MASTER’S THESIS
Industrial Organisation, Regulation and Competition Policy
2015-2016
Rainy days for cartels:
The implementation of umbrella effects in the estimation of cartel damage
Abstract: The recent increase in private enforcement of antitrust law in the European Union has shed light on the fact, that a successful cartel not only harms its direct and indirect customers, but also customers of non-colluding firms. The economic phenomenon is called umbrella effect and arises because competing firms of the cartel face higher demand, as the cartel increases its price and, in turn, increase their price as well, resulting in damage to the customers of said firms. This harm is directly caused by the cartel.
In my thesis, I first analyse the magnitude of umbrella effects in a setting of Bertrand competition with differentiated products. I then study the difference in price set by colluding firms’ and non-colluding firms’ and discuss the impact of the theoretical implementation of umbrella effects in cartel damage estimation.
The results show that umbrella effects are significant and the implementation of umbrella effects in damage estimation leads to more robust results but also to lower overcharges. If umbrella plaintiffs are allowed to sue for damages in the near future, cartel members face turbulent times as the scope of liability towards claimants increases.
Author:
Emanuel Holler 11086157
Supervisor:
Prof. dr. Maarten Pieter Schinkel
October 10, 2016 Amsterdam
Statement of Originality
This document is written by Emanuel Holler, who declares to take full responsibility for the contents of this document.
I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in
creating it.
The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.
Table of Contents
1. Introduction ... - 1 -
2. Legal foundation of umbrella effects ... - 5 -
2.1 United States ... - 5 -
2.2 European Union ... - 8 -
3. Economic foundation of umbrella effects ... - 12 -
3.1 Decomposition of cartel damage ... - 13 -
3.2 Principles of umbrella effects ... - 16 -
3.3 Introduction to cartel damage estimation methods ... - 21 -
4. Analysis of umbrella effects in Bertrand competition ... - 27 -
4.1 Model misspecification by Inderst, Maier-Rigaud & Schwalbe ... - 27 -
4.2 Model solution with Shubik demand function ... - 30 -
5. Implementation of umbrella effects in cartel damage estimation ... - 40 -
5.1 Separating datasets from cartel firms and umbrella firms ... - 41 -
5.2 Pooling datasets from cartel firms and umbrella firms ... - 45 -
5.3 Economic trade-off ... - 47 -
List of Figures
Figure 1: Approach of the implementation of umbrella effects in damage estimation ... - 4 -
Figure 2: Welfare effect of the cartel ... - 13 -
Figure 3: Potential downstream damage caused by the cartel ... - 15 -
Figure 4: Umbrella effects with Bertrand competition ... - 17 -
Figure 5: Cartel effects on a market with three layers ... - 18 -
Figure 6: Cartel price pattern and but-for price ... - 21 -
Figure 7: Umbrella effects, product differentiation and increasing cartel size ... - 35 -
Figure 8: Significance of umbrella effects with increasing product differentiation and cartel size ... - 36 -
Figure 9: Umbrella effect and price difference with increasing cartel size ... - 38 -
Figure 10: Price difference and increase in substitutability ... - 39 -
Figure 11: Comparison of cartel overcharge, umbrella effect and weighted average effect ... ... - 48 -
List of Tables
Table 1: Possibilities to recover compensation for umbrella effects under several European
civil-law systems ... - 10 -
Table 2: Summary of findings on umbrella effects ... - 20 -
Table 3: Difference-in-difference approach stylised ... - 24 -
Table 4: Umbrella effects and cartel size ... - 28 -
Table 5: Umbrella effects and cartel size with symmetric non-cartel price ... - 29 -
Table 6: Umbrella effects and cartel size with price difference ... - 30 -
1. Introduction
Every economic textbook teaches the idea that perfectly competitive markets lead to a socially desirable outcome in which goods are produced in the most efficient way, which results in price being equal to marginal costs. But headlines in the media paint a different picture in which firms actively try to distort competition and increase prices above marginal cost level (Bos 2011, p.1). From a business perspective, this is understandable. Firms have an incentive to hinder competition, because collusive behaviour leads to greater profits compared to competitive behaviour (Hüschelrath & Weigand 2010, p.4). One mean of doing so is by forming a cartel, which can be defined as:
“[…] an agreement or concerted practice between two or more competitors aimed at coordinating their competitive behaviour on the market or influencing the relevant parameters of competition through practices such as, but not limited to, the fixing or coordination of […] selling prices […].” Directive 2014/104/EU of the European Parliament and of the Council on certain rules governing actions for damages under national law (2014), Chapter 1, Article 2, (14)
My analysis focuses on seller cartels, which is a group of firms which uses its market power to increase the price for the cartel good above the competitive level (Bundeskartellamt 2008, p.2). If a cartel is detected, it is the role of public enforcement to fine companies which were members of the cartel. Over the years, said fines have increased in order to discourage firms from colluding. The evidence however shows that almost no industry exists in which collusion has not taken place (Pepall, Richards & Norman 2011, p.235).
In recent years the number of private enforcement cases, which are antitrust lawsuits from private parties – competitors, suppliers or customers of a cartel- has increased rapidly in the European Union. Under private enforcement it is generally accepted that any individual harmed by an infringement of EU competition law may recover damages.1 Yet in practice, compensation claims have been reduced to the vertical production chain of the cartel. Thus only purchasers of the cartel good or their customers may recover damages (Maier-Rigaud 2014, p.1). But the cartel effect is not necessarily bound to the vertical production chain, but may also affect non-colluding firms and their customers. My thesis focuses on the latter.
So called umbrella effects typically arise when a cartel-induced price increase leads to a diversion of demand to substitute products. The producers of these substitutes react by
increasing their prices as well. Their customers suffer losses which are, economically speaking, directly caused by the cartel (Inderst, Maier-Rigaud & Schwalbe 2014, p.2). Throughout my analysis I will refer to firms that fall under the (price) umbrella of the cartel as umbrella firms or colluding firms. The price they set is denoted umbrella price or price set by non-colluding firms. And the damage that customers of umbrella firms suffer is referred to as umbrella damage.
The importance of umbrella effects is best visible in the case of the elevator cartel in several European countries. The Oberster Gerichtshof, the highest court in Austria, granted claimants umbrella damages in Kone vs. ÖBB-Infrastruktur2 and requested a ruling from the European Court of Justice (CJEU), which ruled that cartels are potentially liable for damages that accrue to customers of non-colluding firms. If claimants are allowed to proceed with such compensation claims, cartel members face a new group of potential plaintiffs, increasing the liability for damage recoveries. It is obvious that economists and lawyers alike need to have a thorough understanding of umbrella effects and their impact on damage estimation methods. Yet there is little literature concerning umbrella effects and the introduction of the latter into cartel damage estimation.
