Jump to Content Jump to Main Navigation
Economics for Competition Lawyers, 2nd Edition by Niels, Gunnar; Jenkins, Helen; Kavanagh, James (24th March 2016)

9 Quantification of Damages

Gunnar Niels, Helen Jenkins, James Kavanagh

From: Economics for Competition Lawyers (2nd Edition)

Dr Gunnar Niels, Dr Helen Jenkins, James Kavanagh

From: Oxford Competition Law (http://oxcat.ouplaw.com). (c) Oxford University Press, 2015. All Rights Reserved. Subscriber: null; date: 21 January 2019

Subject(s):
Cartels — Abuse of dominant position — Economics — Private actions — Quantifying harm

(p. 407) Quantification of Damages

9.1  Damages Claims, Economics, and the Law

9.1.1  What is special about the economics in damages claims?

9.01  You have seen in this book how economics helps to understand the effects of business practices, and how this in turn can be used in determining whether competition law has been infringed. The use of economics does not need to stop there, however. In Chapter 8 we showed the economic principles on which competition authorities may base remedies. In this chapter we deal with the economics of quantifying damages. After a competition authority has found an infringement and imposed a remedy (often a fine), parties that have been harmed by the infringement may file a claim for damages against the infringer. Such ‘follow-on’ damages actions are increasingly common in many jurisdictions. In principle, the same economic analysis of effects that is used to assess the existence of an infringement can be used to determine damages. The economic principles discussed in Chapters 4 to 6 are therefore of direct relevance to damages claims as well. After all, the competition investigation assesses harm to competition and consumers, and this harm is in essence what represents the damage caused by the infringement.

9.02  Yet there are some differences. First, a significant proportion of damages claims relate to cartels. These are usually prohibited per se, such that competition authorities have little need for analysing economic effects. This means that much of the economic analysis of the actual harm caused by cartels is undertaken only in the context of follow-on damages claims. Second, it is commonly accepted that effects-based analyses in exclusionary conduct cases focus on harm to the competitive process, not individual competitors. In contrast, damages claims are filed by individual competitors or customers who believe that they have suffered harm. Nobody files a claim on behalf of the competitive process (several jurisdictions allow certain forms of class action, but that is still because the members of the class have been harmed individually). You can see that there is some potential risk here of re-opening the debate on ‘harm to competition versus harm to competitors’. In original actions—where a competitor brings the case directly before a court (also called stand-alone actions)—the infringement and harm must both be established, and the court must consider whether evidence of harm to the competitor also constitutes evidence of harm to competition. The distinction is clearer in follow-on damages cases—the competition authority has already found an infringement (presumably on the basis of harm to competition more widely), and now the competitor must show that it has actually suffered from that infringement. Third, damages claims normally go further in quantifying harm than the analysis undertaken (p. 409) at the infringement stage. This is because a court must ultimately determine an exact monetary amount for the damages award (if any), whereas a competition authority can limit itself to establishing that such harmful effects are significant without quantifying them precisely.

9.1.2  Policy principles behind damages claims

9.03  Damages actions brought before courts by harmed parties are a form of private enforcement of competition law. This complements public enforcement by competition authorities. In the United States, the majority of antitrust cases are private actions. Elsewhere, most cases tend to be taken on by competition authorities, but the importance of private actions has grown. As mentioned above, there is a distinction between original private actions—where infringement and damages must both be established by the court—and follow-on damages actions that are brought before a court after an infringement decision by a competition authority.

9.04  There are various reasons why policy-makers find private actions attractive. One is that they save taxpayers’ money. Competition authorities have limited budgets, and must prioritize their enforcement actions. As economists would say, there is scope for an efficient division of labour: business-to-business disputes about contracts and exclusionary conduct seem to lend themselves to being effectively dealt with through private actions (businesses will be willing to pay for litigation if the stakes are sufficiently high). In the United States, treble damages are awarded to parties harmed by antitrust infringements, providing an extra incentive to bring private actions. Competition authorities can then deal with the rest—in particular, cases involving end-consumers, and cartel cases, where private parties are less likely to initiate a lawsuit as they lack the investigative powers of competition authorities.

9.05  Another reason to encourage private damages actions is that they contribute to the deterrence of anti-competitive practices. Nowadays, hefty fines, and possible prison sentences for individuals, already constitute a strong deterrent. The prospect of having to pay damages on top of the fines may further dampen the enthusiasm to form cartels. A curious economic argument in this regard is that follow-on damages actions may actually be redundant from a deterrence perspective—the same deterrent effect can be achieved by raising the fine itself up to the level at which the damage would ultimately be determined. That might save a lot of litigation costs. However, it pre-supposes (not always realistically) that public enforcement is sufficient to catch all competition law infringements, and that competition authorities have as much information at the infringement and fining stages as would become available in a private damages action.

9.06  A further policy principle behind damages actions, and one that is embedded in EU law, is that of the right to full compensation. As stated in the 2014 Directive on antitrust damages actions:

  1. 1.  Member States shall ensure that any natural or legal person who has suffered harm caused by an infringement of competition law is able to claim and to obtain full compensation for that harm.

  2. 2.  Full compensation shall place a person who has suffered harm in the position in which that person would have been had the infringement of competition law not been committed. It shall therefore cover the right to compensation for actual loss and for loss of profit, plus the payment of interest.

  3. (p. 410) 3.  Full compensation under this Directive shall not lead to overcompensation, whether by means of punitive, multiple or other types of damages.1

9.07  We discuss in this chapter how the compensation principle can be made to work in practice. If followed to the letter it would imply a certain degree of precision in the determination of the harm—there should be neither under-compensation nor over-compensation. It would also mean that compensation should reach the victims of an infringement regardless of where they operate in the supply chain. If the cartel’s direct customers have passed on the cartel overcharge to their own respective customers, the latter should get the appropriate compensation. We show how economics can help with these legal principles.

9.1.3  Searching for the right answer, within practical and legal bounds

9.08  Any damages assessment needs to strike a balance between two objectives: first, finding the most accurate answer—the desire to determine the real damage value as closely as possible, which is how an economist would naturally seek to approach quantification problems; and second, using approaches that are clear and easy to apply and that fit within the existing legal frameworks. Calculating the exact damage arising from an infringement requires complete information about what would have happened in a parallel world where the infringement did not take place—the ‘but for’ or counterfactual situation. Determining the counterfactual is inherently difficult as such complete information cannot exist. Courts generally understand this. As one judgment by the English High Court put it:

It is common ground between the parties, and obviously right, that expert economic evidence is necessary in this case, in order to be able to establish the relevant economic factual background, and then to wander off into the realms of economic fantasy which is necessary in order to arrive at answers to questions of causation and loss, so far as liability is established …

[this requires] expertise in defining relevant parameters and operations of and within the health care industry, and then the application of the fruits of that exercise, and of factual material, in economic modelling to work out what would have happened had Reckitt Benckiser not done that which it may ultimately be found they should not have done.2

9.09  Hence, to assess the counterfactual you need to enter into the ‘realms of economic fantasy’ and carry out economic modelling (don’t think immediately of complicated equations; modelling can be fairly straightforward sometimes). All models are necessarily simplifications of the real world. They rely on assumptions, and can vary in the degree to which they take into account all factors that may influence the counterfactual. This variation is often driven by constraints on data, time or budgets. The fact that models simplify reality does not invalidate their use as evidence in court. As one US court stated: ‘The antitrust cases are legion which reiterate the proposition that, if the fact of damages is proven, the actual computation of damages may suffer from minor imperfections.’3

(p. 411) 9.10  Another US court similarly held that: ‘The vagaries of the marketplace usually deny us sure knowledge of what plaintiff’s situation would have been in the absence of the defendant’s antitrust violation.’4

9.11  The EU Directive on antitrust damages actions states that the difficulty of quantifying harm with precision should not make it impossible to make claims:

Member States shall ensure that neither the burden nor the standard of proof required for the quantification of harm renders the exercise of the right to damages practically impossible or excessively difficult. Member States shall ensure that the national courts are empowered, in accordance with national procedures, to estimate the amount of harm if it is established that a claimant suffered harm but it is practically impossible or excessively difficult precisely to quantify the harm suffered on the basis of the evidence available.5

9.12  Several EU Member States have rules on the degree of freedom for judges when determining damages in cases where precise quantification is difficult. Such rules often reflect the principles of equity, justice, and procedural efficiency. For example, in a follow-on damages claim regarding a car insurance cartel the Italian Supreme Court confirmed that when the exact harm is difficult to prove, the Italian courts can rely on Article 1226 of the Italian Civil Code and award an equitable amount of damages (ex aequo et bono).6 The Supreme Court considered the insurance case as ‘a textbook example’ of where the Italian courts should make use of such a power, due to the fact that it was difficult for the claimant to prove the precise value of the loss it had suffered (essentially the cartel overcharge).

9.1.4  The remainder of this chapter

9.13  The aim of this chapter is to take you through the economic principles and methods that are of relevance to quantifying damages. Section 9.2 sets out a conceptual framework for estimating the harm from hardcore cartels. Section 9.3 does this in relation to exclusionary practices. Section 9.4 presents a classification of methods and models that can be used for quantifying damages, and contains a general discussion on what you should look for in a model. The classification is taken from a report for the European Commission by Oxera et al. (2009), which formed the basis for the Commission’s 2013 ‘Practical guide on quantifying harm for damages’.7 The later sections explore each of the main approaches. Section 9.5 deals with cross-sectional comparisons; section 9.6 with time-series comparisons; section 9.7 with difference-in-differences comparisons; section 9.8 with financial-analysis-based approaches; and section 9.9 with market-structure-based approaches. Section 9.10 addresses the economics of pass-on in the context of the passing-on defence in damages actions. Section 9.11 concludes with an explanation of the principles behind interest and discounting, usually the final (but relatively unexplored) step in damages calculations.

(p. 412) 9.2  Harm from Hardcore Cartels: Conceptual Framework

9.2.1  The main effects illustrated

9.14  Hardcore cartels tend to result in higher prices. This holds not only for price-fixing cartels but also for other types. We discussed this in Chapter 5. Some hardcore cartels target quantities rather than prices because they find it easier to agree on and monitor output quotas. The OPEC petroleum cartel is a notorious example. Restrictions in output normally go hand in hand with increases in price (a simple economic principle: we saw in Chapters 1 and 2 that most demand curves slope downward). Likewise, customer-allocation and bid-rigging cartels give each member a degree of monopoly power over its allocated customers or bids, which creates scope for restricting output and increasing price.

9.15  Figure 9.1—which you will recognize from the charts in Chapter 2—shows the overcharge paid on all the units actually sold (rectangle A), and the corresponding reduction in volume (triangle B). Triangle C represents consumer surplus in this market (the difference between what consumers are willing to pay for each unit bought, and what they actually pay). While this chart is similar to the ones in Chapter 2, note that the counterfactual price here is not necessarily the same as in perfect competition. Markets are rarely perfectly competitive. Equally, the cartel price is not necessarily the same as the monopoly price. Not all cartels manage to set price at the profit-maximizing monopoly level.

Figure 9.1  Stylized illustration of the effects of a hardcore cartel

9.2.2  The cartel overcharge harm

9.16  The overcharge, A, is the quantity of actual unit sales by the cartel multiplied by the difference between the actual cartel price and the counterfactual price (i.e. the price that would have been charged in the absence of the cartel). It is convenient to express the overcharge as a percentage of the actual price of the cartel. If the cartel price is €125, and the counterfactual price is €100, the overcharge would be 20 per cent (€25 is 20 per cent of €125). The overcharge is sometimes expressed as a percentage of the counterfactual price (in this case 25 per cent). This is equally valid in theory, but it is important to be clear (p. 413) about which percentage calculation is used. Expressing the overcharge as a percentage of the actual price makes it easy (and intuitive) to calculate the total amount of overcharge by applying the percentage to the amount that the buyer actually paid for its purchases (often known as the value of commerce). For example, if the cartel sold 1 million units at a price of €125 each and at an overcharge percentage of 20 per cent, the total overcharge would be €25 million. If one specific claimant filed a successful damages action, and it could demonstrate that its total purchases from the cartel amounted to, say, €15 million over the relevant period, the amount it was overcharged would be 20 per cent of €15 million, so €3 million.

9.2.3  Volume effects

9.17  The lost-volume effect (represented by triangle B in Figure 9.1) is known in economic theory as a deadweight welfare loss; it represents an inefficiency to the economy as a whole, as we explained in Chapters 1 and 2. This deadweight loss is greatest if the counterfactual price is equal to the price under perfect competition, but also arises if some other form of competitive interaction would have taken place in the counterfactual, such as oligopoly. From an economic perspective, the deadweight loss is inefficient as the cartel does not serve those customers who would be willing to pay the price offered under more competitive conditions.

9.18  In practice, cartel damages actions are mostly brought by parties that were actual (direct or indirect) customers of the cartel during the infringement period. These actions normally focus on the overcharge harm. Damages from reduced volumes are more difficult to prove. How can you identify those potential customers who did not purchase at all during the infringement period but would have done so at the non-cartelized price? Take the private schools price-fixing case in the United Kingdom.8 Parents who could no longer afford to send their children to private schools at the inflated school fees may have been harmed as well as those who paid the higher fees and did send their children to those schools—but can they prove it? Harm from volume loss may also arise further downstream if direct purchasers of the cartel pass on the overcharge by raising prices to their own customers (we discuss this further in sections 9.3 and 9.10).

9.2.4  Dynamic cartel effects

9.19  In addition to the price and quantity effects illustrated in Figure 9.1, cartels can have longer-term effects on the structure and functioning of the market. The reduction in rivalry between firms can result in lower levels of innovation and a slowing of the rate at which improvements in efficiency are achieved, or at which inefficient firms exit the market. Higher cartel prices may also have a distortive effect in downstream markets—for example, if certain purchasers can no longer afford high input prices and downstream concentration consequently increases.

9.20  To an economist, all these longer-term effects are relevant for the damages estimation, since they may affect the counterfactual price. For example, the counterfactual price might have been even lower (and hence the overcharge even higher) if the market would have seen cost-reducing innovations in the absence of the cartel. However, such factors can be taken into (p. 414) account only in circumstances where estimating these effects is really feasible, and where it is legally possible to include them, since it may be difficult to demonstrate a causal link between the infringement and the alleged longer-term harm.

