Leniency policies have become an important antitrust tool but it is not clear whether they have effectively prevented recidivism or whether firms have learned to collude under, and even make strategic use of them. If “recidivism” is really an industry-level phenomenon, the appropriate policy measures are very different from what is necessary if individual firms, having been detected and punished for colluding, engage in the behavior again. Following Levenstein et al. (2015), this brief discusses the recidivism question as one about post-cartel behavior, i.e. the set of policies required to assure that effective competition emerges post-cartel breakup.
Cartels are one of the main concerns of the European Commission (EC) and the US Department of Justice (DOJ) and so, the US and EU Leniency Programmes (LPs) were designed, in 1978 and 1996 respectively, as a device for the deterrence and dissolution of collusive agreements (see Marvão and Spagnolo (2015a) for an in-depth review on the available evidence of the effects of LPs).
In the analysis of cartel formation, recidivism is an important issue. In the set of 510 cartel members fined by the EC in 1998-2014, Marvão (2015) identifies 89 “multiple offenders” (firms fined for collusion more than once), 10 “repeat offenders” (firms which initiate a cartel after being investigated for another cartel), and 5 recidivists following the definition from Werden et al. (2011): firms which initiate a cartel after being fined for another cartel.
The DOJ dataset compiled by Levenstein and Suslow (2015), spanning 1961-2013, preliminarily finds 113 “multiple offenders” but only 14 “repeat offenders”. Of these 14 firms, 5 that had been previously indicted were caught in the 1990s, but none was indicted again by the DOJ in the 2000s.
Although the number of (discovered) “true recidivists” is not zero, it is less than 1% in these two samples (EU, US). Recidivism seems to arise when there are lapses in enforcement; not surprisingly, some firms take advantage of these lapses to return to old behaviors. Designing policies that are able to prevent recidivism requires understanding whether this is an industry or firm-level phenomenon.
Levenstein et al. (2015) use the above-mentioned EU and US datasets to show that collusion occurs in virtually all sectors of the economy, but with discernable patterns.
In the US, construction and chemicals are frequently cartelized (pre and post leniency). There are a large number of cartels in local markets in some industries, such as retail gasoline stations and dealers and ready-mix concrete. While collusion in these local markets is frequently uncovered, it is not necessarily amongst the same firms.
In the EU, chemicals and transport cartels are also frequent areas of collusive activity (although cartels that are strictly within national boundaries and prosecuted by national competition authorities are not included in the sample).
The authors show that there is a large share of repeat and multiple offenders in chemicals and a surprisingly high proportion of repeat offenders in the manufacture of transport and electrical equipment. The highest proportion of multiple offenders is found in pharmaceuticals and refined petroleum products. The transportation and storage market is a sector with a high incidence of collusion (83 convicted cartel members), but no repeat offenders.
While the determinants of cartel activity are varied and endogenous, some correlations with industry-driven recidivism can be discussed:
- Industry concentration. It increases the ease of tacit collusion and it should increase the likelihood of explicit collusion, but there are many cartel examples in unconcentrated industries. In some industries, it has been argued that high fixed costs make competition unstable, so that, absent collusion, firms price below long-run marginal cost and are unable to cover fixed costs (Pirrong, 1992).
- Culture and history. Spar (1994) argues that the cooperative culture necessary for survival for diamond miners facilitated collusion as the industry matured. Policy fluctuations can also contribute to this problem, as was the case in the US during the Great Depression.
- Inelastic demand. This is empirically challenging to capture if the observed prices have been affected by monopoly power, thus potentially raised to a level at which demand is elastic. In many cases, the direct consumer is a producer, so the downstream cost function and competitive intensity also influence elasticity of demand for the cartelized product. Grout and Sonderegger (2005) estimate the likelihood of collusion in the US and EU and rank industries accordingly. This could be used to target competition authority resources to select industries.
Once a cartel breaks-up, cartel members may decide to compete in the market, merge, tacitly collude, or explicitly collude again. The latter does not mean that the cartel re-forms: a firm may collude in a new industry or product line or with a new set of co-conspirators.
U.S. Steel was involved in 6 different US cartels between 1948 and 1969, with different cartel partners and in different steel products. VSL construction was similarly involved (including as a leader) in multiple US cartels across several decades with distinct, but overlapping partners.