Blair & Maurer (1982) were among the first to analyse umbrella effects and discuss whether plaintiffs who purchased from non-colluding firms are accorded standing in court. Using a model with large firms at its core and a competitive fringe, which acts as price-takers and firms producing a homogenous good, the authors show that the umbrella effect is equal to the cartel effect. Purchasers from non-colluding firms pay the same price overcharge as direct purchasers of the cartel good, which is the difference in price between collusive periods and competitive periods.
A comprehensive work on umbrella effects is provided by Inderst, Maier-Rigaud & Schwalbe (2014). The authors show that umbrella effects arise is a variety of market settings, in both price (Bertrand) and quantity (Cournot) competition with homogenous and differentiated goods. The magnitude of umbrella effects depends highly on market characteristics such as the degree of product differentiation and the size of the cartel.
Davis and Garcés (2010), Oxera (2009) and the European Commission (2013) among others3, cover cartel damage estimation methods which are used in practice. In essence the authors
2 See case C-557/12, Kone and Others [2014].
3 The European Commission provides a practical guide: Quantifying Harm in Actions for Damages based on Breaches of Article 101 or 102 of the Treaty of the Functioning of the European Union (2013). See also van Dijk,
distinguish between cost-based and financial analysis-based methods, market structure-based and comparator-based approaches. While in principle we can use the same (econometric) tools in the damage estimation, in the case of damage caused by umbrella effects, the calculation of a simple cartel-induced overcharge may be grossly misleading (Inderst, Maier-Rigaud & Schwalbe 2014, p.5).
A recent working paper tried to combine both umbrella damages and their impact on damage estimation methods. Blair, Durrance & Wang (2016) show the economics of umbrella effects using a dominant firm model, which is comparable to the model used by Blair & Maurer (1982). The model uses homogenous goods and works under the assumption of an incomplete cartel. The authors proceed to estimate damages by using a cartel dummy variable approach for both cartel firms and non-colluding firms.
My thesis contributes to the literature on umbrella effects and cartel damage estimation methods and connects the two. While previous papers have greatly contributed to the economic understanding of umbrella effects, surprisingly, only one paper has tried to combine the microeconomic effects of umbrella damages and their implementation in (econometric) damage estimation methods. This lack in economic research is problematic. We need not only a coherent economic interpretation of umbrella effects but also a comprehensive analysis of the impact of implementing umbrella effects in damage estimation methods.
I aim at answering three research questions.
(I) Are umbrella effects significant and what factors influence its magnitude?
(II) On what factors does the price difference between the cartel price and the umbrella price depend?
(III) Should data on prices set by umbrella firms be included in the damage estimation, or should damage be estimated separately for cartel firms and umbrella firms respectively and what are the theoretical effects of either including or excluding said data?
The following figure describes my approach to this topic, where I excluded the introduction (1.) and the conclusion (5.).
T., & Verboven, F. (2005) for another theoretical and Friederiszick & Röller (2010), Laitenberger & Smuda (2013) or Nieberding (2005) for an applied approach.
Figure 1: Approach of the implementation of umbrella effects in damage estimation
After this introduction, in chapter 2, I briefly discuss the current legal standings of parties harmed by umbrella effects in both the USA and EU. I decompose cartel damage, show the factors influencing umbrella effects and introduce damage estimation methods in chapter 3. As the framework above could also be interpreted as a stylised approach on how to estimate damage in practice, the next step would be to use data with whatever method of cartel damage estimation has been chosen. Since I analyse the theoretical effect of including umbrella effects in the cartel damage estimation, instead of applying data to my estimation method, in chapter 4.2 I use a model of Bertrand competition with differentiated goods to derive equilibrium prices for both cartel firms and umbrella firms and a competitive equilibrium price. I also show the significance of umbrella effects and I answer research questions (I) and (II). In chapter 5. I discuss research question (III) and I recommend when we should estimate cartel damage based on both cartel prices and umbrella prices and when it would be better to use both datasets separately.
The conclusion in chapter 6 includes the discussion the results of my analysis and potential shortcomings of the model and provides an outlook on future research to be conducted on this topic. 2. Legal foundation of recovering (umbrella) damage 3.1-3.2 Economic foundation of (umbrella) damage 3.3 Damage estimation methods
Black box: cartel prices
and umbrella prices / but-for prices 5. Estimating overcharge using a reduced-form regression 5.3 Economic trade-off, arriving at overcharge 4. Model with differentiated goods in Bertrand competition (theoretically) given by
build from model parameters
review based on model parameters
5.2 Pooling datasets 5.1 Separating datasets
Source: own illustration. III
2. Legal foundation of umbrella effects
Before examining the economics of umbrella effects and their implementation into cartel damage estimation methods, I provide a short overview of the legal standing of umbrella effect plaintiffs in the USA and in the EU. The description of both jurisdictions is constructed from the historic basis from which antitrust laws developed, focussing on the USA. This is followed by short case reviews relevant to the topic of umbrella damage, and, concerning claimants in the EU, an outlook as to whether their standing in court may be successful.
2.1 United States
The origins of competition policy as we know it today can be traced back to the end of the 19th
century, primarily because of the formation of trusts and cartels (here and in the following Motta 2004, p.1 ff.). The rapid expansion of transportation and communication lead to a large single market in the USA, which incentivised firms to grow. This lead to an increase in competition for many industries, which resulted in price wars, to which firms responded by forming cartels and trusts in order to be able to keep prices and margins high. This harmed end consumers, smaller-size producers and other market participants. Although the government initially did not intervene, the growing power of trusts in the American economy alerted progressive social movements, which urged the government to do something (Page 2008, p.4). This led to the creation and adoption of the Sherman Act in 1890. The Sherman Act, however, only covered price-fixing, market-sharing agreements between independent firms, and monopolisation practices by single companies, but not mergers. Until the introduction of the Clayton Act, mergers were allowed, unless formed with the intention of monopolising the market. Firms which wanted to coordinate prices had the option of merging into a single firm and the Sherman Act could not be applied. In order to extend anti-trust legislation to cover mergers, the aforementioned Clayton Act was introduced in 1914. It explicitly forbids price-fixing and opens the way to recovering treble damages in court. Article 15 U.S. Code § 15 of the Clayton Act states:
“[…] any person who shall be injured in his business or property by reason of anything forbidden in the antitrust laws may sue […] and shall recover threefold the damages […]” Although it sounds hopeful for potential claimants who purchased from non-colluding firms, standings in cartel damage cases are, to say the least, uncertain (Blair & Maurer 1982, p.765).