9.2.5  Umbrella pricing and other ‘remote’ cartel effects

9.21  A cartel at a particular stage of the supply chain can cause ripples along the whole chain, such that various parties may potentially be harmed. Yet legal principles such as causality, remoteness and foreseeability tend to put a limit on who can claim for damages in practice. Direct competitors and direct customers of the infringing party are relatively ‘close’ to the effects of the infringement. Suppliers to the infringing parties may also have suffered harm, since cartels usually result in lower levels of output. This means that fewer inputs are required, thus reducing the volumes sold by suppliers. Companies in connected markets can equally be affected—if a brick cartel has raised costs to the construction industry such that there is less construction activity overall, other suppliers to the industry are harmed as well. Or take the example of a hot dog stand outside the brick factory. The brick cartel leads to an output restriction, hence lower production in the factory, fewer workers buying hot dogs during lunch breaks, and the hot dog stand suffering harm. Economically, these are all forms of harm that would fall within the compensation principle. However, courts tend to draw the ‘remoteness line’ closer to where the infringement has taken place.9

9.22  One area where courts have accepted that damages may arise beyond the cartelized products themselves is in relation to the ‘umbrella effect’. This occurs where products or geographic areas outside the cartel agreement compete so closely with the cartelized products that it is likely that their prices have been raised as well, benefitting from the pricing ‘umbrella’ provided by the cartel. In a 2014 judgment concerning the lifts and escalators cartel, the ECJ determined that such umbrella claims are valid in principle:

Where a cartel manages to maintain artificially high prices for particular goods and certain conditions are met, relating, in particular, to the nature of the goods or to the size of the market covered by that cartel, it cannot be ruled out that a competing undertaking, outside the cartel in question, might choose to set the price of its offer at an amount higher than it would have chosen under normal conditions of competition, that is, in the absence of that cartel …

It follows that … a loss being suffered by the customer of an undertaking not party to a cartel, but benefiting from the economic conditions of umbrella pricing, because of an offer price higher than it would have been but for the existence of that cartel is one of the possible effects of the cartel, that the members thereof cannot disregard.10

9.23  To analyse the umbrella effect you first need to assess whether the other products or geographic areas in question do indeed compete closely with the cartel. The techniques used for market definition and determining the closeness of competition discussed in Chapters 2 and 7 are of relevance here. If they are close substitutes, you can analyse the overcharge for these other products or areas directly, using the methods discussed later. Where limited data is available, you can sometimes assume that the overcharge for the other products (p. 415) is a proportion of the cartel overcharge. This proportion will depend on the degree of competition between the two groups; the higher this degree, the closer the umbrella overcharge will be to the cartel overcharge. Note that the effect can also work the other way around: the cartel may pull up the price of the competing products (the umbrella effect), but competing products may also drag down the cartel price. The stronger the suppliers or products outside the cartel are relative to those inside, the more likely is this latter effect of reducing prices.

9.2.6  Presumptions on the existence of cartel overcharges

9.24  If a competition authority decision contains evidence that a hardcore cartel was operational for many years without breaking down, you might expect the cartel members to have charged higher prices than they would have in the absence of the cartel. There have been significant enforcement efforts against cartels by competition authorities worldwide in the last two decades, and substantial fines and other penalties have been imposed (see Chapters 5 and 8). Where hardcore cartels have nonetheless been active—meeting and exchanging information regularly, and using sophisticated methods to circumvent detection—it is not unreasonable to infer that the mere fact that the cartel members took such risks over a long time period indicates that they considered it worthwhile.

9.25  As discussed in Chapter 8, the decision as to whether a company or individual engages in cartel activity can be thought of as a rational profit calculation. In a simple model, the company would be expected to commit the infringement if the expected additional profit from being part of the cartel (i.e. above the profit earned when competing) is higher than the sum of (i) the expected fine if the cartel is detected (which itself depends on the expected probability of being detected), and (ii) the expected damages payout of a subsequent successful private action. According to this logic, and on the assumption that cartel participants act rationally, they would not be taking on the risk of being prosecuted if they did not expect to achieve significant extra profits from the cartel. This basic logic is sometimes (implicitly or explicitly) followed by courts. In a vitamins cartel case, the Dortmund Regional Court applied the presumption that a cartel price is generally higher than a market price:

The damage of a price cartel consists of the difference between the cartel price and the hypothetical competitive price in the absence of the cartel. According to the experience of life (Lebenserfahrung), it can be assumed that a competitive price is lower than a cartel price. The defendant did not show that it would have been different in this case and why. The difference between the competitive price and the cartel price represents a financial damage in the sense of lost wealth.11

9.26  The court found support for this proposition in the fact that prices increased or remained stable during the cartel, but declined after the cartel ceased to operate. Similarly, in a cement cartel case, the Higher Regional Court in Düsseldorf stated that: ‘The longer and more sustainable a cartel was operational, and the wider the area it was designed to cover, the higher the requirements that have to be imposed on a court if it wants to deny that the cartel agreement produced any economic benefits.’12

(p. 416) 9.27  The court emphasized that market mechanisms were unlikely to function properly due to the imposition of cartel quotas. It thus concluded that prices set by the cartel were likely to have been higher than in a competitive market. In a damages action against a sugar cartel in Spain, the Valladolid Provincial Court relied on the competition authority’s finding that the agreement on prices had caused serious harm, and therefore rejected outright the defendant’s expert report which had calculated the damage as zero.13

9.28  The EU Directive on antitrust damages actions captures the above logic by establishing a rebuttable presumption of harm from hardcore cartels, aimed at shifting the burden of proof towards the defendants to some extent:

To remedy the information asymmetry and some of the difficulties associated with quantifying harm in competition law cases, and to ensure the effectiveness of claims for damages, it is appropriate to presume that cartel infringements result in harm, in particular via an effect on prices. Depending on the facts of the case, cartels result in a rise in prices, or prevent a lowering of prices which would otherwise have occurred but for the cartel. This presumption should not cover the concrete amount of harm. Infringers should be allowed to rebut the presumption. It is appropriate to limit this rebuttable presumption to cartels, given their secret nature, which increases the information asymmetry and makes it more difficult for claimants to obtain the evidence necessary to prove the harm.14

9.2.7  Empirical insights into the magnitude of cartel overcharges

9.29  Economists have carried out many empirical studies on overcharges in past cartels. Lawyers involved in damages actions have shown great interest in the results of these studies, in order to get a feel for what sort of orders of magnitude are typically involved in such cases. We note from the outset that some care is required when interpreting this empirical data. Not all studies on cartel overcharges would qualify as sufficiently robust. There may also be a publication bias: empirical studies tend to focus on cartels that are most likely to have had an effect, and some cartels with no effect will not have been captured in these studies (although, as shown below, a small but significant proportion of the cartels studied resulted in no overcharges).

9.30  A study by Connor and Lande (2008) used a comprehensive dataset on cartel overcharges, and has been the most widely cited study on this topic.15 It contained 674 observations of average overcharges from 200 social science studies of cartels from the eighteenth century onwards—for example, it covered a British coal cartel that started in the 1770s and a Canadian petroleum lamp oil cartel in the 1870s. The study found that the median cartel overcharge for all types of cartel was 20 per cent. An earlier study by Connor and Lande (2005) suggested that in around 7 per cent of cartel cases there was no overcharge. Oxera et al. (2009) examined the dataset underlying the 2008 Connor and Lande study, as well as an additional 350 observations provided by these authors, and tested the sensitivity of the results by limiting the sample to cartels that started after 1960 and to overcharge (p. 417) estimates obtained from peer-reviewed academic articles and chapters in published books (this reduced the sample size from over 1,000 to 114).

9.31  Figure 9.2 illustrates the distribution of cartel overcharges across this dataset. The range with the greatest number of observations is 10–20 per cent. The median overcharge is 18 per cent—not far from the 20 per cent found by Connor and Lande. The average overcharge is around 20 per cent, compared with 23 per cent in Connor and Lande. However, since the variation in observed overcharges is large, it is informative to consider the distribution of overcharges as well as the median or average overcharge. In 93 per cent of the cases the overcharge is above zero (as in Connor and Lande, 2005). This supports the theory that in most cases the cartel overcharge may be expected to be positive, but also shows that there is a small but significant proportion of cartels where no overcharge has been found.

Figure 9.2  Distribution of cartel overcharges in empirical studies of past cartels

Source: Oxera et al. (2009), based on Connor and Lande (2008).

9.32  The important policy question is what you do with the results of this empirical literature. Is it merely interesting background information? Do you use it to inform your calculation of the possible order of magnitude of the damage at an early stage in the case? Or do you go as far as using a presumption that the overcharge in the case you are considering is somewhere between 10 per cent and 20 per cent, as that is the most frequently observed range of past cartel overcharges? A prominent example of the latter approach is the Hungarian Competition Act, which establishes a rebuttable presumption that hardcore cartels result in a 10 per cent overcharge.16 We see little economic merit in such presumptions (other than possibly the presumption that for hardcore cartels the overcharge is likely to be greater than zero, without specifying a number, as discussed above and reflected in the EU Directive on antitrust damages actions). The amount of the overcharge in any particular damages case ultimately (p. 418) needs to be determined on the specific facts. This was also the view of the Commercial Court in Brussels in the damages action brought by the European Commission (as a purchaser) against the members of the lifts and escalators cartel. In a 2014 judgment the court accepted the notion that cartels are generally likely to increase prices, but quoted the Oxera et al. (2009) study that for a small but significant proportion of cartels no overcharge has been found and hence that a case-by-case assessment is required.17

9.3  Harm from Exclusionary Conduct: Conceptual Framework

9.3.1  Lost profit: Legal principles determining the relevant economic questions

9.33  Exclusionary conduct prevents existing rivals from competing effectively or forces them to exit altogether. Potential competitors may be prevented from entering the market or restricted to small-scale entry. Buyers, be they end-consumers or intermediate producers or distributors, may be harmed by exclusionary conduct if the reduction in competition leads to higher prices, less choice, or lower quality. A number of legal principles are important when considering the chances of success for certain types of claim for exclusionary damages. To economists there is a broad set of damages that may arise from a competition infringement, and each type is in principle quantifiable; to lawyers the set is narrower. Existing rivals are, in legal terms, relatively close to the exclusionary conduct, meaning that it is easier for them to substantiate a damages claim than it is for potential entrants.

9.34  From an economic perspective, harm to competitors from an exclusionary infringement may arise in two ways. First, it may be in the form of increased costs, where costs include both cash cost items, such as input goods, and other more general items, such as the cost of financing the business. Second, it may be in the form of reduced revenue, where the infringing conduct affects the price or sales volume. The effect of both increased costs and reduced revenue is a reduction in profit. From a legal perspective, it may be important (we understand from lawyers) to make explicit whether this effect falls under actual loss (damnum emergens) or lost profit (lucrum cessans). To an economist there is not much difference. The framework presented here can be used for both categories. In legal terms the evidentiary requirements may be different. Actual loss is generally easier to prove—for example, a small airline that was foreclosed from access to distribution channels by a dominant rival and had to find alternative, more expensive, distribution channels instead, should in principle be able to demonstrate the harm based on the actual additional expenditure incurred (it can show the actual invoices). It is more difficult to prove the quantum of the alleged lost profit—e.g. loss of market share or a fall in sales growth—and a causal link between the unlawful conduct and that lost profit. This is because of difficulties in establishing whether such losses were due to the anti-competitive practice or to other factors, such as incompetence, bad luck, or external economic factors. Most legal systems seem to take a relatively conservative approach when assessing claims for lost profit—or related concepts, such as loss of chance and loss of opportunity—in exclusionary damages cases. Often the damages awarded are limited to actual losses or a narrow interpretation of lost profits. We show some examples below.

(p. 419) 9.3.2  Lost profit: Cases where courts have been cautious

9.35  The Paris Court of Appeal ruled on an exclusionary damages case in 1998 after it had found that Labinal, a supplier to the aerospace industry, had infringed Article 101 and Article 102 by trying to eliminate its only competitor, Mors, from a tender to supply tyre pressure measuring equipment to British Aerospace.18 The court awarded Mors FF34.2 million (around €5.2 million) for the losses caused by Labinal. The calculation was based on the report of a court-appointed expert, with the court confining itself to assessing whether the expert’s conclusions were reasonable and supported by the factual evidence. The expert considered that Mors had incurred harm in the form of: (i) additional administrative and commercial costs; (ii) loss of opportunity to participate in other tenders; and (iii) the inability to recover one-off costs. However, the expert did not consider that Mors should be awarded damages for loss of opportunity to enter adjacent markets since it had failed to prove that it would have entered these other markets had Labinal’s anti-competitive conduct not taken place.

9.36  An exclusionary damages case before the Court of Appeal of Milan followed a finding by the Italian competition authority that the members of the National Association of Employment Consultants had collectively boycotted the claimant’s software packages.19 The court compared the average number of contracts with the claimant that were terminated by the association’s members in the two years of the collective boycott (1997–98) with the average number of contracts terminated in the years before the boycott. On that basis, the court awarded €148,200 in damages. The next question was whether the claimant was entitled to compensation for the slower growth of its business due to the boycott (a form of lost profit). While the claimant had shown that before the boycott its business was growing by more than 10 per cent per year and that this increase had suddenly ceased at the time of the boycott, the court considered that it could not be sure that this growth would have continued in the counterfactual. The past evidence could not be used to support a presumption that the growth rate would have been the same.

9.37  Various English courts, and at one stage the ECJ, ruled on the famous Crehan damages case.20 The claimant was a pub landlord who, in 1991, entered into an exclusive contract with Inntrepreneur to lease two pubs on the condition that he stocked only its beers. After two unsuccessful years, Inntrepreneur terminated his tenancy and sought to recover money owed to it. Mr Crehan counterclaimed that the beer tie agreement infringed Article 101 and sought to recover three heads of damages: (i) losses that he suffered during the period of the lease between 1991 and 1993; (ii) future profits he would have made in the period between 1993 and 2003 in the absence of the beer tie; and (iii) the value in 2003 of the untied leases had he wished to sell these on. Both the High Court and the Court of Appeal accepted that if liability were established, the claimant would have been entitled to recover in full the losses suffered during the period of the two-year lease. However, the Court of Appeal took a more restrictive approach than the High Court in relation to the recoverability of future profits that the claimant would have made between 1993 and 2003. The High Court (albeit hypothetically as it had dismissed the case on other grounds) calculated the total lost profit (p. 420) as £1,311,500. This included all three heads of damages listed above. In contrast, the Court of Appeal held that the claimant would have been entitled to only £131,336 in damages. It considered the lost profit between 1993 and 2003 to be too speculative. The case was subsequently appealed to the House of Lords, which overturned the Court of Appeal’s finding that the beer tie agreement infringed Article 101. As such, the issue of the quantum of damage did not need to be addressed in this last ruling. Mr Crehan’s thirteen years of battles in court ultimately left him empty-handed, but his case was instrumental in embedding in EU law the legal principle of compensation.

9.3.3  The concept of lost profit: An economic framework

9.38  The basic economic framework to determine harm from exclusionary practices—encompassing both actual loss and lost profit in the legal sense—is illustrated in Figure 9.3. The damages are calculated as the difference between the actual and the counterfactual profit of the company. To take a simple example, if the victim of the infringement—say, a small competitor airline that was harmed by a dominant airline’s exclusionary conduct—had actual revenues of €10 million and actual costs of €8 million, its actual profit is €2 million. If its revenues would have been €15 million in the absence of the infringement and its counterfactual costs €12 million, then its counterfactual profit is €3 million. The small airline’s lost profit is therefore €1 million.

9.39  The framework can be rearranged as illustrated in Figure 9.4, which shows a simpler expression for the fall in profit. The lost revenue in Figure 9.4 is calculated as the difference (p. 421) between the counterfactual and actual revenues. The costs avoided due to the infringement are then deducted from the lost revenue to obtain the reduction in profit. For example, if volume falls, certain costs that vary with volume (e.g. fuel or input materials) will decrease as well. A company that experiences a reduction in sales due to exclusionary conduct by a rival will in this sense have an offsetting benefit from a cost reduction, and this cost saving should be deducted from the lost revenue to obtain the lost profit, as illustrated in Figure 9.4. This rearranged expression has the advantage over Figure 9.3 of requiring less detailed knowledge of the company’s cost structure. It is not necessary to calculate all the costs that the company would have incurred in the relevant period. Instead, the focus is on the costs that the company did not incur because of the infringement—that is, the avoided costs. Following the above example, lost revenues for the small airline equal €5 million (€15 million of revenue in the counterfactual minus €10 million of actual revenue). Its avoided costs amount to €4 million (€12 million of costs in the counterfactual minus €8 million of actual costs), reflecting the fact that in the absence of the exclusionary conduct the airline would have achieved higher revenues but also incurred additional costs. Lost revenues of €5 million minus avoided costs of €4 million gives a lost profit of €1 million.

Figure 9.3  Economic framework for calculating harm from exclusionary conduct

Source: Oxera et al. (2009).