In the EU, Akzo Nobel N.V. has been convicted for 9 cartels, which lasted between 1987 and 2007, and in which its co-conspirators were mostly overlapping – e.g. collusion with Arkema in 6 instances (although the latter changed its name during the period). Many of the other co-conspirators were also multiple offenders. While Akzo only received one fine increase for recidivism, it received 7 leniency reductions, of which 3 were full immunity.
Other EC repeat offenders are ABB and Degussa Evonik – both convicted 4 times and received full immunity twice – as well as Brugg and Sumitomo. The latter was convicted for 7 cartels, of which 5, in the automotive wire harness, were self-reported.
What may influence repeated cartel participation, at the firm level?
- Firm’s corporate culture. In such a case, the leadership of the organization expects managers to collude, and collusion occurs in many markets in which the firm operates. Firm norms and expectations of managerial behavior can repeatedly encourage collusion and “disregard” previous fines, as illustrated in the ADM case (Eichenwald, 2000).
- Firm structure. Multi-market collusion literature focuses on the ability of firms to target punishments in particular markets. Multi-market firms may also encourage the spread of collusion if they have learned to collude in one market and share their “best practices” in another. This seems to have been the case, for example, in the spread of the vitamin cartel from vitamins A and E to other vitamins (Connor, 2008). Multi-market collusion is encouraged not only by multi-product multinationals, but also multi-market relationships between what appear to be smaller firms in local markets. For example, if gas stations are owned by multi-market firms such as large oil firms or chains of stations, that may facilitate repeated collusion over time and/or across geographic locations.
In complementarity with LPs, Levenstein et al. (2015) discuss additional (possibly) effective post-cartel policies, aimed at preventing firm-driven recidivism.
- Company Fines and Leniency. Theoretical research has emphasized the aptitude of well-designed and well-run LPs to improve cartel detection and deterrence (for a survey, see Spagnolo, 2008). However, Marvão and Spagnolo (2015b) note the generosity of the current EU LP: the average LP reduction is 45% and leniency is granted to 52% of convicted cartel members. In addition, Marvão (2015) shows that repeat offenders appear to receive larger EC leniency reductions, which suggests that firms can learn the “rules of the game”, colluding repeatedly and reporting the cartel to reduce their penalties. As such, fines need to be tougher and recidivism needs to be dealt with differently.
- Individual Accountability. Senior management in EU cartels does not seem to suffer from their participation in cartels. For example, Robert Koehler became CEO of SGL Carbon in 2012, after being convicted in 1999 of price-fixing in the graphite electrodes cartel. Imposing tougher sanctions, such as individual prison sentences or disqualification of senior executives from employment in their sector or role, may prevent repeated collusive behaviors (in new firms) and thus, increase deterrence levels.
- Follow-On Damages. Private damage suits may increase deterrence. In the US, private litigation plays a major role in the enforcement of antitrust law. Conversely, access to private damages is relatively new in the EU. A recently adopted EU Directive on damages (11/2014) prevents the use of LP statements in subsequent damage actions. However, Buccirossi et al. (2015) show that the effectiveness of damage actions can be improved if the civil liability of the immunity recipient is minimized and claimants receive full access to all evidence collected by the competition authority. Access to previous cartel decisions, for a given firm, will increase the amount of available information and can increase the likelihood and/or amount of successful damage claims.
- Consent Decrees. These impose conditions on the behavior of convicted firms (e.g. maximum price, and transparency). If these are violated, the authorities intervene, thus lowering the cost of prosecuting recidivists. In the US, decrees were routinely used by the DOJ in the 1960s and 1970s, but the practice was abandoned due to concerns of effectiveness and large costs. More recently, in September 2007, the Brazilian Administrative Council for Economic Defense enacted a resolution that allows for the use of consent decrees with the aim to settle cartel investigations. Two have already been executed.
If recidivism is industry-driven, its prevention may require a different set of tools, including those below, to complement leniency.
- Structural Remedies. Competition authorities have repeatedly permitted mergers among former cartel members, often without review, let alone structural intervention. Davies et al. (2014) examine mergers among former cartel conspirators and conclude that only 29% of the mergers were investigated by the EC. Remedies such as disclosure, divestiture of assets, selling minority shares in competitors, or licensure of intellectual property to competitors may change the nature of competition in the market and make collusion more difficult (see Marx & Zhou, 2015 regarding post-cartel mergers). This is particularly relevant if recidivism is industry-driven.