This is because the scope of liability for antitrust violations is quite strict, mainly due to Supreme Court rulings in Hannover Shoe v United Shoe Machinery Corporation and Illinois Brick Co. v Illinois.4 In the former case, Hannover Shoe, a shoe manufacturer, accused United Shoe of monopolising the shoe machinery industry (Thimmesch 2005, p.1659). Hannover Shoe denounced United Shoe’s practice of “leasing and refusing to sell its more complicated and important shoe machinery” (392 U.S. 481 at 483-484), which had supposedly led to a price increase compared to a situation where Hannover Shoe would have been allowed to buy more advanced shoe machinery. United Shoe argued that Hannover Shoe had charged its customers a higher price and the alleged overcharge had been passed on to them, meaning that Hannover Shoe had not suffered any losses. The Supreme Court decided that pass-on, here used as a defence, would not be permitted and the right to sue for damages was granted to Hannover Shoe. In the latter case, the State of Illinois sued the Illinois Brick company and other manufacturers of concrete blocks for price-fixing agreements. Although the State of Illinois had purchased from contractors rather than directly from the cartel, it claimed damages as indirect purchasers of the cartelised product. Its damages claim was rejected, because the Court argued these would subject the defendant to “a serious risk of multiple liability” and would turn cases falling under Section 4 into problems of “massive evidence and complicated theories”.5 Both cases show the Supreme Court’s agenda regarding indirect harm, which were both dismissed. In turn, it meant that cartel damages were limited to direct purchasers of the cartel product only. The Supreme Court had confirmed this view in 1972, arguing “Congress did not intend the antitrust law to provide a remedy in damages for all injuries that might conceivably be traced to an antitrust violation”6 (Hansberry et al. 2014, p.197). The Court feared that treble damages would create an over-deterrence effect and potentially harm competition.7 Limiting the scope of liability to direct purchasers only intended to preserve the optimal deterrence effect of cartel sanctions (i.e. fines paid to competition authorities and cartel damages and interest payments awarded to those harmed by the antitrust violation) and to prevent duplicative recoveries. The first case to deal with umbrella effects was Mid-West Paper Products Co. v Continental Group, Inc.8 Mid-West Paper asserted price-fixing allegations among producers of consumer bags. The case included two kinds of plaintiffs: one were indirect purchasers of empty bags and
4 392 U.S. 481 (1968) and 431 U.S. 720 (1977) respectively. For a comprehensive economic analysis of the
consequences of denying pass-on damages, known as the Illinois Brick rule, see Schinkel, Tuinstra & Rüggenberg (2008).
5 431 U.S. 720 (1977), Id. At 745.
6 Hawaii. vs. Standard Oil Co, 305 U.S. 251 (1972). 7 See Werden, Hammond & Barnett (2011).
products packaged in such bags, and the other were direct purchasers of consumer bags from a subsidiary of one of the defendants and from non-colluding competitors of said defendants.9 The plaintiffs were denied standing to seek damages for overcharges which had resulted from the price umbrella set by the colluding firms (Blair & Maurer 1982, p.766). Part of the Court’s approach included an analysis as to whether the plaintiff’s alleged injury was ‘economically traceable to the defendants’ and the Court of Appeals concluded that “it is impossible to say with any degree of economic certitude if purchasers from a competitor of the price-fixing makers have been injured by the illegal overcharge” (Hansberry et al. 2014, p.199).
Only one district court, namely the Western District Court in Washington, accepted the right of a plaintiff to sue for compensation caused by the umbrella effect in State of Washington v American Pipe & Construction Co.10 Plaintiffs were able to sue for damages if the plaintiffs
were able to establish a causal link between the defendant’s action and damage and if such damage were not remote from the alleged illegal activity (conditions of remoteness and causation), which has to be checked from jurisdiction to jurisdiction. Although the case was not accepted for a summary judgement, it established, at least, the right of umbrella effect plaintiffs to sue for damages.
In line with the latter case, in Associated General Contractors v Carpenters of 1983 the US Supreme Court explained that in order to be successful, umbrella victims needed to link damage suffered by them to the cartel’s activity, and it was the court’s exercise to ‘[…] evaluate the plaintiff’s harm, the alleged wrongdoing by the defendants, and the relationship between them’.11 Thus, in principle, the Supreme Court accepted the right of plaintiffs to file a lawsuit based on umbrella effects (Hansberry et al. 2014, p.199). It defined several criteria which courts were instructed to use to determine a causal link between the alleged antitrust violation and harm and whether harm was not too speculative. The criteria, which in essence are still used today, are the following:
i) the nature of the plaintiff’s alleged injury;
ii) the directness or indirectness of the asserted injury; iii) the speculative aspect of the injury;
iv) the risk of duplicative recoveries;
9 Harvard Law Review, Vol. 93, No. 3 (Jan., 1980), pp. 598-608.
10 U.S. District Court for the District of Hawaii – 280 F. Supp. 802 (D. Haw. 1968). 11 459 U.S. 519 (1983).
v) the danger of a complex apportionment of damages; and/or vi) the existence of more direct victims of the alleged conspiracy.
Although this functional approach by the Supreme Court seems straightforward, the analysis to determine whether a plaintiff has standing to sue for damages and whether causation has been established sufficiently is complex. Because there is no coherent legal approach for umbrella effects, outcomes of cases may differ from jurisdiction to jurisdiction.12
Today the scope of liability for violations of antitrust infringements remains limited in the USA, and courts continue to follow a more restrictive view on whether umbrella plaintiffs are eligible to claim damages, even though, as we have seen, there are cases where plaintiffs have at least been granted the right to sue for damages. The position is still that the optimum deterrence standard combined with actions by courts and private plaintiffs is enough not to over-deter or to under-deter antitrust breaches.
From an economist’s point of view there has been plenty of criticism about only allowing direct purchasers of the cartel’s product to sue for damages, and leaving out umbrella effect victims and indirect purchasers, especially not recognising pass-on losses, an effect which should be even more intuitive to non-economists than umbrella effects. Han, Schinkel & Tuinstra (2009) find that the (price) overcharge does not measure the true harm inflicted by antitrust violations, and only allowing standing to direct purchasers underestimates true harm done.
If current antitrust law were to be in line with the Clayton Act, which was cited above, and every victim of antitrust breach could recover damages, further analysis on umbrella effects, such as mine, should focus on demonstrating the causal link between the harm inflicted and the alleged antitrust violation, and what harm is not too remote and should be granted standing in court. Previous rulings do not seem to be based in economic theory, but more economic analyses could lead to a reorientation of tort principles in current usage. So far, however, victims of umbrella effects have been denied standing, and no recent rulings indicate a change of mind.