Figure 9.4  Rearranged economic framework for calculating harm from exclusionary conduct (equivalent to Fig. 9.3)

Source: Oxera et al. (2009).

9.40  The effect on profits is often approximated by reference to variables such as lost volumes, lost customers, or lost market share. These quantifications are (or should be) consistent with the conceptual framework in the figures above. For example, damages claims based on an estimation of lost sales volume can be translated into a negative effect on profits by applying some average counterfactual profit margin to each unit of sales lost. If the airline demonstrates that it sold 50,000 fewer tickets because of the abuse by its dominant rival, and its average profit margin per ticket was €20 (and would have been €20 in the counterfactual), its lost profit can be estimated at €1 million.

9.3.4  The effect of infringements that increase input prices

9.41  Figure 9.3 can also be rearranged to capture the effect on profit from infringements that increase input prices. See Figure 9.5. This applies to exclusionary conduct that has the effect of raising prices to buyers, and to exclusionary conduct that has the effect of raising rivals’ costs in downstream markets—a common theory of harm in abuse of dominance cases. Figure 9.5 can also be used to analyse the effect on the profits of customers of a cartel. In the top section of Figure 9.5, the fall in profit is calculated as the increase in costs minus the increase in revenues. In the case of a cartel or anti-competitive increase in an input price, the increase in costs to the downstream purchaser (the box to the left) will often be equal to the overcharge. The increase in revenues (the middle box) will include the pass-on of this overcharge to the purchaser’s own customers. Pass-on is achieved through raising price, and the resulting higher revenues may have to be offset against the higher costs caused by the infringement. The lower part of Figure 9.5 is equivalent but splits the profit effect from a higher price into three components: the increase in costs on units actually purchased (the overcharge effect); the (p. 422) increase in revenues on units actually sold downstream (which covers the pass-on effect); and the effect of lost volumes of sales downstream due to the price increase upstream. Within this framework, the total lost profit equals the sum of the lost profit from actual volumes and the lost profit from lost volumes.

Figure 9.5  Rearranged economic framework for calculating harm from price-increasing conduct (equivalent to Figure 9.3)

Source: Oxera et al. (2009).

9.3.5  The counterfactual in exclusion cases: Competitive or barely legal conduct?

9.42  Where damages are claimed for exclusionary conduct, the counterfactual is not always straightforward to determine, as the dominant company may have behaved in a number of different ways in this hypothetical situation. In predatory pricing cases, which of a range of possible non-predatory prices would the dominant company have charged in the counterfactual? In margin squeeze cases, what would have been the dominant company’s non-abusive margin between the wholesale price and the retail price, and would it have achieved this margin in the counterfactual by lowering the former or raising the latter?

9.43  This question is not always relevant to the legal assessment. Take the 2 Travel v Cardiff City Transport damages action before the CAT.21 Following the entry by 2 Travel with a low-cost bus service on a number of routes in Cardiff, the incumbent operator, Cardiff Bus, launched its ‘white services’ on the same routes to compete with 2 Travel. The OFT considered this to be predatory. In the follow-on damages claim, the counterfactual question was how much profit 2 Travel would have made had Cardiff Bus not launched the white services at all. The question was not whether Cardiff Bus might have responded in some other, legitimate, way to 2 Travel’s entry, for example, by increasing service frequency or lowering fares on existing services but still recovering costs.

9.44  However, in other cases this counterfactual question can make a material difference. In a predatory pricing case, should it be assumed that the dominant company would have charged a price that covered its average total cost in full, leaving plenty of space for competitors? Or might it still have priced sharply, just staying on the right side of the law, covering average avoidable costs? This matters for the damages estimate. The higher the counterfactual price, the more profit margin there would have been available for the claimant. From an economic perspective, it depends on what is most realistic: when an incumbent faces new entry, it is not unreasonable to assume that it reacts to this entry somehow, even in a counterfactual situation where such a reaction remains within the bounds of the law. The price it charged before the entry is therefore not necessarily an accurate representation of the counterfactual price; its costs, or prices charged in other markets with existing competitors, may provide a better indication. Economics provides tools to estimate such costs and prices in order to determine the counterfactual, as further discussed in this chapter.

(p. 423) 9.45  The CAT faced this question in Albion Water, a follow-on damages claim against Dŵr Cymru (Welsh Water) involving an excessive wholesale access charge and resulting margin squeeze.22 For both conceptual and practical reasons, the CAT dismissed the notion that the counterfactual price should be set at the level of access charge that would just be on the right side of the law (in this case, the highest legitimate access charge). Instead it chose a value based on an average of a range of lawful charges:

It will be very rare that an infringement decision, whether adopted by a domestic competition authority or by the European Commission, or indeed on appeal as in Case 1046 [where earlier the CAT had confirmed the abuse], will determine the precise borderline between lawful and unlawful conduct. If Dŵr Cymru is right that the claimant in a follow-on damages claim will have to show precisely where that line should be drawn, that will often involve the court in re-doing much of the work done in the earlier infringement decision. Further, it is a task that is almost impossible to accomplish, as is demonstrated in this case. If 16.5p/m3 is not abusive (a point we do not decide), what about 16.6p/m3 or 16.7p/m3, or 16.8p/m3? We do not see how a claimant could prove that one rather than the other is the tipping point between lawful and unlawful conduct.23

There is a range of lawful access prices that Dŵr Cymru could have offered and we should take the figure in the middle of that range. The counterfactual must be based on an assumption that Dŵr Cymru would have offered a reasonable access price, rather than an access price which is the highest it could lawfully have charged.24

9.46  In margin squeeze cases there is a further conceptual question: in the counterfactual, is the wholesale price lower, or the retail price higher? After all, this type of abuse involves squeezing the competitor’s margin from both sides. On which side the squeeze is released can have a significant impact on the estimated damages. If the incumbent had charged a higher retail price in the counterfactual, there would have been more scope for the entrant to gain market share and grow overall demand downstream by undercutting this retail price. In contrast, if the incumbent had cut wholesale prices but kept the retail price at the same low level in the counterfactual, the entrant would have found it more difficult to grow its profits. Again, from an economic perspective it depends on what is most realistic. In telecoms markets, where several margin squeeze cases have arisen (see Chapter 4), retail markets tend to be more competitive than wholesale markets. Incumbents, acting lawfully, may be more inclined to lower wholesale prices than to raise retail prices so as to avoid losing retail market share.

9.4  A Classification of Methods and Models for Quantifying Damages

9.4.1  The classification

9.47  Having set out the conceptual frameworks for determining harm from cartels and exclusionary conduct, we now turn to the practical approaches to quantifying such harm. Economics has developed a wide array of methods and models for this purpose. Figure 9.6 presents a (p. 424) classification into three broad approaches: comparator-based, financial-analysis-based, and market-structure-based. This is taken from the report for the European Commission by Oxera et al. (2009). It encapsulates previous classifications, most notably that of US case law which has explicitly identified three ‘common approaches to measuring antitrust damages’: the before-and-after approach, the yardstick or benchmark approach, and regression analysis.25 The classification here draws clearer distinctions between what is used as the basis for the counterfactual in each method, and the precise estimation technique. Before-and-after and yardstick are in reality two different types of comparator-based approach; the former involves making comparisons over time, the latter across product or geographic markets. Similarly, the regression analysis category does not in itself clarify the basis for the counterfactual. Regression analysis is a technique, and can be used in both the before-and-after and the yardstick approach. Figure 9.6 identifies the basis for the counterfactual that underlies each of the three approaches. It then summarizes the estimation techniques that can be used within each approach. All three approaches can be used for any type of competition law damages case. They are not mutually exclusive and in fact often complement each other, as discussed below.

Figure 9.6  Classification of methods and models for quantifying damages

Source: Oxera et al. (2009).

9.48  Comparator-based approaches use data from sources that are external to the infringement to estimate the counterfactual. This can be done in three ways: by cross-sectional comparisons (comparing different geographic or product markets); time-series comparisons (analysing prices before, during, and/or after an infringement); and combining the above two in ‘difference-in-differences’ models (e.g. analysing the change in price for a cartelized market over time, and comparing that against the change in price in a non-cartelized market over the same time frame). Various techniques can be used to analyse this comparator (p. 425) data, ranging from the simple, such as comparing averages, to the more sophisticated, such as panel data regression.

9.49  Financial-analysis-based approaches—the Commission’s Practical guide calls them ‘cost-based and finance-based methods’—have been developed in finance theory and practice. They use financial information on defendants, claimants or comparator companies to estimate the counterfactual. There are two types of approach that use this information. First are those that examine financial performance. These include assessing the profitability of defendants or claimants and comparing this against a benchmark, and bottom-up costing to estimate a counterfactual price. The second type is a group of more general financial tools, such as discounting and valuation, which can be used also as part of the other approaches (e.g. a time-series comparison of profit margins during and after the cartel).

9.50  Market-structure-based approaches are derived from IO theory and use a combination of theoretical models and empirical estimation (rather than comparisons across markets or over time) to arrive at the counterfactual. This involves identifying models of competition that best fit the market in question, and using these models to assess what prices or volumes would have been in the absence of the anti-competitive conduct. The Practical guide calls these ‘simulation models’. We think this term is too general. While it captures the idea that you use a model to simulate a counterfactual, simulation is a technique that can be used in many other contexts as well.

9.4.2  Some notes on the use of models generally

9.51  No model can fully describe and predict the complete range of market interactions, but nor is it intended to do so. Models can be thought of as maps that simplify the real world to make it understandable and predictable. The simplifications made will depend on the intended use of the map. A geological map and a road atlas will make very different simplifications, even if describing the same piece of land, because they are used for different purposes.

9.52  A model will only be as good as the quality of the input data used to populate it. Thus it is important to ensure that the sourcing of data is free from potential biases and that the data used is consistent over time and over units (companies, business units, or individuals). A sophisticated model based on unreliable or biased data is less useful than a simpler model based on better data. A critical question for you, or a court, to ask when presented with a model is therefore whether the data used is of sufficient quality, and whether a simpler model could be used. There are many statistical tests to assess the reliability of the model and its results.26 When presented with an empirical study, you are entitled to require it to show the basic diagnostic tests. One important aspect of reliability is statistical significance: does the estimated value (say, of the overcharge) reflect the true value? A 90 per cent or 95 per cent confidence interval is often used as a threshold for this purpose—the range of uncertainty around the estimated value is such that there is only a 10 per cent or 5 per cent probability that the true value does not lie within this range. The t-test and p-value are most commonly used for this purpose. If the p-value is 0.05 or smaller, the true value lies within the 95 per cent confidence interval around the estimated value, and the estimate can be considered statistically significant.

(p. 426) 9.4.3  What can economic models say about causation?

9.53  Quantifying the harm and establishing a causal link between the harm and the infringement are essential parts of any damages action. What, if anything, can economic models used for quantifying damages say about causation? Econometric analysis seeks to identify statistically significant relationships between a dependent variable and various explanatory variables (see also Chapter 2). The econometric analysis itself does not prove causality. A model may be constructed with two completely unrelated variables that happen to have a high correlation (i.e. they move similarly over time, like inflation and accumulated rainfall). Such an analysis may identify a relationship that is statistically significant but economically meaningless.

9.54  Nevertheless, econometrics can help address the issue of causation because it can take into account many possible explanatory factors (subject to data availability). This is important for damages actions since the difficulties in proving causation frequently arise when a model purports to show a relationship between two variables but ignores other explanatory variables. A model may show that a competitor’s sales have fallen during the period of an exclusionary abuse, but fail to address other possible explanations such as a general drop in sales in the market, the entry of a new competitor, or managerial incompetence. A good econometric model would seek to ‘control’ for those other explanations—that is, incorporate them into the model as additional explanatory variables. That way, the various effects can be isolated from one another, and the model may well show that, while the other factors explain some of the sales loss, the remainder of the loss is related to the infringement.

9.55  In US antitrust damages cases, where the use of econometrics is more common than in Europe, the issue of causation is often dealt with in this way. The US courts sometimes expect to see a regression analysis in order to have robust estimates and to isolate the effects of the infringement from other effects:

The goal of a prudent economist in performing the ‘before and after’ analysis is to determine the hypothetical or ‘counter-factual’ prices that would have prevailed during the conspiracy period, but for the conspiracy. In applying the ‘before and after’ model of damages, it is fundamentally necessary to explain the pattern of forces outside the violation period using factors that might have changed (i.e., supply, demand, and differences in competition) to predict the prices during the conspiratorial period. In this context, as in most economic problems, failure to keep ‘other things equal’ is one of the known pitfalls in the path of the serious economist … A prudent economist must account for these differences and would perform a minimum regression analysis if utilizing the ‘before and after’ model.27

9.56  In a number of cases, economic evidence has been rejected on the grounds that it did not sufficiently account for other possible explanations for the harm. In Stelwagon v Tarmac Roofing, the expert’s opinion was not accepted since it ‘failed to sufficiently link any decline in Stelwagon’s MAPs [modified asphalt products] sales to price discrimination. The sales may have been lost for reasons apart from price discrimination—reasons that [the expert]’s analysis apparently did not take into account.’28 In other cases courts did accept the econometric evidence. In Conwood v US Tobacco, a case of alleged monopolization, the court accepted that the expert had tested for ‘all plausible explanations’ for the claimant’s low (p. 427) market share in his regression analysis, thus allowing the court to isolate the effect of the infringement itself.29

9.4.4  Choosing a single damages value

9.57  Methods and models cannot be ranked a priori. The choice of model is driven mainly by data availability. It is not uncommon for an economic expert to use more than one model if there are different sources of available information. It is also common to see the experts on the claimant and defendant sides using different models. Ultimately a court needs to decide on the specific amount of damages (if any) to be awarded. The main question is normally whether specific models have been applied reasonably and robustly to the case at hand, not whether one approach is inherently superior to another. It may be that several models are robust but give non-identical results, simply because they rely on different assumptions and have used different data. A court may have to decide which assumptions or data sources it prefers.

9.58  An alternative approach when presented with multiple robust estimates is to ‘pool’ them. This involves combining the results of two or more models into a single value. The economics literature has shown pooling to work well.30 The combination of individual forecasts of the same event has often been found to outperform the individual forecasts. One approach is simply taking the mean of the available estimates. This is often done for macroeconomic forecasts. Hence, if three robust models estimate the harm to be €10.1 million, €11.2 million, and €12.0 million, respectively, the pooled model result using a simple average would be €11.1 million. This combined value can then be used as the best estimate of the actual harm.

9.59  An advantage of pooling is that when the models rely on different data, combining their results means that the final value reflects more of the underlying data (and hence more of the available information) than a single model alone. While it is theoretically possible to conceive of a ‘unified’ model that incorporates all the data sources of the individual models, it is often difficult to implement this in practice. Instead, pooling of different model results creates a form of ‘unified’ estimate, since it draws on all the approaches undertaken. In addition, pooling the results can help reduce biases in the individual models, as positive and negative biases may offset one another. If the biases are all in the same direction, combining results would not eliminate them, but nor would it exacerbate them. Thus, combining results can generally be expected to reduce biases, since at least some of the biases are likely to be in different directions. Pooling does need to be applied with care. It is most frequently used in cases where a single expert is using multiple approaches, or where multiple experts—such as a group of court-appointed experts—are attempting to estimate the same value for the same purpose. Pooling the results from different experts on opposing sides can also work, but only if their approaches start from similar premises and datasets.