- Monitoring and screening. Some antitrust authorities have implemented monitoring and screening techniques to identify anticompetitive behavior in a given industry. These initiatives involve the analysis or monitoring of the characteristics of products or market structures that are thought to be more prone to collusion (mostly due to repeated offenses). Some examples are watch lists (e.g. Australia, UK, Chile), price observatories (e.g. Belgium, Spain, France), statistical screens (e.g. US FTC, Korea FTC), gasoline retail in Brazil and public procurement in Sweden (see Abrantes-Metz (2013) for further details on screens).
While literal recidivism, i.e. the formation of a cartel after having been convicted of illegal collusion, appears to be rarely detected in the EU and US, there remain policy gaps closing which could improve competition post-cartel.
A variety of post-cartel policies should be explored for their ability to increase the likelihood that workable competition, rather than tacit collusion or single firm dominance, will emerge. These reduce the reliance of competition authorities on leniency-driven self-reports, which will in turn make leniency more effective and less amenable to strategic use by firms determined to collude.
- Abrantes-Metz, Rosa (2013). “Proactive vs Reactive Anti-Cartel Policy: The Role of Empirical Screens.” Available at SSRN: http://ssrn.com/abstract=2284740.
- Buccirossi, Paulo, Catarina Marvão, and Giancarlo Spagnolo (2015). “Leniency and Damages,” CEPR Working Paper DP 10682.
- Connor, John M. (2008). Global Price Fixing, 2nd ed. Berlin: Springer.
- Eichenwald, Kurt (2000). The Informant. New York: Random House.
- Grout, Paul and Silvia Sonderegger (2005) “Predicting Cartels,” Office of Fair Trading, Economic Discussion Paper.
- Levenstein, M., Marvão, C., Suslow, V., 2015. Serial Collusion in Context: Repeat Offenses by Firm or by Industry? OECD Global Forum on Competition. DAF/COMP/GF(10/2015)
- Levenstein, Margaret C., and Valerie Y. Suslow (2015). “Price Fixing Hits Home: An Empirical Study of U.S. Price-Fixing Conspiracies,” working paper.
- Marvão, C., 2015. The EU Leniency Programme and Recidivism. Review of Industrial Organization, 48(1), 1-27
- Marvão, Catarina and Giancarlo Spagnolo (2015a). “What do we know about the effectiveness of leniency policies? A survey of the empirical and experimental evidence,” in Beaton-Wells, C and C Tran (eds.), Anti-Cartel Enforcement in a Contemporary Age: The Leniency Religion, Hart Publishing.
- Marvão, Catarina and Giancarlo Spagnolo (2015b). “Pros and Cons of Leniency, Damages and Screens”. Competition Law and Policy Debate (forthcoming)
- Marx, Leslie M., and Jun Zhou (2015). “The Dynamics of Mergers among (Ex) Co-Conspirators in the Shadow of Cartel Enforcement,” working paper.
- Pirrong, Stephen Craig (1992). “An application of core theory to the analysis of ocean shipping markets” Journal of Law and Economics, 35(1): 89-131.
- Spar, Debora (1994). The Cooperative Edge: The Internal Politics of International Cartels, Ithaca: Cornell University Press.
- van Driel, Hugo (2000). “Collusion in Transport: Group Effects in a Historical Perspective.” Journal of Economic Behavior and Organization, 41(4): 385–404.
- Werden, Gregory, Scott Hammond, and Belinda Barnett (2011). “Recidivism Eliminated: Cartel Enforcement in the United States since 1999,” Georgetown Global Antitrust Enforcement Symposium, Washington DC, Sept. 22, 2011.
Leniency policies, widely used by antitrust authorities, aim to deter and dissolve cartels by granting a fine reduction (up to immunity) to reporting cartel members. What are the characteristics of the reporting cartel members? Marvão (2014) addresses this question by developing and testing a model where cartel members are heterogeneous in terms of the value of the cartel fine they expect to receive. The author shows that the first reporting firm in a cartel tends to be the cartel leader (in the US) or a repeat offender (in the EU). Reporting is also shown to be more likely in cartels which affect a larger market (in the US) and in cartels which have a lower number of members but which affect a geographical area wider than the EEA (in the EU).