2.2 European Union
The development of competition law in the European Union is closely intertwined with the historical development of its structure and has been influenced by the success of the
12 Hansberry et al. (2014) list Loeb Indust. v Sumitomo Corp., 306 F. 3d 469 (7th Cir. 2002); Laumann v Nat’l. Hockey League, Case No. 1:12-cv-01817 (S.D.N.Y.2012).
implementation of U.S. competition law.13 The historical development of competition laws has seen two levels of jurisdiction: one on a national level and one on a supra-national level (Motta 2004, p.9). Laws concerning anti-competitive practices on a supra-national level are found in articles 101 and 102 of the Treaty on the Functioning of the European Union (TFEU), which have to be implemented in national law within set time periods. Article 101 TFEU prohibits, inter alia, price-fixing. The formation of a cartel falls under so-called hardcore restrictions and is prohibited in all circumstances.
The principle of free competition, which we regard as the best way of allocating resources, was embedded in the 1951 Treaty of Paris (Motta 2004, p.13). Since then, free competition has been preferred over centralisation. Some of the points covered by article 101 TFEU can be traced back to said treaty, namely practices between firms which directly or indirectly distort competition. Said distortions include cases of abuse of dominance clauses, cartels and mergers. The main objective of the European competition policy which has been developed over the years is the elimination of any discrimination in the economic system. This is why the Court of the Justice of the European Union (CJEU) in Courage v Crehan14 and Manfredi15 ruled that generally any individual is eligible to claim compensation for harm suffered due to an infringement of antitrust law. But, as in the USA, the scope of liability for antitrust violations is disputed.
Only a few cases concerning umbrella effects have been brought in the European Union, namely cases in the UK and one in Austria.16 The latter paved the way to an EU-wide recognition as the Oberster Gerichtshof, the high court of Austria, requested a ruling from the European Court of Justice (ECJ). The authors compare different legal systems in the EU with respect to the chances of success under the civil-law systems of France, the United Kingdom, Poland and Austria, which are summarised below:
13 For a comprehensive analysis of European competition law, see: Jones & Sufrin (2014); for its historical
development: Chirita (2013).
14 C-453/99, Courage Ltd v. Bernard Crehan, 2001 E.C.R. I-06297.
15 C-295/04 to C-298/04, Vincenzo Manfredi v. Lloyd Adriatico Assicurazioni SpA, 2006 E.C.R. I-06619. 16 See Moy Park Limited & Others v Tessenderlo Chemie N.V., Case no. 1202/5/7/12, registered on 28 Sep. 2008
before the CAT; W.H. Newson Holding Limited & Others v IMI Plc & Others, Case no. 1194/5/7/12, registered on 17 May 2012 before the CAT and transferred on 24 July 2012 to the High Court of England and Wales, Nokia
Corporation v ALU Optronics Corporation & Others, case number HC09C04421, High Court of Justice Chancery
Table 1: Possibilities to recover compensation for umbrella effects under several European civil-law systems
Jurisdiction Possibility to recover damages Case examples
French civil law
§ chances of recovering damages seem to be very low
§ Economist’s expert reports can be challenged by courts (1) and standard evidence of the demonstration and quantification of damage is high, leading to claims by direct purchasers rarely succeeding
§ Proving direct causation between anticompetitive practice and price increase is complex and the burden in this respect remains with the plaintiff
§ Passing-on is presumed and plaintiffs must provide evidence of not passing on overcharges to their customers (2)
(1) Arkopharma v Roche, Nanterre Commercial Court, 11 May 2006 (2) Doux Aliments v Ajinomoto Eurolysine, 15 June 2010 and Gouessant and SOFRAL v Ajinomoto Eurolysine, 15 May 2012, both French Cour de cassation, English common law system
§ More promising prospects for plaintiffs
§ Again, the causal link between anticompetitive conduct and increase in price by non-colluding firms must be shown and needs to be within the scope of foreseeable risk
§ quantification of damage by estimating but-for prices (prices which would have prevailed in the cartel’s absence) is accepted by English courts (1) and the complexity of damage calculations is manageable for umbrella effect plaintiffs
(1) Arkin v Borchard, [2005] EWCA Civ 655
Polish civil law system
§ no claim or case has been brought to Polish courts to date § problem of data compilation, civil litigants are not permitted to
view evidence, parties can only freely lend evidence to other parties, which in a cartel case does not seem likely
Austrian civil law
§ in Austria and other member states civil tort rules apply, and plaintiffs must, again, demonstrate fault, damage and direct causal link between the two
§ one case, the elevator cartel which is discussed below, has been brought, leading to an enquiry from the Austrian Supreme Court to the CJEU (decision on umbrella effects still pending) § as of right now, claimants of umbrella compensation will not be
able to sue for damages, as those indirect losses, which are merely a side effect of the cartel, are not granted under (current) civil law
Oberster Gerichtshof, 17 Oct. 2012, 7 OB 48/12b, KONE AG,
Otis GmbH, Schindler Aufzüge und
Fahr-treppen GmbH, Schindler Liegenschafts-verwaltung GmbH, ThyssenKrupp Auf-züge GmbH v ÖBB-Infrastruktur AG
Source: Hansberry et al. (2014), p.201 ff.; Strange, Butler & Ceglia (2012), p.2 ff.
Between 2003 and 2007, several European manufacturers of elevators and escalators, namely Kone, Otis, Schindler and ThyssenKrupp, colluded in several Member States of the European Union. The members of the cartel coordinated their activities with respect to more than half of the volume of new machinery in the Austrian market. In those cases, at least one third of the
market was split by allocating contracts between the colluding firms. ÖBB Infrastruktur, a company owned by the Austrian government, purchased elevators both from cartel members and non-cartelist suppliers and tried to recover damages for umbrella effects also. The court of first instance rejected the claim, as judges classified the increase in price to be an indirect effect of the cartel for which cartel members would not be liable. ÖBB Infrastruktur went to the Court of Appeal, which overturned the initial ruling and allowed the demand for compensation pursuant to Article 101 TFEU. It argued that the increase in price by non-colluding firms was objectively predictable and inevitable.17 The Austrian Supreme Court, as a last instance, ruled that each company in a market sets its prices with respect to its corporate strategy and business agenda. Because a cartel outsider can increase its prices for a variety of reasons, which, according to the Court, are not foreseeable by the cartel, the chain of causality between anticompetitive conduct and increase in price is interrupted. The Court concludes that following national civil law, cartel members cannot be sued for these losses. The Supreme Court, however, raised a question about the efficacy of Article 101 TFEU and directed it to the Court of the Justice of the European Union. In its answer, the CJEU argued that:
“[…] the victim of umbrella pricing may obtain compensation for the loss caused by the members of a cartel, even if it did not have contractual links with them, where it is established that the cartel at issue was […] liable to have the effect of umbrella pricing being applied by third parties acting independently, and those […] aspects could not be ignored by the members of that cartel.” (Kone C-557/12, §34)
However, it is for the referring court in each member state to determine whether a direct causal link between the alleged antitrust infringement and the harm suffered by umbrella victims exists. If said causal link is not accepted by national law, plaintiffs would not be allowed to sue for damages.