9.60  The Court of Session in Edinburgh accepted the pooling approach in a commercial dispute between two paper manufacturers that led one to claim for lost profits.31 Inveresk had sold one of its brands to Tullis, but in the five-month transition period its staff hindered (p. 428) operations and thereby damaged the brand. Tullis sued for lost profits as disaffected customers switched to other suppliers. The claimant’s expert used various methods to estimate the lost sales from the existing customer portfolio: a before-and-after comparison of sales using simple interpolation; several before-and-after comparisons using econometric analysis; and a difference-in-differences analysis, using a comparison with sales to an unaffected group of customers before and after the infringement. The judgment, which, perhaps as yet unusually in Europe, contains a lengthy discussion of the econometric analysis, agreed with the expert that pooling of the three main model results was appropriate to come to a final damages value: ‘It is accepted economic practice to use more than one benchmark, in order to reflect more of the underlying data and to reduce the effect of biases in individual approaches.’32

9.61  In the end the court awarded £4,250,000 in damages in line with the pooling of the estimates made by the claimant’s expert.

9.5  Comparator-based Approaches: Cross-sectional

9.5.1  Choice of comparator

9.62  Cross-section comparisons can be made between companies, product markets, geographic areas, or a combination. In Conwood v US Tobacco, a monopolization case in moist snuff (dipping tobacco) resulting in one of the highest antitrust damages ever awarded ($1.05 billion), the plaintiff’s expert used comparator market shares from US states where no exclusionary practices had taken place, and from the market for loose-leaf tobacco, in which the defendant was not active.33 The regression analysis showed that the plaintiff had a higher sales growth in these comparator markets. The ideal cross-sectional comparison includes data from the relevant market and from unaffected markets that are otherwise similar. If a regional infringement had the effect of increasing prices nationally, comparing data from two regional markets within the country would give a biased estimate of the damage since the comparator groups would be ‘contaminated’ by the infringement. In a case relating to a German paper wholesaling cartel, both the higher regional court and the German Federal Court of Justice felt unable to use cross-sectional comparisons between cartelized and other regional markets for paper wholesaling to estimate the overcharge because there was some evidence of cartels existing in other regional markets, which were therefore possibly affected by overcharges as well.34

9.63  The strength of a cross-sectional comparison lies in how like-for-like the comparison is and how many of the differentiating factors have been controlled for in the modelling. In a US case concerning a refusal to deal with a prospective purchaser of the Chicago Bulls basketball franchise by the Chicago stadium owner, the experts and court calculated the counterfactual fair market value of the Chicago Bulls franchise using recent sales prices of comparable National Basketball Association franchises.35 A total of ten such transactions were considered by the court to be sufficiently comparable. Factors that were considered in (p. 429) the comparison between franchises included the size and population growth of the home city; the city’s interest in basketball (e.g. its history of supporting teams, stadium attendance, and advertising support—all factors that the court said are difficult to quantify objectively and ultimately gave little weight); and whether the deal concerned an ‘expansion’ franchise (i.e. a new team, which is normally less valuable than an existing team). Likewise, in the Conduit case in Spain involving exclusionary practices by the telephony incumbent in relation to directory enquiries, the court accepted the UK market as a comparator for the claimant’s lost market share in Spain, given the similarities between the two markets.36 The market for directory inquiries had recently been liberalized and Conduit was one of the new entrants in both countries, but with greater success in the United Kingdom. The damage estimate itself was not accepted by the court.

9.5.2  Comparison of averages

9.64  Several estimation techniques can be employed to derive the counterfactual price using cross-sectional comparators, ranging from the simple to the more sophisticated. A relatively simple comparison of averages uses the average price in an unaffected comparator group as an estimate for the counterfactual price. See Figure 9.7. The price in the cartelized market is €12, while the average price in comparator markets is €10. This indicates that the overcharge is €2 (or 16.7 per cent of the cartel price). The measure used to denote the ‘average price’ is usually the simple average or arithmetic mean (as the €10 here), but could also be the median or the modal price. The arithmetic mean is calculated by dividing the sum of all observations by the number of observations (€10 is the simple average of €10, €10.25, €9.50, €9.50, and €10.75). The median price is identified such that half of the companies in the comparator group charge a price below this median (p. 430) price and half charge above it (here the €10 in comparator market 1). The modal price is one that is observed most frequently in the comparator group (here €9.50 in markets 3 and 4). The choice among the three measures would depend on the nature of the market and the pricing pattern. If there are ten comparator markets, nine with a price of €10 and one with a price of €25, the modal or median price—both €10—might give a more accurate representation of the average price than the arithmetic mean of €11.50 which is influenced by the outlier (you may of course also conclude that this outlier is not such a good comparator market after all).

Figure 9.7  Example of a cross-sectional comparison of averages

9.65  Whichever metric is used, the counterfactual price can then be compared with the actual price charged in the infringement market to calculate the overcharge. If there is sufficient data, a statistical test can be undertaken to check whether the counterfactual price is significantly different (in the statistical sense) from the actual price charged. We mentioned the t-test and the p-value earlier. Testing for statistical significance—and making the results of the tests transparent—is good practice in economics and statistics. It helps in understanding the precision of an estimate. A test for statistical significance accounts for the number and the variation of observations in the comparator group. The more observations you have, and the lower the variation, the more likely you are to obtain significant results.

9.66  What about Figure 9.7, where the actual price was €12 and the simple average of the comparator markets was €10? The variation in prices in the comparator group is relatively small—between €9.50 and €10.75—and quite different from the cartel price of €12, but you have only five comparators. The p-value for this comparison is 0.026, which is below the threshold of 0.05, and means that the actual difference lies within the 95 per cent confidence interval around the estimated difference. In practice, where both the comparator and cartel market observations are more varied, and some comparator prices even exceed the cartel market price, you would normally need a greater number of data points to obtain estimates that are statistically significant at standard thresholds.

9.5.3  Cross-section regression analysis

9.67  Regression analysis is a more sophisticated statistical method that can explain the variation in data using multiple factors (we explained the basics in Chapter 2). It addresses the main shortcomings of simple comparisons of averages by controlling for differences in market or company characteristics. Consider a situation in which you have prices for a number of companies (n), some of which are in the infringement market and some of which are outside it. A cross-section regression analysis will then be based on an equation such as Pi = α‎ + βXi + δDi + ei. On the left-hand side of the equation is the variable to be explained, in this case price. Thus the variable Pi represents the price of company i (this i ranges from 1 to n as there are n companies in total). On the right-hand side are all the variables that can explain price. Xi includes characteristics of each company or its market, such as input costs, quality, or size—these are factors other than the infringement that may influence price, and hence should be controlled for. Di is the ‘dummy’ variable which is equal to one if company i belongs to the infringement market, and zero if the company belongs to a comparator market. This dummy variable is ultimately of main interest to the analysis as it gives you the estimate of the overcharge. The first term α‎ is the intercept, and the last term ei is the error term (we explained these in Chapter 2; they are important for statistical purposes but not directly for inferring the overcharge).

(p. 431) 9.68  The model can be estimated when sufficient data points are available for all the P and X variables. The D values are taken directly from what is known about the scope of the infringement market. The regression analysis seeks to identify the statistical relationship between these variables. The parameter of main interest in this regression is δ‎, which represents the average change in price attributable to the fact that a company belongs to the infringement market. In essence, this regression approach is similar to a simple comparison of average prices, but it isolates the part of the price difference that is related to the infringement rather than the other factors captured in the equation. Regression analysis, where it is feasible, generally leads to more robust results than simple price comparisons.

9.6  Comparator-based Approaches: Time Series

9.6.1  Choice of comparator period

9.69  Comparisons over time are probably the most commonly used approach to quantify damages. The approach is intuitive. In the LePage’s monopolization case in the United States—which we also discussed in Chapter 4—the court found that the ‘impact of 3M’s discounts was apparent from the chart introduced by LePage’s showing that LePage’s earnings as a percentage of sales plummeted to below zero—to negative 10 per cent—during 3M’s rebate program’.37 In the three years before 3M introduced its bundled rebates, LePage’s had a healthy operating income and significant growth in sales. After the introduction of the rebates, LePage’s lost sales to key customers and saw its operating income turn into losses. The court was satisfied that there was substantial evidence that the anti-competitive effects of 3M’s rebate programs caused LePage’s losses. Another relatively simple comparison over time was made in a damages action before the Regional Civil Court of Graz following an earlier ruling by the Austrian Cartel Court against five driving schools.38 The Cartel Court found that for a period of two months the schools had conspired to charge identical prices for the most popular driving courses, infringing the Austrian Cartel Act. The claim was brought by the Bundesarbeitskammer (the Federal Chamber of Workers) on behalf of customers of the driving schools. The Bundesarbeitskammer argued that the loss suffered by customers could be quantified as the 22 per cent difference between the price charged by the driving schools for the two months of the cartel’s duration and the lower average price in the market after the cartel had ended. The court accepted this calculation.

9.70  Although the time-series comparison is often described generically as the ‘before-and-after’ approach, it is more accurate to make an explicit distinction between three variants: before and during; during and after; and before, during, and after. Ideally, time-series comparisons should be made using information from both the pre- and the post-infringement period so that more data is used. This allows ‘anchoring’ the predicted prices for the infringement period and increases the likelihood of robust findings. ‘Before’ and ‘after’ data each have advantages and disadvantages. Post-infringement data is more likely to be available because it is more recent. One potential problem with using post-infringement data, however, is that it may take some time for the cartel or anti-competitive behaviour to unwind fully or for the market to return to competitive pricing, so some of the earlier ‘after’ data may be contaminated. You may need (p. 432) data over a longer period and to test for different end points of the infringement. One of the advantages of pre-infringement data is that the market equilibrium that it represents is not normally contaminated by the existence of the infringement. However, it may sometimes be difficult to ascertain when the infringement actually began. A further disadvantage is that the pre-infringement period may be so long ago that good data is difficult to obtain.

9.71  Time-series analysis has the advantage over cross-sections that the comparison involves the same companies and markets within and outside the infringement period. You still have to control for other factors that could result in differences in the market over time. Time-series regression analysis allows for this, and the same approach as described above can be used. Instead of looking at variations across n different companies, you analyse variations across n different time periods. The statistical techniques may be different, because data over time often displays characteristics not seen in cross-sectional data. There may be a seasonal trend or serial dependence (autocorrelation) in time-series data, which means that a high value now (at time = t) is associated with a high value tomorrow (t + 1). This is a potential statistical problem as the different observations over time may not be independent. More advanced time-series econometrics can control for trends and serial dependence.

9.6.2  Simple time-series techniques: Averages and interpolation

9.72  Comparisons of averages over time are similar to those described under cross-sectional comparisons. The average price in the market concerned during the infringement period is compared with the period before or after. As previously, a statistical test (e.g. a t-test) can be conducted to determine whether the difference is statistically significant.

9.73  Interpolation involves joining the price points before and after the relevant period to indicate what the prices would have been in the intervening period (we called this ‘anchoring’ earlier). In its simplest form the connecting line will be straight, as in the example in Figure 9.8. (p. 433) This is loosely based on a cartel damages case we have worked on. The top line shows the development of actual prices paid by the claimant. The starting point of the cartel is the date when, according to the European Commission’s infringement decision, the first meeting between cartel members was held. The starting price for the interpolation is based on an average actual price in the months before this start date. The end price is an average actual price in the months after the Commission’s dawn raids took place and the cartel broke down. You can see that the actual price plummets at about that time. The dashed line between these points shows the counterfactual price according to the interpolation exercise. More sophisticated versions of interpolation can incorporate seasonal patterns if that is a feature of the market. Figure 9.8 incorporates a different adjustment, namely one for exchange rate movements, since in this case the cartel fixed prices in one European currency and the claim related to prices in another currency which had devalued during the cartel period. This results in a new counterfactual line, as shown in the figure. It can be seen that the price increase due to the exchange rate movement explains a small part of the spike in prices during the cartel period, but there is still a significant overcharge effect.

Figure 9.8  Interpolation to determine the counterfactual (prices before, during, and after the cartel)

9.6.3  More sophisticated time-series techniques: ARIMA, dummy regression, and forecasting

9.74  Time-series models can be univariate (‘pure’) or multivariate. A univariate time-series model does not attempt to formulate a behavioural relationship between the variable under consideration (e.g. price) and other potential explanatory variables (e.g. costs). Instead, the historical pattern of the relevant variable itself is used as a predictor of its own future values. A multivariate time-series model, on the other hand, includes other explanatory variables and assesses the relationship between them to predict the relevant variable.

9.75  ARIMA (autoregressive integrated moving average) is a widely used pure time-series technique. Rather than simple interpolation it uses the pattern of past values of the variable under investigation to forecast its future values. Figure 9.9 illustrates this technique with an example based on an exclusionary conduct case we have worked on. Historical sales volumes of the claimant (a competitor to the infringing party) are modelled using ARIMA. The results of this are then used to forecast volumes during the period of the infringement—October 2009 to October 2013 in this example. The forecasts act as estimates of the counterfactual volumes, which can then be compared with actual volumes to estimate the harm resulting from the infringement (the lost volumes must be multiplied by the relevant profit margin per unit to determine the lost profit). If the comparison is between a period during the infringement and a period after, the process can be reversed such that the model backcasts (as opposed to forecasts) to the start of the infringement. In this example, the ARIMA modelling was on past data, but the figure also shows what happened after the infringement, which served as a cross-check of the results.

Figure 9.9  Example of forecasting using an ARIMA model (sales volumes of a competitor before, during, and after an exclusionary abuse)

Source: Oxera et al. (2009).

9.76  Time-series regression analysis uses the same basic equation as cross-section regression analysis: Pt = α‎ + βXt + δDt + et. Again this model specification seeks to explain prices P through a range of explanatory factors X, and a dummy variable D representing the infringement period. But can you spot the difference with the cross-section equation we saw earlier? The subscript to the variables is now t, representing different time periods, rather than i, representing different firms or markets (there are n time periods, so t ranges from 1 to n). Time-series regression controls for changes in the other explanatory variables X over time (such as changes in input costs). The dummy variable takes the value of one during (p. 434) the infringement period and zero in outside periods. In static time-series regression the price P in period t is assumed to be explained by the other factors X in that same period t. In dynamic time-series regression the price in period t can also be explained by the price (and hence other factors) in previous periods (t – 1, t – 2, etc.). Dynamic modelling is appropriate if there is sufficient data and there are reasons to believe that inter-temporal effects play an important role in the market in question (e.g. if prices are negotiated with the previous year’s price as a starting point).

9.77  There is sometimes heated debate among economists about the best way of doing time-series econometrics. The dummy-variable approach as described above is commonly used. The main alternative option is the forecasting approach, whereby you estimate the relationship between price and other factors for the period before or after the cartel only, but not during the cartel. Once estimated, you use the parameters to forecast or backcast what would have happened during the cartel.39 Proponents of the forecasting approach would argue that it is difficult to infer any meaningful estimates from the cartel period itself since the relationship between price and other factors may be distorted by the cartel. Only outside the cartel period do you observe how price is determined in the ‘normal’ course of business. For example, the cartel may have influenced costs as well as prices, so that costs during the cartel are not a good independent variable to explain price. Say the cartel lasted from 2006 to 2012 and you have data on prices, costs and other market characteristics from 2006 to 2014. Under the forecasting approach you estimate a model using the data for 2013 and 2014, establishing the relationship between price, costs and the other factors. With this (p. 435) model you then predict what price would have been in the 2006–12 period. A drawback of the forecasting approach is that the 2013–14 period may not represent the earlier cartel years very well if the market has changed significantly over time. Another disadvantage is that the after period may be too short to allow for a robust regression (monthly data for the two years may be adequate, but with quarterly or annual data two years would not be a sufficiently long period). Instead, the dummy-variable approach uses the data for the whole period. You can see that both approaches have advantages and disadvantages. Good practice would be to try both and assess which one fits the data best.