Analysis of Leniency Policies
Cartels are a perennial problem and are one of the main concerns of the European Commission (EC) and the US Department of Justice (DOJ). As cartels are secret, measuring the rate of success of cartel detection is challenging. The increased number of detections in recent years may be the result of a higher desistance rate and/or a higher incidence of cartels. The US and EU Leniency Programmes (LPs) were thus designed to work as a device for the deterrence and dissolution of collusive agreements and have been in place since 1978 and 1996, respectively.
The DOJ’s decision on cartel fines is made in accordance with the “U.S. Sentencing Guidelines” and is, in the vast majority of cases, followed by plea-bargaining. The US Leniency Programme grants full immunity to the first firm coming forward, whereas the other firms receive no leniency reduction. However, plea bargaining is present in over 90% of cartel offences and the settlements often lead to a reduced fine for the subsequent cartel members. Firms are also liable for the damages caused by the cartel’s activity. In addition, the Amnesty Plus Program benefits prosecuted cartel members who disclose previously undetected cartels.
EU fines are set in accordance with the “EU Guidelines on the method of setting fines” and are adjusted to account for aggravating and mitigating circumstances. The total fine is capped at 10% of the total worldwide turnover of the firm in the previous year. In the current LP, the first reporter receives immunity from fines and the subsequent firms receive a reduction of 10-75%, depending on their place in the reporting queue.
The empirical literature on LPs policies is relatively short and recent. It focuses on the adequacy of the leniency reductions and presents conflicting results. However, an understanding of the characteristics of the reporting firms, and of the cartels in which they take part, is vital to make policies provide the correct incentives for firms so as to dissolve and dissuade cartels.
The Issue of Repeat Offenders
The current EU fine guidelines state that a repeat offender is any firm that was previously found to infringe Articles 101 or 102 of the EU Treaty. The DOJ defines repeat offenders as any firm that “after release from custody for having committed a crime, is not rehabilitated”. While repeat offenders are a serious issue, the LP Notices are not explicit as to whether or not they should receive a lower leniency reduction, if any.
Repeat offenders are also a highly debated issue. In Marvão (2012), it is shown that recidivism is one the factors which influence the granting and scale of EU leniency reductions. Connor (2010) has suggested that there is evidence of a significant incidence of recidivism, and identifies 389 recidivists worldwide in the period between 1990 and 2009. This number constitutes 18.4% of the total number of firms involved in 648 international hard-core cartel investigations and/or convictions. Werden et al. (2011) have contested Connor’s definition of recidivism and his calculation of the numbers of multiple and repeat offenders. The main discrepancy between the two arguments appears to be in how cartel members who merge and form a new firm are dealt with. Werden et al. (2011) follow the legal practice (DOJ and EC) and suggest that no repeat offenders have been fined in the US, since 1999.
The Model by Marvão (2014)
The aim of Marvão (2014) is to understand the specific characteristics of reporting cartel members and of the cartels in which they take part.
If firms are similar in everything but their own beliefs on the likelihood of being caught by the authorities, firms may have different incentives to report the cartel. Different beliefs may be generated from public statements issued by EU or US officials, knowledge of the budget allocated to the detection and conviction of cartels, and the proportion of convictions in cartel investigations, among others. Harrington (2013) formalizes this behaviour but his underlying assumption of homogeneity of firms only allows for symmetric equilibria.
Marvão (2014) extends the game in Harrington (2013) to include firm heterogeneity. In the first game stage, a two-firm cartel collapses for internal reasons. In the second stage, each firm receives a private signal on the expected probability of detection and conviction by the authorities. Given the signal received, and the expectations on the other firm’s behavior, firms decide to report if the signal is above their threshold level. In addition to the individual fine, the cartel sanction includes a payment for overcharges and other costs inherent to being fined. These costs may include attorney fees, negative impact on consumer’s perception (which may lead to lower sales), managers being fired, future punishment by other firms and possible future damage claims (from customers). Each cartel member can apply to the LP and receive a fine reduction.
The model shows that the cartel member with the highest expected fine will be the first to report the cartel, provided that it receives a sufficiently high and unbiased signal on the probability of being caught.
Empirical Evidence in Marvão (2014)
The theoretical model is tested with the use of data on cartel convictions. The US data employed in the empirical analysis is an excerpt from John Connor’s Private International Cartels dataset (1984-2009; 799 cartels). The EU data was self-collected by the author and includes 81 cartels in the period of 1998 to 2011.