Lastly, I refer to the argumentation by the European Commission regarding the merger between Hutchison 3G Austria and Orange Austria.18 In the chapter on the post-merger reaction of competitors, it says:
“[…] Generally accepted and robust economic theory demonstrates that the profit-maximising response of competitors to a price increase [of the merged firm] would be to increase prices themselves. The rationale behind this expectation is the following: if the merged entity were to raise prices, some customers would consider switching to one of the other […] providers who would not have done so in the absence of the merger. […]
17 Oberster Gerichtshof, 17 Oct. 2012, 7 OB 48/12b., p.14.
These newly available customers then increase the demand faced by the other competitors as a result of which they have an incentive also to increase prices themselves.”
In the paragraphs quoted, the European Commission refers to the work of Deneckere & Davidson (1985), whose model I use in my analysis of umbrella effects, as explained in the introduction. The two authors explicitly argue in their paper that the merger they analyse may also represent a cartel (Deneckere & Davidson 1985, p.473 ft. 2). If a price increase can be expected for competitors of a merged firm and the European Commission accepts that said increase has been caused by the merged entity, the question arises as to why this causal link is not applied to cartels and used to explain umbrella effects.
In summary, the chances of recovering damages in the USA are low. In the EU, however, the CJEU’s response to the Austrian Supreme Court’s request has brought significant changes regarding umbrella effects under EU law, yet the practical implications of the ruling remain to be seen as, in both jurisdictions, claimants are required to show the causal link between the alleged antitrust infringement and the damage it caused. But the European Commission has already accepted this causation in public enforcement. If national law with respect to private enforcement were to be changed by one of the member states in order to facilitate umbrella claims, this might lead a chain reaction if more jurisdictions regard umbrella plaintiffs as eligible candidates to sue for damages and facilitate showing the causal link between the alleged injury and suspected harm.
Economically speaking, showing the causation is straightforward. Although economic models do not necessarily reflect reality, it seems unwise to ignore simple economic facts when it comes to the legal enforceability of competition law. Incorporating this economic causation into competition law is a necessary step in the near future of private enforcement.
3. Economic foundation of umbrella effects
Although umbrella effects are legally disputable, from an economic perspective it should be obvious that a stable cartel harms not only its customers but also those who purchase from non-colluding firms in the same relevant market or in substitute markets at cartel prices (Maier-Rigaud 2014, p.249). In the subsequent chapter, I will first decompose cartel damage and show how customers of colluding firms are harmed. This is followed by an explanation of the economic principles behind umbrella effects, with the aim not only to introduce to the basic economic intuition behind umbrella effects but also to show the variety of factors which
influence its size. Lastly, I will give a short, coherent overview of cartel damage estimation methods, which can be used also to implement umbrella effects.
3.1 Decomposition of cartel damage
Firms which form a cartel tend to increase prices or reduce quantity supplied in order to maximise joint profits. If successful, the cartel price can be approximated by the price set by a monopolist which is found at the intersection of (aggregate) marginal revenues and marginal costs. In comparison to a competitive market this reduces quantity and increases price.19 Because prices are higher in a cartelised market than in a competitive market, cartelists are able to redistribute some of the consumer surplus to the cartel. Also some consumers decide not to purchase the product at the new cartel price. Both effects are illustrated in the figure 2 below, in which a cartelised good is sold directly to the end consumer:
Figure 2: Welfare effect of the cartel
The new market equilibrium can be seen in the figure. The area labelled Rent transfer is the economic rent transferred from consumers to producers. Consumers now pay the cartel price
19 Even with functioning competition, the competitive price might be higher than in perfectly competitive markets.
This would be the case with oligopolistic markets. For simplicity reasons, I assume a perfectly competitive market, where !"= $%. S D Quantity Price DWL !" !# $" $#
Source: own illustration based on Davis & Garcès 2009, p.349.
!& instead of the competitive equilibrium price !" and they purchase '& instead of '".20 Area DWL (deadweight loss) is the welfare loss. This is consumer surplus which is lost in the process of cartelisation because it is not redistributed to any agent in the market. The deadweight loss arises as a result of the price increase induced by the cartel. Also there are consumers who are willing to pay a price between !& and !" but due to cartelisation those consumers are not served. This is also called the lost-volume effect, as consumers buy a smaller amount of the cartel good compared to competitive periods (Oxera 2009, p.14). This is inefficient, as those consumers would be willing to pay the price under more competitive conditions.
From a welfare perspective, economic rent transfer is not actually a welfare loss or in itself inefficient. It is merely a transfer between consumers and producers. For consumers, however, this constitutes damage because they buy the cartelised good for a higher price. This damage is referred to as an overcharge, the difference in price, thus !&− !". The total damage for consumers, assuming the cartel sells to direct purchasers only, is the sum of economic rent transfer and the deadweight loss, which is greater than the economic damage and also greater than the profit achieved by the cartel.
Although the total harm is the sum of both areas Rent transfer and DWL, the DWL is generally ignored in the estimation of cartel damage and consumers can only claim the overcharge multiplied by the quantity purchased: '& !& − !" . This assumption of damage to consumers being equal to the overcharge is not a bad one [in cases with direct customers], as area DWL is relatively small compared to area Rent transfer (Davis and Garcés 2010, p.349). Also it is hard to argue what quantity would have been purchased in the case of a lower cartel price. The lost-volume effect is more difficult to quantify than is the overcharge harm (Oxera 2009, p.15). But if we assume that the market consists of n-layers and the cartel sells to intermediaries who depend on the cartel product as input, the decomposition of damages becomes more complex. We now have to distinguish between direct and indirect damage. Figure 3 below depicts said situation for a market with three-layers:
20 In my analysis and in this chapter I assume no cartel effects on product quality, thus prices are at a constant
Figure 3: Potential downstream damage caused by the cartel
The cartel damage results in three parts, namely direct and indirect damages. Following van Dijk & Verboven (2007) 21, who assume downstream firms purchasing from a cartelised market, we can decompose the damage into three terms. The first two terms below describe direct damages for customers of the cartel, the third and last depicts indirect damages:
• The first effect is the direct overcharge (1) by the cartel, which leads to an increase in costs and downstream firms suffer a decrease in profits.
• The second effect is the “output” effect (2), which occurs because in the absence of cartels the downstream firms would have sold an extra ('& − '") units and earned a margin (!& − !").The output effect is rarely taken into consideration in cartel damage estimation cases.
• The third effect is the increase in profits earned by the downstream firms, as part of the cartel overcharge is passed on to layers farther below in the production chain. It is called the pass-on effect (3) and it decreases the damage suffered by the downstream firms. Customers of the downstream firms however, suffer damage, depending on how much of the cartel-induced price increase is passed on by the downstream purchaser.