9.6.4  Time-series analysis to calculate the value of commerce

9.78  Determining how much the claimant has purchased from the cartel in the relevant period is the first step in quantifying damages. This value of commerce is normally a matter of fact: a claimant should be able to prove its purchases with invoices or purchase records on a database. Yet in practice this not always so straightforward. In our experience even the most sophisticated multinational companies do not always have a good record of their purchases of particular products, especially if the cartel was operating many years ago. For legal or tax purposes companies have to retain commercial data for a particular minimum period—this varies across jurisdictions, and is usually between five and ten years—but beyond that the records tend to become patchy. This is where economics can help on value of commerce. Gaps in data may be ‘plugged’ using simple interpolation, or even sophisticated time-series regression analysis. Sometimes there are sensible rules of thumb that can be applied. For example, in a damages claim against a car glass cartel, if a car manufacturer has incomplete records of the amount of car glass it has purchased but good records of the number of vehicles it has produced, a simple rule of thumb is that the manufacturer will have purchased at least one windscreen per car produced.

9.79  A US court accepted this approach in the LCD (liquid crystal display) cartel damages case.40 One of the claimants, mobile phone handset manufacturer Nokia, did not have full records of its LCD purchases. Its economic expert calculated the purchases based on the number of mobile devices that Nokia had sold over the relevant period. Approximately $11.6 million of the $52.3 million in claimed damages was based on these ‘inferred invoices’. The court agreed with the manner in which the expert had identified that Nokia’s purchase records were incomplete, and had subsequently derived the value of LCD purchases from data on mobile device shipments:

[The expert’s] finding that Nokia’s available LCD purchase data for years 2001 to 2004 is incomplete and her corresponding estimate of these purchases based, in part, on Nokia’s SEC reports, is grounded in facts … From her review of the LCD invoice data contained in Nokia’s R3 database, [the expert] noted that LCD purchases were ‘very low in the earlier years of the conspiracy compared with the later years: purchases were fewer than 50 million units in 2001, and grew to over 300 million units in 2007.’ … Finding this level of growth ‘questionable’, [the expert] ‘used other sources of data to check whether the Nokia invoice dataset reflected all of Nokia’s LCD purchases’, including Nokia’s ‘mobile device shipment data’ contained in its SEC Reports … ‘Relying on the fact that Nokia must purchase at least (p. 436) one LCD to produce a mobile device’, [the expert] compared the LCD invoice data and the shipment data. Based on this comparison, she concluded that ‘the invoice dataset does not contain all records of all LCD purchases.’ … [The expert] used the SEC reports to calculate the number of mobile phone handsets Nokia shipped for certain years, and then inferred the number of LCDs purchased by Nokia.

9.7  Comparator-based Approaches: Difference-in-Differences

9.80  Difference-in-differences analysis combines the cross-section and time-series approaches. It requires data both over time and across infringement and non-infringement markets, often referred to as a panel. The estimation techniques for panel data are similar to those often used for evaluating clinical trials and the effect of policy choices, in that one group has a ‘treatment’ applied to it (the infringement) while another that is not treated is used as a control group.41 The difference-in-differences analysis compares what happens to each group before, during and after the treatment. By using the control group, the analysis removes the impact of any changes that affect both treatment and control groups.

9.81  Figure 9.10 illustrates the difference-in-differences approach. It uses the average price in the treatment group (i.e. the infringement market) in the period during the infringement (A), and the corresponding averages for B (infringement market after the infringement), C (non-infringement market during), and D (non-infringement market after). The difference (A – B) reflects the change in prices in the market concerned during and after the infringement, while (C – D) reflects the change in the comparator market. Not all of the difference (A – B) is due to the infringement. There are other factors with an effect on price, and these can be inferred from the change in price in the comparator market, so (C – D). The difference in the differences in the average prices, that is, (A – B) – (C – D), is therefore used to identify separately the change in prices in the relevant market that is due to the infringement. Say A = €10, B = €7, C = €8, and D = €6. In the cartel market the average price fell from €10 to €7 after the cartel. However, not all of the €3 difference can be ascribed to the (p. 437) cartel, as there may be other factors resulting in lower prices, such as technological improvements or lower input costs. This can be seen from the comparator market, where average prices fell by €2 (from €8 to €6) over the same period. Only the difference between the two differences—€1—can be attributed to the cartel, so the overcharge is 10 per cent.

Figure 9.10  The difference-in-differences approach illustrated

9.82  Conceptually, the difference-in-differences technique is an improvement on pure cross-sectional and time-series models since it exploits the variations over time as well as across firms. This increased variability helps in the estimation of the effect of the infringement, and can also account for certain factors that might affect prices in the two markets. The data requirement for a difference-in-differences panel data regression is normally greater than that for cross-sectional or time-series regression, for the simple reason that you need data both across markets and over time.

9.8  Approaches Based on Financial Analysis

9.8.1  The role of financial analysis in determining the counterfactual

9.83  The tools for financial and profitability analysis that we discussed in Chapter 3 can also be used for damages estimations. In damages cases there is not always a clear-cut distinction between finance and non-finance methods, since a form of financial analysis is often involved at some stage. Some of the financial-analysis-based approaches can be seen as one form of application of comparator-based approaches, as counterfactual values are also often derived from comparator markets or time periods: for example, margins or profits during and after the infringement. Financial methods can also be used to construct counterfactual prices or profits by assessing information on the cost of production, cost of capital, and profit margins of the relevant market participants. This latter application of financial methods is different from the methods discussed thus far as it does not use comparators, but rather builds up the counterfactual using a combination of theory, assumptions, and empirical information. The market-structure-based approaches discussed in section 9.9 share this methodological feature.

9.84  Financial analysis can be used in several ways. First, the deterioration in the financial performance of claimants as a result of the infringement can provide an estimate of the harm caused to them. Second, the improvement in the financial performance of the defendant as a result of the infringement can provide an estimate of the benefits derived from the infringement. From a legal perspective, this is not a direct basis for determining compensatory damages, but in certain circumstances it may be used to inform the valuation of the damage suffered by the victims of the infringement (e.g. in cartel cases). Various techniques can be used for both types of analysis of financial performance—in particular, profitability analysis and valuation. A relevant benchmark would need to be identified as well, reflecting the profitability in the counterfactual. Finally, the counterfactual price level can be estimated by assessing the infringing parties’ production costs, and adding a return that reflects the degree of competition in the counterfactual.

9.8.2  Financial performance of claimants

9.85  The damage incurred as a result of the infringement should ultimately be reflected in the financial performance of the claimant. Hence, a comparison of the claimant’s actual financial performance with the financial performance that would be expected in the absence of the infringement can be used to provide an estimate (p. 438) of the damage. Financial performance is usually measured in terms of either profitability or company valuation. Valuation is closely related to profitability, since valuations of assets are usually based on the expected profits that can be achieved with those assets.

9.86  One example is the (partly successful) damages action before the Versailles Court of Appeal in France, where an excluded competitor claimed damages for lost profit based on its financial performance.42 This followed on from an earlier decision by the French Competition Council, which found that the defendants together had deliberately delayed the communication of information that was necessary for the claimant to operate in the market for media services. Since this information could not be obtained from any other source, the Competition Council concluded that the defendants’ conduct contributed to raising barriers to entry, thus constituting a breach of the French equivalent of Article 101. The claimant sought damages as a result of loss of clientele (€828,103), and damages resulting from the difference between its business plan and its actual financial results (€2,027,571). The court considered that while the claimant was entitled to recover damages as a result of its loss of clientele, the quantum of those damages should be reduced due to the claimant’s lack of experience in the business area in which it was starting up. The court therefore awarded only €100,000 to compensate for the lost opportunity to penetrate the market more quickly (it provided no explanation as to how it arrived at this figure). Moreover, it rejected the claim for damages resulting from the difference between the business plan and actual financial results, considering that since loss of clientele and loss of expected profits are one and the same category of harm, they can be compensated only once.

9.87  Another example is the exclusionary conduct case before the Court of Appeal of Milan, where the Italian competition authority had concluded that Telecom Italia abused its dominant position by preventing the claimant from entering the market for services for closed user groups.43 The claimant wished to provide a service that would have linked the telephone exchanges between its customers’ offices using a network infrastructure exclusively composed of switching nodes and dedicated lines leased from the defendant. The claimant was to pay Telecom Italia a fixed charge for the lease of dedicated local and trunk lines. However, it was found that Telecom Italia refused to lease the lines required to link the offices of the claimant’s customers, thus abusing its dominant position. A group of court-appointed experts calculated the claimant’s actual losses on the basis of documented costs that it had incurred. In relation to lost profits, the experts took into account a business plan drawn up by the claimant, but considered the projected figures relating to the acquisition of new customers to be too high. Moreover, they considered that the claimant’s future expansion would have been limited by the fact that it had not made sufficient investments in publicity and other promotional activities, that it lacked sales staff, and that it had experienced significant delays between the signing of new contracts and the activation of the service. Similarly, as regards loss of opportunity, the experts doubted the claimant’s argument that it would have had significant first-mover advantages, because there were no major barriers to entry into that market and it would have been difficult for the claimant to maintain customer loyalty once established.

(p. 439) 9.88  This last case shows how courts tend to treat with caution counterfactuals based on business plan analysis, a theme we saw before in the discussion of lost profits. Business plans can suffer from optimism bias. You can get situations where several competitors are excluded by a dominant rival, with each claiming that it would have achieved a certain market share in the counterfactual, but with those lost market shares adding up to more than 100 per cent. However, with some careful scrutiny of the parameters and assumptions used, and a cross-check against actual market indicators, contemporaneous business plans can still give you a good indication of what was expected to happen in the absence of the infringement.

9.8.3  Financial performance of defendants

9.89  In some cases compensatory damages can also be inferred from the financial performance of the defendant. For example, a cartel overcharge means that a particular cash flow is transferred from the buyer (claimant) to the seller (defendant). Therefore, the financial performance of the seller would be expected to be better than in the absence of the overcharge. The value of the cash flows that are transferred to the defendant may provide an estimate of the overcharge paid by the claimant. In the case of exclusionary abuses, the financial performance of the defendant can be used as a proxy for the value of the business opportunity from which a claimant was excluded. Consider the Chicago Bulls franchise case referred to earlier.44 The defendant prevented the claimant from acquiring an asset by refusing to enter into a contract for the provision of supplementary services. Fishman and Illinois Basketball Inc. (the unsuccessful bidders for the franchise) brought an action against the Chicago Professional Sports Corporation (the successful bidder), its shareholders, William Wirtz (the owner of the Chicago stadium), and others for refusing to contract with Illinois Basketball for the lease of the stadium and hence foreclosing competition in the market for the franchise. In this case, the actual financial performance of the defendant was considered in the damages quantification. The value of the damage to the claimant was calculated as the value of cash flows from the Chicago Bulls franchise generated by the winning bidder, effectively assuming that the claimant would have obtained this same value but for the infringement.

9.90  Another example is the Danish Supreme Court exclusion case in the ferry sector where the financial performance of the defendant was considered in the determination of the harm caused to the claimant, a competitor.45 De Danske Statsbaner (DSB) is the state-owned train operator which owned Gedser Harbour and operated ferry transport services to Germany. As the owner of the port, DSB collected port fees from another ferry operator, GT Linien, for operating a service there. However, it did not collect fees for the use of the port by its own vessels since these were exempt from this duty under Danish law. On appeal, the Danish Supreme Court upheld an earlier Eastern High Court judgment which found that DSB had abused its dominant position in the market for ferry transport between Denmark and Germany by collecting port fees from GT Linien without charging such fees to its own vessels, and that GT Linien was entitled to recover damages. In quantifying the damages, the Supreme Court based its estimate in part on reconstructed accounts of Gedser Harbour prepared on the claimant’s behalf, since the defendant was an integrated (p. 440) port authority and did not produce separate accounts. While the claimant argued that it should be entitled to recover DKK25 million (around €3.3 million), the court, after detailed consideration of the financial evidence, agreed with the defendant that the reconstructed accounts did not sufficiently take into account depreciations, reserves set aside for investments by the port, and interest on its invested capital. The court therefore awarded the claimant only DKK10 million (€1.3 million) in damages, but this was still based on the analysis of the financial performance of the defendant.

9.91  Financial data on the defendants can serve as a useful cross-check of damages estimates. For example, if a time-series comparator analysis produces a 25 per cent cartel overcharge, but the management accounts of the cartel members show profit margins on the product concerned of only 5–10 per cent in the same period, something may be missing from the analysis. A closer look is required. It may be that the time-series analysis has overstated the overcharge, or that the margin data does not capture all the effects of the cartel (e.g. the cartel may have resulted in higher costs as well as prices).

9.8.4  Bottom-up costing analysis to estimate the counterfactual price

9.92  This technique involves estimating the counterfactual price on the basis of bottom-up analysis of the costs and returns of the claimant or defendant. It typically involves determining unit costs of the product and adding a mark-up, expressed as a cash margin or a percentage profit margin on the costs. The resulting counterfactual price is compared with the actual price to obtain the per-unit value of the overcharge by the defendant or loss of revenue by the claimant. A 2008 exclusionary conduct case before the Düsseldorf Higher Regional Court provides an example of this approach.46 This stand-alone damages claim was brought against a state-owned local utility company by a direct competitor, in relation to a competitive bid for the supply of gas to a particular client. Despite having offered the lowest price, the claimant lost the bid because the utility had threatened the client with an increase in its district heating price if it were to source gas or electricity from other suppliers. The court considered the defendant’s conduct to be an abuse of a dominant position in the district heating market. It awarded damages corresponding to the claimant’s lost profits in relation to this particular contract, amounting to around 5 per cent of supply costs. This figure was determined through a bottom-up analysis, using detailed information that the claimant had provided on its own supply costs. The court considered that the claimant had proved that at the relevant time it would actually have been able to supply the contracted amount of gas, and that under normal circumstances it would have expected to earn a 5 per cent profit margin over and above supply costs.

9.93  Another example of the bottom-up costing approach is the Albion Water case discussed earlier, which involved an excessive wholesale access charge and corresponding margin squeeze.47 To determine the access fee that Dŵr Cymru would have charged in the counterfactual, the CAT took an average of three different measures of the company’s costs, resulting in a figure of 14.4 p/m3. It also determined that, in line with commercial and regulatory principles applied in the water sector, this access charge would have been increased year on year in accordance with the retail price index. The CAT then assessed whether Albion (p. 441) Water would actually have accepted an access fee of 14.4 p/m3 in a commercial negotiation in the counterfactual. After all, the company’s initial claim was that a charge of only 7 p/m3 would have been justified. In light of the factual evidence, the CAT considered that Albion Water’s approach to negotiations was not based on ‘die-in-ditch principles but on pragmatic business sense’, and that it would have agreed to 14.4 p/m3 even if this was above what it thought was justifiable.48

9.9  Approaches Based on Market Structure and Industrial Organization Theory

9.9.1  What are market-structure-based approaches?

9.94  IO models can be used to simulate market outcomes—particularly prices and volumes—in the counterfactual scenario. As you have seen in various places in this book, IO theory has developed a range of models of competitive interaction that predict a variety of market outcomes, from the least competitive (monopoly) to the most competitive (perfect competition). As with some of the financial-analysis-based approaches discussed above, market-structure-based approaches to quantifying damages differ from comparator analysis in that they build up the counterfactual based on a combination of theoretical models, assumptions, and empirical estimation, rather than comparisons across markets or over time.