US data on the individual turnover are not available, but sales and overcharges are likely to be larger for the cartel leader. Although this creates a further incentive to report the cartel, the US DOJ guidelines state that leaders cannot receive immunity from fines. It is thus surprising that the results show that, in US cartels, the leader seems to be more likely to report and receive immunity from fines. The cartel leader is identified as the firm mentioned in the DOJ decision as a ringleader or mentioned in the history of the case as the cartel disciplinarian/bully. This result suggests that different definitions of ringleaders are used, or that the rule is not always enforced by the DOJ.
In the EU, it is only the coercer of the cartel who is not allowed, since the LP of 2002, to receive immunity from fines. Although the EU public statements on cartel convictions do not identify the leader or coercer of the cartel, it is likely that the coercer is also the leader of the cartel. However, with no explicit data on the leader, the results cannot be obtained.
Surprisingly, the US results show that repeat offenders are more likely to receive immunity from fines. Even more concerning is the fact that this likelihood is larger with each additional repeat offender in the cartel.
The EU results show that firms that have colluded more than once are more likely to report the cartel and receive immunity from fines. This effect is particularly strong if the report occurs after the end of the cartel.
It may be that repeat offenders are larger in terms of sales or have better knowledge of how to interpret the signals received, perhaps due to their previous collusive agreements, and thus, are better at choosing the timing of the report and what evidence to provide the authorities with. Although it is in the authorities’ interest to give incentives to the reporting of a cartel, legislation should ensure that the deterrence effect is not diminished by the existence of excessive leniency reductions.
Reports are more likely to occur in US cartels which serve markets with a moderate and, to a lesser extent, large number of buyers; as well as in cartels which are shorter and smaller. This is perhaps because collecting evidence is easier and/or quicker. In addition, firms which are convicted in both US and EU are more likely to be the first reporter in the US if they received a lower EU fine, perhaps because they are quicker to report the cartel to the DOJ.
EU Reports are more likely to occur in longer and smaller cartels. The latter result is noteworthy as it contrasts with the work done in Sjoerd (2005) and Brenner (2009), where the number of cartel members is never significant.
In EU cartels reported after their end, the reporter is less likely to have received other reductions. Although these reductions could be due to firms claiming not to know that the agreement was illegal, it could also be that firms apply for other reductions if they do not expect to receive a (large) leniency reduction.
When the perceived probability of conviction is high, firms are more inclined to report the cartel. This prosecution effect is magnified by the existence of the EU and US Leniency Programmes. In addition, a pre-emption effect exists as when firms believe that other firms will report, there is an incentive to be the first reporter and apply for a fine reduction within the LP. Therefore, identifying the most likely reporter in a cartel is key to designing a successful LP.
Marvão (2014) shows that the main sources of fine heterogeneity are recidivism and leadership of the cartel, which illustrate the need for more proactive competition authorities.
Reports are also more likely in cartels that affect a larger market (in the US) and in cartels that have a lower number of members but which affect a geographical area wider than the EEA (in the EU). Leniency Programmes should thus be in line with these incentives, by focusing on dissolution of cartels in these markets and by increasing firm’s beliefs on the likelihood of conviction. This could be done, for example, through unannounced inspections, screenings and requests for information or for a meeting with a firm representative. These measures, provided that they are credible, would supplement and enhance leniency.
- Brenner, S., 2009. An empirical study of the European corporate leniency program. International Journal of Industrial Organization 27 (6), 639–645.
- Connor, J. M., 2010. Recidivism revealed: Private international cartels 1990-2009. CPI Journal 6, 2.
- Harrington, J. E., 2013. Corporate leniency programs when firms have private information: The push of prosecution and the pull of pre-emption. Journal of Industrial Economics 61 (1), 1–27.
- Marvão, C., 2012. The EU Leniency Programme: Incentives for self-reporting. Trinity College Dublin. Working paper.
- Sjoerd, A., 2005. Crime but no punishment. An empirical study of the EU 1996 leniency notice and cartel fines in Article 81 proceedings. Master’s thesis, Economic Faculty of the Universiteit van Amsterdam.
- Werden, G., Hammond, S., Barnett, B., 2011. Recidivism eliminated: Cartel enforcement in the United States since 1999. Research Paper