21 The authors provide an overview on passing-on and is inclusion in damage estimation methods as well as an
analysis of its legal standing.
!
"
B"
C indirect customers cartel firms"
A"
F"
G"
H"
I direct customersSource: own illustration based on Friederiszick & Röller (2010).
Overcharge (1) Output effect (2)
In a market with more than two layers, damage attributable to the cartel consists not only of the overcharge (and DWL) as in figure 2. It should be obvious that the output effect and the pass-on effect in particular, increase damage suffered by customers alpass-ong the productipass-on chain. However, we have seen that pass-on plaintiffs, especially in the USA, stand little chance when it comes to claiming damages. But if passing-on damages or more generally damages suffered by indirect purchasers is being ignored, the overall damage attributable to the cartel is far greater than the overcharge damage which can partially be recovered.22 Even if pass-on damages can be recovered by those who have suffered them, we still assume that damage only accrues along the production chain.
3.2 Principles of umbrella effects
But it is straightforward to show that the latter does not hold true and damage also arises for parties which, at the first glance, do not interact with either cartel firms, nor customers of the cartel. This chapter explores the economics behind umbrella effects. I show that a price increase by the cartel is met by a price increase by umbrella firms with a simple textbook example of Bertrand competition. I then elaborate on the magnitude of umbrella effects.
The basic economic principles explaining why umbrella firms increase their prices as well as the cartel can be seen in a standard textbook example of Bertrand competition with strategic firms. In standard demand systems, prices are strategic complements, which means that a price increase by a competing firm is optimally followed by an increase of one’s price (Inderst, Maier-Rigaud & Schwalbe 2014, p.11 ff.). Applying this logic to a (representative) cartel firm and a (representative) non-colluding firm, we would expect the non-colluding firm to increase its price also, depending on the firms’ best response functions. The best response functions indicate how a firm sets its optimal price for every choice of price by its rival.23 If demand is linear, so are the best response functions of the two representative firms. Of course there must be more than two firms in the market for a cartel and a cartel outsider to exist. In this example, I assume three firms to be in the market, two of which collude. In figure 4, the best response functions are drawn and the cartel effect and umbrella effect are highlighted:
22 For an encompassing analysis see Han, Schinkel & Tuinstra (2009).
23 Deneckere & Davidson, Davis & Garcés among other use the term reaction function, whereas Inderst,
Maier-Rigaud & Schwalbe use best-reply function. I prefer using the term best response function (Pepall, Richards & Norman 2011, p.161). All terms describe the same concept namely how the firm trades off an increase in the price of its own product, resulting in a higher margin with an increase in the quantity produced.
Figure 4: Umbrella effects with Bertrand competition
In the competitive Nash equilibrium firms set price !", the price which can be found at the intersection of the respective best response functions +,& and +,-. Basically, each firm anticipates what price its competitors will set and includes this in their own profit maximisation. If firms now form a cartel, that does not include all of the firms in the market, the equilibrium changes. If we assume the cartel members to behave jointly optimally, the best response function of the (representative) colluding firm in the figure shifts to the right and is denoted
+,′&. The best response function by the (representative) non-colluding firms remains
unchanged. The new equilibrium solution is, again, found at the intersection of both functions yielding equilibrium prices !& and !-. The cartel effect is given by the increase in price (!& −
!"), as marked in the figure. The umbrella effect is characterised by (!-− !"). Thus we have
seen that it is straightforward to show umbrella effects in a simple framework.
The model shown above is considered a static model. Static models of partial cartels address the question whether a stable cartel exists, and if this is so, what firms will participate in the cartel. In the example above, I assumed that the cartel, consisting of two firms and one firm remaining outside the cartel, is stable. A cartel is stable, if they are internally and externally stable (Schwalbe 2011, p.3). Internal stability refers to the decision by a firm inside the cartel
!
"!
# !# !$ !" !$Source: own illustration, based on Deneckere & Davidson (1985), p.477. Umbrella effect Cartel effect 45° ()# ()" ()′"
whether it should remain in the cartel or not. External stability examines incentives for firms to join the cartel in the first place (D’Aspremont et al. 1983, p.21 ff.). Cartels are internally and externally stable if the firms inside the cartel prefer to abide by cartel agreement and firms outside the cartel prefer to stay outside. For my analysis I assume that stable cartels exist within my theoretical model and so does the basic example of Bertrand competition above.
Who is harmed by a stable cartel is best visualised in a market, again with three-layers. There are manufacturers, retailers, who purchase an input good from the manufacturers and end consumers, who purchase a homogenous good from the retailers. The input good however, provided by the firms in the top layer is differentiated. The following is a stylistic figure of said market:
Figure 5: Cartel effects on a market with three layers
The figure shows horizontal and vertical cartel effects. Horizontal effects concern producers of substitutes in the same relevant market or producers of substitutes at cartel prices. Vertical effects arise at different steps in the production chain to firms and end consumers who are direct or indirect purchasers of the cartelised good or substitute goods.24 Neither horizontal nor
24 For the sake of simplicity, potential effects on manufacturers, retailers and customers of complementary products
are not included here. It can be envisaged that suppliers of complementary goods suffer damage due to a loss of
!
non-colluding firm(s) indirect customers cartel firms" "
"
relevant market"
substitute market at cartel price"
A B C D E"
F"
G"
H"
I"
J!
Kvertical effects of the cartel occur to customers of firms competing against the cartel (Inderst, Maier-Rigaud & Schwalbe 2013, p.12). We refer to those effects as umbrella effects.
Vertical effects, as I have explained in 3.1, accrue to direct or indirect purchasers of the cartelised good. It should be obvious that direct purchasers (firms F and G) of the cartel are harmed due to the cartel-induced price increase or quantity reduction. As those firms buy input from the cartel and produce a homogenous good, the magnitude of cartel damage accrued depends on how competitive the market is, as it determines how much of the cartel-induced price increase is passed on to the indirect purchasers. As we have seen, pass-on might decrease the damage which can be recovered from cartel members.
Horizontal effects are the reason not only direct purchasers of the cartel F and G and their indirect customers are harmed by the cartel’s activity. The cartel-induced price increase leads to a diversion of demand from cartel firms A, B and C to their competitors F and E (Maier-Rigaud 2014, p.249). Facing an increased demand, a profit-maximising non-colluding firm will typically increase its price as well. This economic phenomenon is known as the umbrella effect. In order for umbrella effects to arise, the cartel has to be incomplete, thus not all firms are part of the cartel. This increase in price by firms D and (to an extent) E harms firms H, I and J because they pay higher than competitive prices. Although firms H, I and J do not (in)directly interact with the cartel or are at the same stage of the value chain, they are harmed nonetheless. This is why umbrella effects are strictly-speaking not directly attributable to either horizontal or vertical effects (Inderst, Maier-Rigaud & Schwalbe 2013, p.12).25
Firms under the umbrella of the cartel tend not to only increase their respective prices, but rather also the quantity produced. But the increase in production will not offset the cartel effect. If it did, the cartel would not have been formed in the first place. Although firm D (and potentially E) profit from the cartel’s existence, as both increase prices, it is important to remember that both parties are not free-riding under the umbrella of the cartel, which would not be allowed under antitrust law, but merely act as profit-maximising firms.
sales and profits induced by the cartel’s price increase (Clark, Hughes & Wirth 2004, p.16). The question as to how significant umbrella effects are in these constellations is an interesting one and warrants to further research.