9.95  The choice of model is important, given that the outcomes can vary significantly depending on the assumptions adopted. For any given factual situation where a cartel or other infringement has led to an increase in price, the more intense the competition is assumed to be in the counterfactual, the higher is the estimated damage from the infringement (as the counterfactual price is lower when competition is fiercer). In Concord Boat Corporation v Brunswick Corporation, a US case, the expert on the plaintiff’s side applied the Cournot model to determine the counterfactual.49 Under the standard Cournot model with two identical firms in the market, the model predicts that each firm will have a 50 per cent market share. The expert used this as the counterfactual and calculated damages for all periods when the defendant possessed more than 50 per cent market share. This particular use of the Cournot model was criticized by the court on the basis that it did not take into account differences between the quality of the two suppliers’ products, or external shocks, that could have led to the defendant possessing a market share of more than 50 per cent even in the counterfactual.

9.9.2  How close is the cartel outcome to monopoly, and how competitive is the counterfactual?

9.96  In cartel cases, the actual market outcome might resemble that of a monopoly. Cartel members co-ordinate output and prices to try to maximize joint profits. The best they can achieve is the monopoly profit. Yet there are several reasons why joint profit maximization by a cartel may not succeed. As we saw in Chapters 5 and 7, cheating, monitoring problems, and external demand and supply conditions may hinder effective co-ordination. Hardcore (p. 442) cartels may still achieve an overcharge, but the assumption that this actual cartel outcome is equivalent to a monopoly generally holds only in an upper-bound scenario.

9.97  What can be inferred from actual prices observed in periods of cartel breakdown? Are prices seen during the periodic price wars that occur in some cartels close to the counterfactual competitive outcome? The answers to these questions are not clear-cut. An important finding in economic theory is that periodic bouts of deviation and punishment, taking the form of sharp falls in prices, can be expected even in well-functioning cartels (Green and Porter, 1984). These temporary price wars give an indication of how low competitors in the market are prepared to go. However, such prices may not necessarily be sustainable in a competitive situation in the absence of the cartel.

9.98  Questions of this type were addressed in the paper wholesale cartel case before the German Federal Court of Justice in 2007.50 This judgment did not concern the quantification of damages, but of the overcharge by the cartel in the original investigation (German competition law at the time required fines to be based on estimates of actual overcharges). Following a finding of illegal price agreements in 2004, the wholesalers involved in the cartel were fined €57.6 million by the German competition authority. The cartel spanned ten regions in Germany from 1995 to 2000, and charged higher prices to smaller customers. The Federal Court of Justice disagreed with the method used by a lower court to estimate the overcharge in this case, which consisted of comparing the cartel price with the price charged by suppliers who were attempting to undercut the cartel price. The court found that such price cuts could not serve as a reference for the competitive market price since they were still dependent on the cartel price. It concluded that the prices after the cuts were still likely to be higher than the competitive price, and therefore that this method would underestimate the overcharge. In the absence of comparable reference markets, the court was of the view that the counterfactual price for estimating the overcharge should be established by way of an overall economic analysis. The judgment suggested a bottom-up costing approach, whereby an average profit margin—informed by comparator markets—is added to costs.

9.9.3  Use of IO models to determine or cross-check the counterfactual

9.99  As discussed in Chapter 7, competition authorities use IO models in merger control to simulate post-merger outcomes. Using IO models in damages cases is like merger simulation in reverse: in damages cases the counterfactual has more competition than the factual, while for merger cases competition is reduced in the counterfactual. In practice, the degree to which the theoretical IO models are calibrated can vary (calibration in this context means assigning values to the key parameters of the model so that it fits the actual data). At one extreme, little actual data is used and the analysis relies largely on assumptions regarding the main parameters of the theoretical model to simulate market outcomes. At the other extreme, all the main parameters of the theoretical model are estimated using actual data (e.g. estimates of the demand function and profit margins). The optimal approach will depend on data availability, and this in turn determines the weight you can place on the analysis.

9.100  Even where data availability is limited, IO models can still provide the basis for a cross-check of results from other quantification methods. If there is evidence to suggest that the counterfactual market structure has the characteristics of perfect competition or Bertrand oligopoly with homogeneous goods, the theory suggests that the cartel overcharge can be expected to be as high as the cartel members’ price–cost margin (since in competitive markets firms would set prices close to cost). If the counterfactual market structure is more like standard Cournot oligopoly, the overcharge can be approximated by reference to the cartel members’ price–cost margin and the number of firms in the market. Take the example of a cartelized four-firm market. The factual situation might be approximated through a monopoly model while the counterfactual is akin to Cournot oligopoly. A comparison of price outcomes under the two IO models enables a rough approximation of the possible overcharge (referred to as two-model estimation in Figure 9.6). If the cartel price–cost margin was 20 per cent, the cartel overcharge would theoretically have been 12 per cent. If the cartel margin was 40 per cent, the overcharge would theoretically have been 24 per cent. These results are derived from a formula for the cartel overcharge where the factual is monopoly and the counterfactual an n-firm Cournot oligopoly—the formula is m times (n – 1) divided by (n + 1), where m is the cartel margin (this formula assumes linear demand and symmetric, constant marginal costs). So with four firms, n is 4, and with a margin of 20 per cent, the result is 0.2 times (4 – 1 = 3) divided by (4 + 1 = 5), which equals 0.12 or 12 per cent. According to this formula, the overcharge increases as the number of firms in (p. 443) the counterfactual increases (because more firms implies a more competitive counterfactual), and as the cartel profit margin increases (because there is more profit margin to be ‘competed away’ in the absence of the cartel). These results are based on a comparison with a factual characterized as a monopoly. As noted above, cartels may not be as effective at raising prices as a monopoly. If this is the case, the above formula will tend to overestimate the overcharge, but can still be used as an upper bound when cross-checking the results from other approaches.

9.9.4  Example of a market-structure-based approach to estimate lost profits: Bus fights in Cardiff

9.101  In 2012, the CAT awarded £33,818.79 in lost profits to 2 Travel based on a relatively straightforward market structure model (a case we also discussed in section 9.3).51 2 Travel had been the victim of predatory conduct by Cardiff Bus, the incumbent bus operator owned by the local government. Its original claim had been for £50 million of lost profit. In April 2004, 2 Travel began a no-frills bus service on four routes in Cardiff, with a fifth route planned to begin operating at a later date. These were in-fill services: commercial bus services operating in between the tendered morning and afternoon school services that had been awarded to 2 Travel, and for which it could use the same vehicles. In response, Cardiff Bus introduced its own no-frills bus service (the ‘white services’) on the same five routes. This was in addition to its existing, ‘liveried’, bus services, which already operated in Cardiff on routes partly overlapping those of the new entrant. 2 Travel ceased operations in December 2004 and Cardiff Bus closed down its white services shortly thereafter. 2 Travel, which also had several operations outside Cardiff, went into liquidation in May 2005. In 2008, the OFT determined that the launch of the white services was an abuse of dominance.

(p. 444) 9.102  In 2011, 2 Travel claimed £50 million in lost profit under a number of different heads of damage. The largest of these heads—various lost commercial opportunities—were dismissed as too speculative and out of line with the factual evidence. In essence, 2 Travel claimed that, had it succeeded in establishing a foothold in Cardiff, it would have grown not only in Cardiff but in other areas as well. The idea was that Cardiff Bus’s actions caused it to go bankrupt before any of this commercial success could be realized. However, the factual evidence showed that even before entering the Cardiff market, 2 Travel had been a ‘poorly run and administered [business] in almost constant financial difficulty’.52 Internal communications by the company’s finance director recommended going into liquidation rather than enter the Cardiff market. The CAT also found that the (limited) lost revenue resulting from the infringement would not have been enough to prevent insolvency. In the end, the only lost profit claim upheld by the CAT was that for the four routes that 2 Travel actually operated in Cardiff between April and December 2004.53

9.103  The OFT determined that Cardiff Bus had launched the entire white services operation for the sole purpose of driving 2 Travel out of the market. As such, the experts for both sides, and the CAT, agreed that the relevant counterfactual was one in which Cardiff Bus did not operate its white services at all. The experts and the CAT agreed that, in the counterfactual, the passengers who actually travelled on the white services would have used either 2 Travel’s services or Cardiff Bus’s liveried services. It was also common ground between the parties that the overall number of passengers would not have grown following entry. Since Cardiff Bus’s liveried services on the same routes were operating in both the factual and the counterfactual scenarios, it was agreed that none of the actual passengers on these services would have transferred to 2 Travel. Therefore, the maximum number of additional passengers (i.e. in addition to those who already travelled on 2 Travel) that 2 Travel could possibly have attracted were the 150,727 passengers who travelled on the white services over the relevant period.

9.104  The overall approaches to determining how many of these passengers would have travelled with 2 Travel, and the revenue that would have been generated, were similar among both experts and the CAT. However, their analyses differed on some key assumptions. The expert for Cardiff Bus developed a market-structure-based model of 2 Travel’s operations on the relevant bus routes that incorporated a range of facts and assumptions about the main parameters, including the number of services that 2 Travel actually ran from April to December 2004; the number and types of additional passengers that 2 Travel would have gained; and the costs that 2 Travel would have incurred. To determine counterfactual market shares, the expert for Cardiff Bus assumed that passengers get on the first bus that comes along. This is based on a common finding in the bus market. Passengers are more time-sensitive than price-sensitive, especially for frequent services such as those on the four routes in question (for infrequent services—running only once or twice per (p. 445) hour—passengers are more inclined to check the timetable before turning up at the bus stop). The passenger allocation between the two operators was thus approximated by considering the relative frequencies of their services. For example, if one operator runs six buses an hour on a particular section of the route, and another operator runs two buses an hour on that same section, it is assumed that the first operator would get 75 per cent of the passengers and the second 25 per cent. In the main scenarios, it was estimated that 2 Travel would have been able to achieve an average market share of 19 per cent across all sections of the four routes, since this was the share of frequencies relative to those of the Cardiff Bus liveried services on the same routes. This is a simple market structure approach to determine the counterfactual.

9.105  Once the counterfactual passenger numbers had been calculated, they were multiplied by 2 Travel’s fares to estimate its counterfactual revenue. Using an approach similar to that employed by the expert for Cardiff Bus, the CAT allocated the actual number of white service passengers to 2 Travel based on frequency. It thus estimated that, in the counterfactual, 2 Travel would have attracted an additional 41,255 passengers between April and December 2004. At an average fare of just under £0.82 per passenger per journey, the CAT estimated the lost revenues to be £33,818.79. This estimate lies within the range that was provided by the expert for Cardiff Bus, and was an order of magnitude lower than the amount claimed by 2 Travel.

9.106  Having reached a view as to what revenue 2 Travel lost as a result of the infringement, the additional costs that the bus operator would have incurred in order to generate that revenue needed to be considered to reach a figure for its lost profits. There was some disagreement about the treatment of incremental costs in the counterfactual analysis. 2 Travel argued that it could not operate its services in full because of a driver shortage caused by Cardiff Bus’s conduct. The expert for Cardiff Bus considered that if, in the counterfactual situation, 2 Travel had operated to its full registered timetable, it would have required extra drivers. These drivers would have constituted incremental costs compared with 2 Travel’s costs in the actual situation. It follows logically that the additional driver costs should be deducted from the estimates of lost profit for the counterfactual scenarios in which 2 Travel operated more services than it actually did. 2 Travel and its expert had not taken such incremental costs into account. Ultimately, the issue of incremental costs was not directly addressed in the judgment, since the CAT rejected the counterfactual scenario with 2 Travel’s additional services. However, the principle remains that if a claimant would have undertaken greater activity but for the infringement, and hence have achieved higher revenue, one has to deduct from this revenue the incremental costs associated with the greater activity (in line with the framework set out in section 9.3).

9.10  Pass-on of Overcharges

9.10.1  The policy debate about the passing-on defence

9.107  There have been extensive policy debates about whether the passing-on defence should be permitted. This defence means that purchasers or competitors who have suffered harm in the form of higher costs are not entitled to damages if they have passed the higher costs on to their own customers in the form of higher prices. In the United States, pass-on has been ruled out as a defence by the federal courts, and typically only direct purchasers can claim (p. 446) cartel damages (a number of US states do allow indirect purchasers to claim damages).54 Economic incentives have played a role in this policy decision. The reasoning is that direct purchasers are best placed to file a claim, as they will generally have the best information available and may also be more likely to have the resources to make a claim (at least compared with end-consumers). Ruling out the passing-on defence gives direct purchasers better incentives to file claims.

9.108  In Europe, the general approach reflected in the 2014 Directive on antitrust damages actions (Articles 12 to 15) is that the passing-on defence should be allowed, and that direct and indirect purchasers can claim for damages. This is in line with the compensation principle: if direct purchasers have passed on an overcharge to their own customers (i.e. the indirect purchasers), the former should not be over-compensated, and nor should the latter be under-compensated. However, the EU policy to allow the passing-on defence and indirect purchaser claims does come with some further clauses. Article 12 of the Directive makes it clear that a claimant who has passed on an overcharge may still have suffered a loss of profit (this would be mainly in the form of lost sales volumes, as we saw in section 9.3). Article 14 establishes a rebuttable presumption that the overcharge has been fully passed on to the specific indirect purchaser who brings the claim—we return to this later. Article 15 seeks to ensure that claims made at different layers of the supply chain do not lead to multiple liability. This makes economic sense. Regardless of how many layers of the supply chain an overcharge is passed on to, and in what proportion, the sum of the overcharge harm claims from the various layers cannot exceed the level of the overcharge itself. In Figure 9.1 at the start of this chapter, area A is the total overcharge, so the total harm suffered across the various layers of the supply chain cannot exceed A. There should be no double-counting of harm (which in legal terms would result in a form of unjust enrichment by some parties in the chain). For example, if it is found that direct purchasers of the cartel have passed on 75 per cent of A, and their respective customers have passed on 90 per cent of their price increase to end-consumers, the direct purchasers have suffered a harm equal to 25 per cent of A, their customers a harm of 7.5 per cent of A (75 per cent of A but with 90 per cent of that passed on), and end-consumers 67.5 per cent (90 per cent of 75 per cent of A). By the same token, however, the total overcharge harm caused along the chain is still 100 per cent of area A, so if all parties in the different layers of the chain (including final consumers) found some way of making a joint claim and distributing the damages award among themselves, area A would be the right overcharge amount to claim for in that joint action.

9.10.2  Pass-on in theory: The relationship between prices and costs in economic models

9.109  In the standard economic models of competition, oligopoly and monopoly, there are defined relationships between price and marginal cost. These provide the basic theoretical understanding of how cost changes translate into price changes. In all these models, companies are assumed to maximize their profit given a certain level of marginal costs and the degree and nature of competition they face. The resulting equilibrium prices can be expressed as a function of marginal cost. For example, in perfect competition price equals marginal cost, (p. 447) so if the marginal cost to all suppliers increases due to a cartel overcharge on an input, the price will increase correspondingly.

9.110  Cost pass-on refers to the proportion of a cost change that is translated into a change in the final price. It is usually represented as a percentage pass-on rate: the change in price expressed as a percentage of the change in the marginal cost. If costs per unit increase by €10 and the price increases by €5, the pass-on rate is 50 per cent. This percentage pass-on rate is straightforward to interpret, and can be applied directly to the total overcharge. For example, if the overcharge is €3 million, and the percentage pass-on rate is 50 per cent, this means that €1.5 million of the overcharge has been passed on.

