ’, International Journal of Industrial Organization , vol. 26 ( 3 ), pp. 704 – 15 . Google Scholar Crossref Search ADS WorldCat Ivaldi , M. , Jullien , B., Rey , P., Seabright , S. and Tirole , J. ( 2003 ). ‘ The economics of tacit collusion ’, Final Report for DG Competition, European Commission. Jacquemin , A. , Nambu , T. and Isabelle , D. ( 1981 ). ‘ A dynamic analysis of export cartels: the Japanese case ’, Economic Journal , vol. 91 ( 363 ), pp. 685 – 96 . Google Scholar Crossref Search ADS WorldCat Kuhn , K.‐U. and Rimler , M.S. ( 2006 ). ‘ The comparative statics of collusion models ’, available at: http://ssrn.com/abstract=906942 or http://dx.doi.org/10.2139/ssrn.906942 (last accessed: 15 February 2013). Lambson , V.E. ( 1994 ). ‘ Some results on optimal penal codes in asymmetric Bertrand supergames ’, Journal of Economic Theory , vol. 62 ( 2 ), pp. 444 – 68 . Google Scholar Crossref Search ADS WorldCat Leslie , C.R. ( 2006 ). ‘ Antitrust amnesty, game theory, and cartel stability ’, Journal of Corporation Law , vol. 31 , pp. 453 – 88 . OpenURL Placeholder Text WorldCat Levenstein , M.C. and Suslow , V.Y. ( 2006 ). ‘ What determines cartel success? ’, Journal of Economic Literature , vol. 44 ( 1 ), pp. 43 – 95 . Google Scholar Crossref Search ADS WorldCat Levenstein , M.C. and Suslow , V.Y. ( 2011 ). ‘ Breaking up is hard to do: determinants of cartel duration ’, Journal of Law and Economics , vol. 54 ( 2 ), pp. 455 – 92 . Google Scholar Crossref Search ADS WorldCat Markham , J.W. ( 1951 ). ‘ The nature and significance of price leadership ’, American Economic Review , vol. 41 ( December ), pp. 891 – 905 . OpenURL Placeholder Text WorldCat Marquez , J. ( 1994 ). ‘ Life expectancy of international cartels: an empirical analysis ’, Review of Industrial Organization , vol. 9 ( 3 ), pp. 331 – 41 . Google Scholar Crossref Search ADS WorldCat Marshall , R.C. , Marx , L.M. and Samkharadze , L. ( 2011 ). ‘ Dominant‐firm conduct by cartels ’, Discussion Paper, Penn State University . Miklós‐Thal , J. ( 2009 ). ‘ Optimal collusion under cost asymmetry ’, Economic Theory , vol. 46 ( 1 ), pp. 99 – 125 . Google Scholar Crossref Search ADS WorldCat Motta , M. ( 2004 ). Competition Policy: Theory and Practice , Cambridge, MA : Cambridge University Press . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Mouraviev , I. and Rey , P. ( 2011 ). ‘ Collusion and leadership ’, International Journal of Industrial Organization , vol. 29 ( 4 ), pp. 721 – 41 . OpenURL Placeholder Text WorldCat Osborne , D.K. ( 1976 ). ‘ Cartel problems ’, American Economic Review , vol. 66 ( 5 ), pp. 835 – 44 . OpenURL Placeholder Text WorldCat Patinkin , D. ( 1947 ). ‘ Multiple‐plant firms, cartels and imperfect competition ’, Quarterly Journal of Economics , vol. 61 ( 2 ), pp. 173 – 205 . Google Scholar Crossref Search ADS WorldCat Posner , R.A. ( 1970 ). ‘ A statistical study of antitrust enforcement ’, Journal of Law and Economics , vol. 13 ( 2 ), pp. 365 – 419 . Google Scholar Crossref Search ADS WorldCat Posner , R.A. ( 1976 ). Antitrust Law: An Economic Perspective , Chicago, IL : University of Chicago Press . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Posner , R.A. ( 2001 ). Antitrust Law , 2nd edn, Chicago, IL : University of Chicago Press . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Robinson , P.H. ( 1987 ). ‘ A sentencing system for the 21st century? ’, Texas Law Review , vol. 66 ( 1 ), pp. 1 – 61 . OpenURL Placeholder Text WorldCat Rotemberg , J.J. and Saloner , G. ( 1990 ). ‘ Collusive price leadership ’, Journal of Industrial Economics , vol. 39 ( 1 ), pp. 93 – 111 . Google Scholar Crossref Search ADS WorldCat Rothschild , R. ( 1999 ). ‘ Cartel stability when costs are heterogeneous ’, International Journal of Industrial Organization , vol. 17 ( 5 ), pp. 717 – 34 . Google Scholar Crossref Search ADS WorldCat Scherer , F. M. and Ross , D. ( 1990 ). Industrial Market Structure and Economic Performance , 3rd edn, Boston, MA : Houghton Mifflin Company . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Selten , R. ( 1973 ). ‘ A simple model of imperfect competition, where 4 are few and 6 are many ’, International Journal of Game Theory , vol. 2 ( 1 ), pp. 141 – 201 . Google Scholar Crossref Search ADS WorldCat Silving , H. ( 1967 ). Constituent Elements of Crime , Springfield, IL : Charles C. Thomas . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Suslow , V.Y. ( 2005 ). ‘ Cartel contract duration: empirical evidence from inter‐war international cartels ’, Industrial and Corporate Change , vol. 14 ( 5 ), pp. 705 – 44 . Google Scholar Crossref Search ADS WorldCat Symeonidis , G. ( 2003 ). ‘ In which industries is collusion more likely? Evidence from the UK ’, Journal of Industrial Economics , vol. 51 ( 1 ), pp. 45 – 74 . Google Scholar Crossref Search ADS WorldCat Vasconcelos , H. ( 2005 ). ‘ Tacit collusion, cost asymmetries, and mergers ’, RAND Journal of Economics , vol. 36 ( 1 ), pp. 39 – 62 . OpenURL Placeholder Text WorldCat Verboven , F. ( 1997 ). ‘ Collusive behaviour with heterogeneous firms ’, Journal of Economic Behavior & Organization , vol. 33 ( 1 ), pp. 121 – 36 . Google Scholar Crossref Search ADS WorldCat Zimmerman , J. E. and Connor , J.M. ( 2005 ). ‘ Determinants of cartel duration: a cross‐sectional study of modern private international cartels ’, Working Paper, Department of Agricultural Economics, Purdue University . Author notes " The financial support of the Economic and Social Research Council (ESRC) is gratefully acknowledged. We thank the editor and referees, Morten Hviid, Franco Mariuzzo, Fred Wandschneider and other members of seminars at the Centre for Competition Policy for helpful comments. © 2013 The Author(s). The Economic Journal © 2013 Royal Economic Society
TI - Ringleaders in Larger Number Asymmetric Cartels
JO - The Economic Journal
DO - 10.1111/ecoj.12062
DA - 2013-11-01
UR - https://www.deepdyve.com/lp/oxford-university-press/ringleaders-in-larger-number-asymmetric-cartels-6DdePppV7X
SP - F524
VL - 123
IS - 572
DP - DeepDyve
ER -