25 There is an exception to firms H, I and J being harmed, an effect that is called a negative umbrella effect. Under
specific circumstances, it may be that firms H and I (and J) profit from the cartels’ existence. In the market described above, firms F and G buy input goods from the cartel directly, whereas firms H and I purchase their input from umbrella firms. If all firms in the downstream market produce a homogenous good and compete for (indirect) customers, firms H and I have a competitive advantage over the other two firms, if those are unable to pass-on overcharge from the cartel. In this thesis I will not cover negative umbrella effects and my model does not allow for them. However, this would post an interesting future research question.
The intensity of diversion of demand and therefore the umbrella effect depends on the demand from downstream firms, of course. The magnitude of umbrella effects is determined by a variety of factors (Inderst, Maier-Rigaud and Schwalbe 2014, p.5 ff.). Umbrella effects mainly depend on the type of competition (Bertrand or Cournot), the type of good (homogenous or differentiated goods), market coverage of the cartel, the degree of product differentiation. The difference between both types of competition is that with Bertrand competition, a price increase by one firm is usually followed by a price increase of the other, as goods are strategic complements. With Cournot competition, a decrease in output by one firm leads to an increase in output by the other, as prices are considered to be strategic substitutes (Ivaldi et al. 2003, p.23). The type of competition can either dampen or increase the magnitude of umbrella effects; it is the combination with other factors which would allow for a generalisation. With homogenous goods, the authors find positive umbrella effects. With Bertrand competition the umbrella effect is equal to the cartel effect. With Cournot competition, the cartel effect will partially be offset by the increase in quantity by umbrella firms. We expect a less pronounced umbrella effect with Bertrand competition with differentiated goods, as my analysis will show, because products are not perfectly substitutable. Cournot competition with differentiated goods gets the same result as with homogenous goods. The price increase by the cartel is only partially offset by the quantity expansion by umbrella firms, depending on the degree of substitutability. Both market coverage and degree of substitutability are positively correlated with umbrella effects, which I show in my analysis.26 If the cartel were only to have a small coverage of the market, only a small part of the demand would be diverted to non-colluding firms (Inderst, Maier-Rigaud & Schwalbe 2014, p.13). The same is true for a low degree of product substitutability. The table below summarises the authors’ findings for both Bertrand and Cournot competition and homogenous as well as differentiated goods:
Table 2: Summary of findings on umbrella effects
Type of competition Homogenous good Differentiated good
Bertrand competition
Price increase is the same for both cartel and non-colluding firms, increase in input by non-colluding firms is not enough to outweigh decrease of price set by cartel. If firms are not capacity-constrained, no umbrella effect arises, as umbrella firms meet all demand that is diverted to them.
Depending on degree of
substitutability and market coverage of the cartel: higher umbrella effect with higher substitutability, lower umbrella effect with smaller market share
26 It turns out there is a conceptual mistake in the paper by Inderst, Maier-Rigaud & Schwalbe (2014). My analysis
will therefore first show where the authors went wrong and then proceed by solving a model based on another, more general demand function. In doing so I confirm the statements made above, which is why they are listed here. The conceptual mistake does not change or falsify the qualitative statements by the authors.
Cournot competition
Decrease of quantity by the cartel leads to an increase in price, fringe firms increase output, but not enough to offset the cartel effect
Price increase by the cartel will not be offset by output increase of non-colluding firms. The outsiders’ quantity expansion is less
pronounced when products are more differentiated
Source: own illustration based on Inderst, Maier-Rigaud & Schwalbe 2014, p.6-14.
In this chapter we have seen that umbrella effects arise in a variety of settings. This underlines the need for more research on the topic. My analysis focuses on Bertrand competition with differentiated goods.
3.3 Introduction to cartel damage estimation methods
I now turn to the estimation of the cartel damage. The methods described in this chapter can also be used to analyse umbrella effects. The basic idea behind cartel damage estimation methods is to estimate the cartel overcharge, which is the difference between the cartel price and a hypothetical competitive price - the so-called but-for price - a price which would have prevailed if the cartel had not existed. The following is a graphical illustration of this:
Figure 6: Cartel price pattern and but-for price
The cartel overcharge in percent multiplied by the quantity purchased gives the cartel damage for which compensation can be claimed from the cartel members.
There are a number of both theoretical and empirical ways to estimate a cartel overcharge or estimate but-for prices in order to calculate the cartel overcharge. Following the European Commission’s practical guide for Quantifying Harm in Actions for Damages based on Breaches of Article 101 or 102 of the Treaty of the Functioning of the European Union (2013, here and in the following: EC practical guide 2013), I briefly discuss three main methods of quantifying
90 95 100 105 110 115 120 125
Jan 06 Jan 07 Jan 08 Jan 09 Jan 10 Jan 11 Jan 12 Jan 13 Jan 14 Jan 15
Source: own illustration based on consumer price index for sugar (2010=100), Federal Statistical Office Germany.
Cartel overcharge
cartel damages. Although the guide does not go into detail using econometrics or microeconomic theory, it provides a general overview on the approaches used in practice and demonstrates advantages and disadvantages for each of the methods. However, I discuss one of the methods, namely the cartel dummy variable approach, more detailed, as I use the latter for my analysis.
The European Commission divides methods into the following two categories: cost-, financial analysis-based and simulation model-based approaches and comparator-based methods. The structure of the Commission is comparable to the one by Oxera (2009), as their study was conducted for the Commission. Davis & Garcès (2010) discuss the same approaches, the only difference is the designation of each of the main methods. The sources for this chapter, among those not depicted above, agree with the generally used methods in practical damage estimation. Starting with cost-based and financial analysis-based approaches, these models have been developed in finance theory and practice. Of course, the line between financial and non-financial methods can be blurred, because forms of non-financial analysis are often involved at some stage of the analysis. Cost-based methods try to split up prices into their essentials, mainly production costs and a mark-up to represent profits. The difficulty in employing a cost-based method is that cartel members do not face intense competition during the alleged cartel period, which, if data for that period were to be used, might lead to an overestimation of the but-for price. Another difficulty in itself might be to gather company-specific data and working under the assumption that in competitive periods price equals marginal costs, whereas the actual ratio of cost to price might be different (Davis & Garcès 2010, p.363). Financial-based methods take the claimant’s financial performance and try to establish a counterfactual scenario by calculating the claimant’s profitability had the cartel not existed. This can be done by comparing profitability to firms which are active in other markets or by calculating profitability for the claimant outside the infringement period (EC practical guide 2013, 114-118).