9.10.3  Pass-on in competitive markets: Industry-wide versus firm-specific cost increases

9.111  A distinction must be made between firm-specific and industry-wide cost increases. In perfect competition, an overcharge that affects all competitors in a downstream market (i.e. is industry-wide) would be passed on in full. You may find this result counterintuitive, and so do some business people: ‘my market is highly competitive, surely I cannot pass on any cost increase?’ But it simply follows from the fact that, under perfect competition, prices equal marginal costs. All downstream firms that remain in the market therefore see no change in profit level.55

9.112  In contrast, for a cost increase that affects only one, or some, of the competitors in those markets, the expected pass-on rate would be close to zero, since those competitors that do not face the increase can leave their prices unchanged. Those that do face the increase must absorb it to stay competitive or exit the market. This may also be the case if, for example, an entire industry is affected by the overcharge, but that industry competes with another industry that uses a different upstream input not subject to the overcharge. For example, if an upstream cartel operates only in Europe, and downstream companies compete with non-European producers which are unaffected by the cartel, the European downstream companies may face difficulties passing on the cartel input overcharge. A similar logic was applied by the Valladolid Provincial Court in Spain in a 2009 damages action brought by a biscuit producer against a sugar cartel.56 Spanish sugar producers were found to have colluded to fix the prices of sugar for industrial use between February 1995 and September 1996. The court considered that biscuit producers in Spain were harmed because they compete in European markets with foreign biscuit producers which did not purchase their sugar in Spain. By implication the Spanish biscuit producers had to absorb the overcharge on sugar or else lose market share. In 2014 the Spanish Supreme Court applied the same reasoning in a parallel case against another member of the sugar cartel.57 This claim was brought by a group of producers of biscuits, desserts and confectionary, including Nestlé, LU, and Wrigley. The Supreme Court noted that the Spanish confectionary industry exports a significant proportion of its production. The cartel overcharge on Spanish sugar (p. 448) would therefore have affected the industry’s competitiveness vis-à-vis foreign competitors and would have made pass-on difficult, with a negative effect on profit margins.

9.10.4  Pass-on in monopoly and other models of competition: An illustration

9.113  Another well-known theoretical finding is that a monopolist with linear demand and constant marginal cost passes on exactly 50 per cent of the cost increase. You may also find this result counterintuitive: why doesn’t the monopolist pass on cost increases in full? The reason is not benevolence but self-interested profit maximization—if costs change, so does the profit-maximizing price (we explained this in Chapter 2). Figure 9.11 illustrates this logic. It shows the demand and marginal cost curves in the downstream market, where marginal cost includes the cartelized input. Marginal cost per unit equals 2. If the downstream market were perfectly competitive, the price would be 2 and output 8. If it were a monopoly, the price would be 6 and output 4. Now the upstream cartel causes the industry-wide marginal costs to increase from 2 to 4 (in reality a 100 per cent overcharge would be highly unlikely, but this is a stylized example). You can see that in perfect competition, the new price is 4 and output 6. So the cost increase resulting from the cartel is fully passed on—hence there is no overcharge harm to downstream manufacturers that have purchased from the cartel (as noted before, the downstream output decrease from 8 to 6 constitutes a different type of harm—a volume loss—but this is generally more difficult to claim in practice). In monopoly, marginal costs now equal marginal revenue where output is 3. The new price is 7. So the cost increase of 2 units has resulted in a 1-unit price increase, from 6 to 7—that is, a 50 per cent pass-on.

Figure 9.11  Pass-on of a cost increase

9.114  In oligopolistic markets you typically get results that are in between perfect competition and monopoly. In the standard Cournot oligopoly model (with constant marginal cost and linear demand), the pass-on rate for an industry-wide cost change can be expressed as n divided by (n + 1), where n is the number of firms (ten Kate and Niels, 2005). Therefore, for two firms the pass-on rate is two-thirds, while for seven firms it would be seven-eighths. The pass-on rate increases with the number of firms, which is consistent with the 50–100 per cent range between monopoly and perfect competition, as identified above. The standard (p. 449) Bertrand oligopoly model with homogeneous goods has the same outcome as perfect competition, and pass-on is 100 per cent. In the Bertrand model with differentiated goods, and in monopolistic competition, firms behave like monopolists for their own product, so with linear demand and constant marginal cost the pass-on rate would be at least 50 per cent. However, it also depends of the degree of differentiation and the number of competitors. Less differentiation and more competitors gets you closer to 100 per cent pass-on.

9.10.5  Pass-on in theory: Further insights

9.115  In some cases the overcharge may have caused significant changes in the dynamics of competition in the downstream market—smaller operators may have been forced to exit, for example. In theory, this may give rise to pass-on rates greater than 100 per cent, as increased downstream concentration may result in higher downstream prices (although the term pass-on is not accurate in such a situation, as, in reality, a chain of events has taken place). Such factors would need to be assessed on a case-by-case basis.

9.116  In the various models of competitive interaction, companies set their profit-maximizing price with reference to marginal costs. Fixed costs do not directly determine price in the same way as marginal costs, at least in the short run (in the longer run, many fixed costs tend to become variable). Therefore, a change in fixed costs due to an infringement may not be passed on in the same way. However, fixed costs can influence whether a firm can viably operate in the market in the first place—that is, the margins between price and marginal cost need to be at least sufficient to recover fixed costs. An increase in fixed costs may, in the longer term, induce exit and lead the remaining firms to increase price, in which case there may be full pass-on eventually. The effect of changes in fixed costs should also therefore be assessed on a case-by-case basis, and the duration of the infringement becomes an important factor to consider.

9.117  Finally, buyer power of downstream customers can influence the ability of downstream suppliers to pass on the overcharge on the upstream input. If strong buyers can credibly switch to alternatives, this may limit the ability to pass on cost increases, in line with the theoretical insights presented above. However, if buyer power has already been exercised and has meant that prices equal marginal costs, the situation may be similar to that in a competitive market where pass-on is near 100 per cent. Again, in this situation, there may also be a volume effect that can give rise to a different type of damage from the overcharge. Pass-on may also depend on the type of negotiation with strong buyers. In some arrangements between suppliers and large retailers, prices may be renegotiated only if there are changes in costs of significant input items. In this case it therefore matters whether the cartelized upstream product constitutes a large and visible input into the downstream product, or whether it represents only a small fraction of total costs. We discuss this below.

9.10.6  Small versus large cost increases

9.118  A question that arises frequently relates to the pass-on of very small cost increases, or small cost items. Car glass makes up only a small fraction of the total price of a car; methionine makes up only small fraction of the price of chicken feed (and an even smaller fraction of the price of a chicken). On the other hand, candle wax makes up a significant proportion of the costs of a candle. Does this make a difference for the pass-on rate? Assume that the overcharge on an input is 20 per cent. If the input price represents 50 per cent of the downstream product price (akin to the candle wax example), full pass-on of the overcharge (p. 450) would mean a 10 per cent increase in the downstream price—that is, if the candle price is 100 cents and the candle wax costs 50 cents, and the cartel raises this by 20 per cent to 60 cents, 100 per cent pass-on means a new candle price of 110 cents, so 10 per cent higher. On the other hand, if the input price represents only 5 per cent of the downstream price, the latter increases by 1 per cent (costs were 5 cents, the cartel increases this to 6 cents, so a 1 cent increase if fully passed on). You can see how a 10 per cent price increase downstream is more significant (and noticeable) than a 1 per cent increase.

9.119  In the standard economic models it does not matter whether a particular input represents 5 per cent or 50 per cent of total marginal costs, as prices are set with respect to these total marginal costs. In practice it does matter. However, the direction of the effect is not clear-cut. If the affected input cost makes up only a small proportion of the final-product price, there may be no pass-on if the downstream producer chooses not to reset prices. This may occur where downstream prices are set, or negotiated, with respect to only major and more visible input costs, or where there are ‘menu’ costs associated with changing and communicating the final-product price (most restaurants don’t print a new menu every time the costs of meat or vegetables change). Alternatively, small changes in the input price may well be fully passed on in some circumstances if their magnitude is sufficiently small as to avoid any significant demand reduction. In the above example, the downstream producers may get away with a 1 per cent price increase but not a 10 per cent one. This reasoning was used by the Paris Commercial Court in a 2007 vitamins cartel case (Juva v Hoffman La Roche).58 The court considered vitamins to be a small part of the finished good and that a small price increase would be sufficient to offset the overcharge. It also noted that the price of the claimant’s finished good had increased by more than the prices of the vitamins, and that its sales volumes had also grown.

9.120  Economic theory is divided on this point. One economic approach is to assess the relationship between price and overall costs (or overall marginal costs), and infer from this that the pass-on rate is the same for all individual cost items, be they large or small.59 After all, when setting prices, companies may look at their total costs in the round, and hence even small items are considered in this way. Total costs are simply the sum of individual cost items. According to this logic, if the overall pass-on rate of marginal costs is found to be, say, 80 per cent, this rate can also be applied to each of the individual marginal cost items, including the small ones. The counterview is that there may well be differences between small and large cost items, both in commercial reality and in theory. In practice, companies may change prices with regard to only major or more visible input costs, while small-cost items are ignored. Can there still be pass-on in such circumstances?

9.121  Take the example of a law firm that has acquired an espresso machine for its staff area, at a price of €299. It turns out that espresso machine manufacturers had formed a cartel such that the law firm paid an overcharge of 10 per cent, so €29.90. Has the law firm passed this additional cost on through its fees to clients? Let’s say this successful law firm increases its hourly rates by 10 per cent at the start of the new year. When setting the new rates, the firm’s management committee did not pay much heed to the cost of the espresso machine—that is, it did not increase the fees with the espresso machine in mind. However, it did look at the (p. 451) overall expenditure of the business—in particular, lawyer salaries and office costs, which include the cost of the espresso machine. The fee increase will earn the law firm a couple of million euros in extra revenue, clearly dwarfing the €29.90 it overpaid on the espresso machine. In that sense the overcharge is easily recovered. But does that mean the law firm has actually passed on the costs of the espresso machine to its clients? Or did it absorb these costs? And would its fee increase have been any different had the espresso machine been cheaper? There are different views—legal, economic, and even philosophical—on the question of whether the fee increase can be regarded as pass-on.

9.10.7  The effect of pricing practices and price friction

9.122  In any specific case you would expect there to be an analysis of the way companies in the downstream market concerned actually set prices. The theoretical relationships between costs and prices may not always hold, particularly over short time periods. Some companies price on a cost-plus basis, while others may have explicit contracts with customers through which increases in input costs are agreed to be passed on in full. Prices in some industries are changed on an annual (or other periodic) basis rather than continually in response to cost changes, or are determined at the end of a production cycle, long after input prices have been agreed.

9.123  Cost-plus pricing occurs where a seller calculates the cost of the product and then adds a fixed mark-up. If a retailer applies a 25 per cent mark-up on all products, a product that costs €9.00 will sell for €11.25. If an upstream cartel raises the wholesale price to €10.00 (10 per cent overcharge), the retail price is increased to €12.50. The retailer has thus not only passed on the €1.00 input cost increase in full, it has raised its own price by €1.25, implying a pass-on rate of 125 per cent. Yet from both a legal and an economic perspective it is not clear that the extra €0.25 actually represents pass-on as such; it is more a resetting of price in which the change in cost was just one of the relevant factors. Another question is whether the mechanistic pricing rule—in this case, the 25 per cent mark-up—has been applied consistently during the cartel period; the retailer may have had to adjust its percentage mark-up to reflect changing market conditions.

9.124  Another pricing practice is to use specific price points, such as €9.95 rather than €10.05, or a rounded price of €11.00 or €12.00 rather than €11.25. Sellers may also prefer simplicity and continuity in pricing. The price of the print edition of The Times newspaper remained at £1 for several years (it is now £1.20), while the price of a Big Mac (an international pricing benchmark monitored through The Economist’s Big Mac Index) in Spain was €3.50 for years.60 Price changes in all these situations can be lumpy, and may not be directly related to changes in underlying costs. The rate of pass-on may be low as a result, at least in the short term.

9.10.8  Empirical evidence on pass-on

9.125  Empirical studies on pass-on in competition law damages cases are rare. Fields in economic literature where pass-on has been studied empirically include tax incidence, exchange rate movements, and the transmission of intermediate goods prices. They provide some support for the theoretical insights discussed above. Various studies have confirmed very high pass-on rates where the downstream market is highly competitive—for example, in petrol (p. 452) retailing and various agricultural products. Full pass-on is sometimes achieved only after a lag of several months, which may or may not affect the damages estimate, depending on the length of the period considered (the longer the period, the greater the pass-on). In addition, some studies have found that the pass-on rate is higher for price increases than for price decreases. In the standard IO models, such asymmetry does not exist, but for practical reasons prices may sometimes be ‘sticky’ downwards. This is also known as the ‘rockets and feathers’ phenomenon.61 One US-based study found that the pass-on rate for an industry-wide cost shock in raw milk was 92–94 per cent in wholesale prices and 85–87 per cent in retail prices (Dhar and Cotterill, 1999). The study also tested separately for the effects of firm-specific cost shocks. It found that one supplier with significant market power had a firm-specific pass-on rate of 50–60 per cent (in line with the assumption for the monopoly level), while other suppliers with limited market power had a low pass-on rate of 13–19 per cent (in line with the theoretical result where cost changes are firm-specific).

9.126  Some studies on exchange rate pass-on provide further support for the assumption regarding high pass-on rates where downstream markets are competitive. Exchange rate pass-on is conventionally defined as the percentage change in an imported good’s local-currency price for a given percentage change in the nominal exchange rate. One study on the US automobile market found that the pass-on rate for Japanese cars was 15–30 per cent when an exchange rate shock occurred, while for German cars it was above 65 per cent (Goldberg, 1995). During the period in question, Japanese car manufacturers competed mainly in the small-car segment, and were therefore constrained by domestic competitors that were not subject to the exchange rate shock. Their pass-on rate was relatively low. In contrast, German car manufacturers mainly served the luxury-car segment, where there tended to be fewer competitors. Their pass-on rate was high.

9.127  Several empirical studies have analysed how retailers pass on changes in VAT and other taxes. Overall, with some exceptions, these studies indicate high degrees of pass-on by retailers, often between two-thirds and 100 per cent. For example, Blundell (2009) estimated that 75 per cent of a 2.5 per cent VAT rise in the United Kingdom would be passed on to consumers. The Deutsche Bundesbank (2008) investigated the price effects of the VAT increase in Germany from 16 per cent to 19 per cent. Following an econometric analysis, controlling for seasonal effects, exchange rates and additional government measures, the study concluded that higher tax rates had been largely passed on to consumers.

9.128  What is the relevance of these studies for specific damages cases? As with the empirical studies on overcharges discussed earlier, the literature on pass-on provides some interesting background on what kind of pass-on rates you may encounter in practice in different types of industry. It also provides some high-level support for the theoretical insights presented here. However, the literature is not sufficiently clear-cut to derive any presumptions about actual pass-on rates. These will have to be determined on a case-by-case basis.

9.10.9  Pass-on and volume effects

9.129  Where a purchaser has passed on the overcharge in part or in full it may still have suffered a volume harm. After all, higher prices downstream often means lower sales. Legally this is a separate type of harm from the overcharge. The EU Directive on antitrust damages actions (p. 453) (Article 12) states that claimants are still entitled to compensation for loss of profits due to a full or partial passing-on of the overcharge. The ECJ acknowledged the possibility of such a loss-of-volume harm in the presence of complete pass-on in a 1997 judgment on port fees that were illegally levied in the French territory (this was not a competition law ruling).62

9.130  Claims for volume loss from pass-on of an overcharge are not yet common, in part because they are difficult to prove (though perhaps also because claimants would often prefer not to admit to a high pass-on rate when making their case). The downstream price increase would have to be significant for it to have a noticeable effect on downstream volumes. In the earlier example, if the downstream price increase was only 1 per cent it would be difficult to identify any effect on volume. Having an estimate of the own-price elasticity of demand for the downstream product would help. In that case even the effect of a 1 per cent price increase may be calculated, if not actually proven empirically. For example, if the elasticity is –2, a 1 per cent increase in the price would have resulted in a 2 per cent reduction in volume. Once the lost volumes have been estimated, the lost profit can be derived in line with the conceptual framework discussed in section 9.3.