Market structure-based methods, also referred to as simulation models, use economic models to simulate the difference between competitive and collusive prices. Depending on the case at hand, models can range from monopoly or oligopoly models to perfect competition or any conceivable market structure in between. Data can be used to determine the outcome of the model, i.e. data on demand and supply can be used to estimate demand and supply functions which then can be solved for equilibrium prices (Davis & Garcès 2010, p.364). Drawbacks inherent in this approach might be that the required amount of data is not (publicly) available for specifying complex models or that the assumptions which are necessary in order to be able
to solve the model might affect its results. Also whether legal systems take the use of an economic model into consideration depends on the features of the case in question.
Comparator-based methods try to estimate a non-infringement scenario based on:
§ data on the same market but not from the infringement period (before and/or after) § data on different but comparable geographic markets
§ data on different but comparable product markets.
Of course, said data can also be combined (EC practical guide 2013, 33).
The before and/or after methodology employs time series data, usually price data of the cartelised good. It analyses prices paid for the good in periods of competition and compares them to periods of collusion. The cartel damage is then calculated as the difference in prices multiplied by quantity purchased (Davis & Garcès 2010, p.354). The method is very simple and might result in a good approximation of cartel damage. There are two main critical points about this method: one is that knowing the cartel period is vital. Prices set by the cartel may prevail after the alleged cartel period. In that case economists trying to estimate damage should consider potential ripple effects and distorted competition after the break-down of the cartel, which may include punishments from former cartel members.27 If data is employed after the
cartel has allegedly stopped, but collusive prices prevail or price wars occur, employing the before and/or after method might lead to an underestimation or overestimation of the cartel damage. The other critical point is that the approach assumes that market characteristics have not changed over time. If this is not the case, it may well be that the difference between competitive and collusive periods is not due to the infringement. If so, it is possible either to choose a non-time-based comparison in the same market but choose a different comparator market or to adjust the analysis accordingly so that it takes account of changes in demand and supply. Whether data from both periods, before and after the cartel, should be used, has to be determined depending on the market characteristics (EC practical guide 2013, 38-45).
The other two comparator-based methods are the comparison with a different geographic market and/or the comparison with a different but comparable product. Those approaches are also known as yardsticks. When conditions do not allow for using a before and/or after approach, the but-for price can be constructed using a benchmark of a comparable product in the same geographical market or the same product in a different geographical market, or both
27 For a valuable discussion on the effects of the time of termination of a cartel, see Hüschelrath, Müller & Veith
a comparable product in a different geographical market (Davis & Garcès 2010, p.360). The market chosen has to have similar conditions compared to the cartelised market had it not been cartelised. The limitations of yardsticks are similar to the ones of a before and/or after method. Again, we have to assume that economic variables and their relationship with one another do not change. If it did, yardstick approaches would not be able to differentiate between the cartel-induced price increase or fluctuations in price which are caused by other, exogenous variables. Both before and/or after methods or yardstick methods are a helpful tool nonetheless, but the effect of the cartel should be apparent using different methods in order to verify the robustness of the results (Davis & Garcès 2010, p.352).
The combination of data both over time and across markets is called the difference in difference method and can be visualised as such:
Table 3: Difference-in-difference approach stylised
Competitive period Collusive period Damage estimation
Cartelised market A B
(B-A) – (D-C)
Comparative market C D
Source: own illustration based on Oxera (2009).
The damage estimated not only takes the cartel effect (B-A) into account, but rather acknowledges changes in price in the comparator market which would have affected the price had the cartel not been present. Combining the comparator-based methods allows checking for changes in demand and supply, an obstacle which simple before and after methods would not overcome. The approach relies on the assumption that both markets are affected similarly by the changes in the comparator market (EC practical guide 2013, 56-58). Also, it is difficult to obtain data from both markets in order to be able to apply the said estimation method, and data might be biased if umbrella effects are present and care must be taken to ensure that no firms in the comparator market fall under the price umbrella of the cartel, effectively underestimating the damage which has actually occurred.
The shortcomings of the comparator-based approaches above are that in some cases change in demand and supply cannot be incorporated in the estimation method. One way to overcome this criticism is by running a reduced-form regression with a cartel dummy variable, which builds upon a structural demand and supply model (Davis & Gacès 2010, p.357). The goal of using a reduced-form approach is to isolate the effect of the anticompetitive conduct, which is why economists employ data also from the collusive period, which is why it is sometimes referred to as the before, during, and after method. In order to correctly estimate an overcharge, the
analysis conducted must be able to precisely distinguish between anticompetitive effects and other, exogenous effects on price. In practice, the dummy variable approach is widely used to estimate overcharges, checking for exogenous factors which may influence the cartel’s price-setting behaviour.
The general idea is to construct the price regression from a demand and supply model, which can be solved under the assumption that demand equals supply in equilibrium. In short, following Nieberding (2005), we could assume demand (1) and supply (2) are represented by the two following equations: 28
'/0=∝"+∝34/+∝56/
'/7= 8"+ 834/+ 859/+ 8:;/ Quantities are represented by '/0 and '
/7, 4/ is the market price; 6/, 9/ and ;/ are exogenous
demand and supply shifters, respectively. Because equilibrium prices and quantities are jointly determined, we would have to employ proper econometric techniques (e.g. 3SLS) if the equations are identified. The expected signs of each parameter will depend on the reason why they are included in the model. Nieberding’s example is if 9/ were to be a cost-shifter, a reasonable expectation would be that for its parameter 85 to be negative, reflecting that when cost increases, quantity decreases (everything else being equal). But, by using the abovementioned equilibrium assumption, (1) and (2) can be solved, which yields the reduced-form model (3):
4/= <"+ <39/+ <56/+ <:;/
Mainly two approaches are applied when using a reduced-form regression. One is the cartel dummy method, which uses data from all available periods. The other is the forecasting approach, which predicts but-for prices using regression coefficients with observed values of the independent variables in all the periods except the cartel period. The approach excludes this data in order to yield a reliable and precise result, but will not be further discussed here (Nieberding 2005, p.369).
The cartel dummy variable approach is characterised by Rubinfeld and Steiner (1983) as follows:
“In it [, the dummy-variable (DV) approach], one would estimate the econometric model for all time periods for which there was data, both conspiratorial and
28 Since I follow Nieberding (2005) in his approach and do not want to purport that said derivation is my own
work, I use Nieberding’s variables and coefficients only changing the variable names from capital letters to small letters.