9.10.10  Passing the buck: Remaining policy and practical questions

9.131  Court rulings on the level on pass-on are as yet rare. Best practice on how to quantify pass-on is still developing. In many cases, determining the level of pass-on will involve applying the theoretical insights on market structure, combined with an assessment of how downstream companies actually set prices. If there is sufficient data, a quantitative analysis of pass-on may be undertaken as well, considering metrics such as the correlation between claimant costs and downstream prices, or the development of the claimant’s downstream profit margins over time. In any event, as with the overcharge, both claimants and defendants will have to do some homework to estimate the likely degree of pass-on; there are no presumptions that follow from economic theory.

9.132  What about the rebuttable presumption in the EU Directive on antitrust damages actions? This states (in Article 14) that wherever claimants are in the supply chain, it can be presumed that the overcharge has been passed on to them:

  1. 1.  Member States shall ensure that, where in an action for damages the existence of a claim for damages or the amount of compensation to be awarded depends on whether, or to what degree, an overcharge was passed on to the claimant, taking into account the commercial practice that price increases are passed on down the supply chain, the burden of proving the existence and scope of such a passing-on shall rest with the claimant, who may reasonably require disclosure from the defendant or from third parties.

  2. 2.  In the situation referred to in paragraph 1, the indirect purchaser shall be deemed to have proven that a passing-on to that indirect purchaser occurred where that indirect purchaser has shown that:

    1. (a)  the defendant has committed an infringement of competition law;

    2. (b)  the infringement of competition law has resulted in an overcharge for the direct purchaser of the defendant; and

    3. (c)  the indirect purchaser has purchased the goods or services that were the object of the infringement of competition law, or has purchased goods or services derived from or containing them.

(p. 454) 9.133  This presumption reflects a policy choice to make claims by indirect purchasers easier. It does not have any economic grounding. It cannot be presumed from the outset that the overcharge has ended up with any particular layer of the chain. The presumption may also complicate matters if there are simultaneous claims by direct and indirect purchasers. Moreover, pass-on is generally difficult to prove for defendants (and hence the presumption difficult to rebut) as this requires data from direct and indirect purchasers about their pricing downstream.

9.134  A final policy point is that competition authorities should be careful what they write in their decisions. Government agencies are often under pressure to demonstrate the consumer benefits of their interventions. This is one reason why infringement decisions sometimes include statements along the lines of the effects of the anti-competitive conduct (a cartel or abuse of dominance) ultimately being felt by end-consumers. Such statements can be unhelpful to a claimant who operates at an intermediate layer of the supply chain, since they imply that all cost increases have been passed on to end-consumers. For example, in a damages action in France against the vitamins cartel, the Nanterre Commercial Court noted that the earlier European Commission decision and press release had stated that the cartel ultimately affected end-consumers, and inferred from this that direct purchasers were able to pass on the cost increase.63 It is open to question whether the Commission’s statement about end-consumers being harmed was based on any actual analysis specific to the case at hand, or whether, perhaps more likely, it was meant as a general statement about cartels ultimately being bad for end-consumers. What is clear is that the intention of competition authorities to show that their interventions benefit consumers can sometimes, inadvertently, get in the way of follow-on damages actions by providing apparent support for the passing-on defence.

9.11  Interest and Discounting

9.11.1  Relevance of interest and discounting

9.135  A competition law infringement may have started a long time ago and lasted many years. Its effects may be felt for years into the future. Quantifying damages therefore involves calculating monetary amounts for a number of years in the past, and possibly in the future. The economic principle of the time value of money means that you cannot just add up amounts of money from different years—€100 today is worth less than €100 last year, and more than €100 next year. Translating past and future monetary amounts into a present value requires discounting, as we also saw in Chapter 3. For this you apply a discount rate. Future amounts are discounted applying this rate, and past amounts are uprated. Applying interest on past damages values is a form of uprating. The time value of money has also been recognized in the EU Directive on antitrust damages actions, which regards the application of interest as being in line with the principle of compensation:

The payment of interest is an essential component of compensation to make good the damage sustained by taking into account the effluxion of time and should be due from the time when the harm occurred until the time when compensation is paid, without prejudice to the (p. 455) qualification of such interest as compensatory or default interest under national law and to whether effluxion of time is taken into account as a separate category (interest) or as a constituent part of actual loss or loss of profit. It is incumbent on the Member States to lay down the rules to be applied for that purpose.64

9.136  US courts do not award interest on past damages; only post-judgment interest is awarded (and sometimes pre-judgment interest if it can be shown that the defendant has unduly delayed the litigation process).65 In jurisdictions across Europe and elsewhere the legal rules and practices regarding the award and calculation of interest vary significantly. The rules are also frequently at odds with economic principles. One specific issue is whether the interest is paid on a simple or a compound basis. Another is whether a statutory interest rate or some market rate is applied. There is also much debate, among economists as well as lawyers, about which rate is most appropriate if no statutory rate is prescribed: a risk-free rate, the cost of capital, or something in between? The choice of interest rate and method can have a significant effect on the damages quantification, especially in cases involving long-running cartels. In the damages action brought by National Grid, the operator of the electricity transmission network in Great Britain, against the members of a gas insulated switchgear cartel—active from 1988 to 2004—the overcharge claim amounted to £95 million without interest, and £276 million with interest.66

9.11.2  Principles of interest and discounting

9.137  Conceptually, the discount rate should take into account the time value of money, inflation, and risk. Inflation means that prices tend to rise over time and hence the same nominal amount of money decreases in value. Risk means that future expected cash flows are uncertain. Applying interest on damages is a form of uprating cash flows. If an infringement has caused the victim a loss of €100 during each of the past five years, each year’s loss needs to be uprated using the discount rate to determine the current value of this harm. Suppose that the discount or interest rate is 10 per cent per year. The harm from the first of the five years (i.e. the first €100) needs to be uprated five times, which is comparable to paying cumulative interest on that amount for five years. The current value of that amount is €161.05 (€100 times 1.10 to the power of 5). The harm from the second year needs to be uprated for four years (€100 times 1.10 to the power of 4, which equals €146.41), and so on. The present value of the total harm over the five years is €671.56.67 Furthermore, if it is demonstrated, and accepted by the court, that the now ceased infringement will still cause losses to the victim in the subsequent three years (because the victim cannot immediately recover the market position it would have had in the absence of the infringement, for example), those future losses form part of the harm suffered. They need to be added to the present value of (p. 456) the harm over the first five years. Suppose the future losses are €75, €50, and €25 (as the victim is gradually regaining market share to its ‘normal’ levels), and the same discount rate applies. The €75 occurs in the current year, so does not require discounting. The €50 occurs in the next year, so needs to be discounted once, and is worth €45.45 in present terms (€50 divided by 1.10). The €25 in two years’ time is worth €20.66 in present terms (€25 divided by 1.10 to the power of 2). The present value of the total harm over the whole eight years in this example (five past years, the current year, and the two future years) is now €812.68.

9.138  The choice of discount rate can have a significant influence on the damage value. In the above example, if the discount rate were 5 per cent instead of 10 per cent, the present value of the damage from the five past years would be €580.19 instead of €671.56. If it were 15 per cent the value would be €775.37. The higher the discount rate, the greater the present value of the past losses when uprated, but the smaller the present value of the future losses when discounted at this rate. We turn to the choice of interest or discount rate next.

9.11.3  Choice of interest or discount rate

9.139  Various jurisdictions require statutory rates of interest—generally prescribed by civil or contract/tort law provisions—to be used for uprating damages. In Germany it is five percentage points above the Bundesbank base rate; in the Netherlands it is seven percentage points above the European Central Bank base rate.68 The date from which interest can be claimed varies across jurisdictions, and can refer to the start of the infringement (as in the EU Directive, cited above), the start of the legal action, or the date of the award. The relevant legal framework usually determines which part of the cash flows in the damages valuation should be uprated by the statutory interest rate, and for which cash flows (if any) a discount rate can be chosen according to economic criteria.

9.140  Economic and finance theories have developed a range of principles on how to determine the discount rate. For future cash flows, it may be appropriate to use the claimant’s cost of capital as the discount rate. This takes into account the claimant’s time value of money and business risk—that is, the fact that future factual and counterfactual scenarios, and hence estimates of losses, are uncertain. Discounting expected future losses would provide an estimate of their value as at the award date. As regards uprating past losses, there are several possible approaches. One is to use the claimant’s cost of capital. The rationale is that during the period in which the damages were incurred, a claimant earning ‘normal’ returns would have expected to earn profits consistent with the cost of capital. Thus, damages uprated at the cost of capital would capture the expected return that the investors in the claimant company could have earned on the amounts lost, had these amounts been available for investment—i.e. it compensates investors for the use of their capital. An alternative approach is to apply the risk-free rate, which theoretically represents the guaranteed rate of return on an investment that has no risk. This is usually approximated by the rate of return on a virtually risk-free investment such as a government bond. The rationale for using the risk-free rate is that the repayment of damages is certain once awarded (subject to an inability to pay by the defendant), thus ensuring that the claimant is compensated for the time value of money without a risk premium within the interest rate. From an economic perspective, it may be appropriate to apply a risk-free rate to any damages from the point (p. 457) in time where the claimant has certainty that it will be awarded damages. This can be the court award itself, or an earlier point such as when the cartel is discovered and the claimant knows it will be able to prove harm. Without such certainty, applying a measure which captures risk (such as the cost of capital) may be more appropriate. This is still an area of debate among economists.

9.141  An example of a situation where a court preferred one discount rate over another is the Chicago Bulls franchise case that we saw before.69 The original claim uprated historical cash flows at the risk-free rate (in this case they were negative as the claimant had to commit additional equity to the business in the counterfactual scenario). The court agreed with most aspects of the damages valuation (including the comparators used), but not with the uprating approach. It held that historical cash flows (which in this case reflected equity contributions) needed to be uprated at the cost of equity capital to reflect the fact that the claimant would have incurred an opportunity cost of capital on the committed equity.

9.11.4  Economics and law on interest: Reconciling them ought to be more ‘simple’

9.142  Interest can be applied as simple interest or compound interest. With compounding, the calculation includes interest on accumulated interest from prior periods. For example, 10 per cent is applied to €100 in the first year, giving €110, and in the second year the 10 per cent is applied to that €110 from the first year, giving €121. From an economic perspective, compounding interest is the usual, and conceptually correct, approach to discounting. When you put your money in the bank, you expect interest to be paid on the whole balance, so on past interest too. And yet very often legal frameworks require simple interest to be applied. This means interest calculated solely as a percentage of the principal sum. So 10 per cent is applied to €100 in the first year, giving €110, and in the second year the 10 per cent is again applied to the €100, giving a total of €120. In this example the difference between the two methods is only €1. However, for longer time periods and higher interest rates the differences become substantially greater. Whether you apply simple or compound interest usually has a greater effect on the damages estimate than which particular rate you use.

9.143  In most jurisdictions simple interest is the norm. As economists we are somewhat puzzled by this. The Netherlands seems to be one of the few jurisdictions where the civil code prescribes compounding of interest.70 EU case law seems to emphasize simple interest but has also referred to compound interest. In a 2001 case the General Court stated that:

Regarding the rate of interest, it should be pointed out that, according to a principle generally accepted in the domestic law of the Member States, in an action for the recovery of a sum unduly paid based on the principle prohibiting unjust enrichment, the claimant is normally entitled to the lower of the two amounts corresponding to the enrichment and the loss. Furthermore, where the loss consists of the loss of use of a sum of money over a period of time, the amount recoverable is generally calculated by reference to the statutory or judicial rate of interest, without compounding.71

(p. 458) 9.144  However, the General Court also found that, in this particular case, the actual amount to be calculated would be better reflected by applying a compound interest rate, and therefore preferred the latter. A UK House of Lords ruling, Sempra Metals, contains a useful discussion of these points.72 It notes some comments made by legal representative bodies that ‘the obvious reason for awarding compound interest is that it reflects economic reality’, and that ‘computation of the time value of the enrichment on the basis of simple interest will inevitably fall short of its true value’. On the other hand, it noted that ‘The virtue of simple interest is its simplicity. That cannot be said of compound interest, which can be calculated in different ways leading to different results.’ (We do not necessarily follow this last statement; indeed, compound interest is easier to compute in spreadsheets than simple interest.)

9.145  Case law in the United Kingdom also seems to increasingly recognize that the use of the statutory interest rate is not necessarily aligned with business reality. In a case outside competition law, the High Court held that:

The Judgments Act [statutory] rate is fixed for the benefit of unpaid judgment creditors. It is not normally an appropriate rate of interest to award in the context of a dispute between two businesses … If Claymore or a company such as Claymore, had sought to borrow £750,000.00 over the period since June 2004, Claymore would have had to pay interest at more than 1% over base rate.73

9.146  Ultimately, there is plenty of scope for economics and law to come closer together in the area of interest and discounting in damages cases. Given the significant difference the application of interest can make to the damages estimate, this theme merits greater debate and reflection than it has received to date.

Footnotes:

1  Directive 2014/104/EU of the European Parliament and of the Council of the European Union of 26 November 2014 on certain rules governing actions for damages under national law for infringements of the competition law provisions of the Member States and of the European Union, OJ L 349/1, 5 December 2014, at [3].

2  Secretary of State for Health and others/Pinewood v Reckitt Benckiser, approved judgment, 4 December 2012, 2012 EWHC 3913 (Ch). We advised one of the parties in this case.

3  South-East Coal Co v Consolidation Coal Co 434 F 2d 767, 794 (6th Cir. 1970).

4  J Truett Payne Co v Chrysler Motors Corp 451 US 557, 565; 101 S Ct 1923; 68 L Ed 2d 442 (1981).

5  Directive 2014/104/EU of the European Parliament and of the Council of the European Union of 26 November 2014 on certain rules governing actions for damages under national law for infringements of the competition law provisions of the Member States and of the European Union, OJ L 349/1, 5 December 2014, at [17].

6  Fondiaria SAI SpA v Nigriello (Italian Supreme Court, 17 February 2007).

7  European Commission (2013), ‘Practical Guide on Quantifying Harm for Damages Based on Breaches of Article 101 or 102 of the Treaty on the Functioning of the European Union’, Staff working document, 11 June.

8  Office of Fair Trading (2006), ‘Independent schools agree settlement: competition investigation resolved’, press release, 19 May.

9  In 2015 the Austrian Supreme Court rejected a claim for €23 million by a lift installer against the members of a lifts and escalators cartel. The installer claimed that the cartel had led to fewer new lift installations and maintenance contracts. The court considered the claim to be insufficiently substantiated, and any harm to the installer from the output reduction to be too remote. Austrian Supreme Court, Case 4Ob95/15x Judgment of 16 June 2015.

10  Case C-557/12, Kone AG and Others, Judgment of 5 June 2014, at [29–30].

11  LG Dortmund AZ 13 0 55/02 Kart Vitaminkartell III [2004] (Dortmund Regional Court, 1 April 2004).

12  Oberlandesgericht Düsseldorf, Berliner Transportbeton I, KRB 2/05.

13  Audiencia Provincial de Valladolid, Sentencia num. 261/2009, Judgment of 9 October 2009.

14  Directive 2014/104/EU of the European Parliament and of the Council of the European Union of 26 November 2014 on certain rules governing actions for damages under national law for infringements of the competition law provisions of the Member States and of the European Union, OJ L 349/1, 5 December 2014, [recital 47]. The presumption itself is at [17].

15  Professor Connor has updated the study a number of times since. The main findings are not very different.

16  Hungarian Competition Act (Act LVII of 1996, as amended from time to time), [88/C].

17  Nederlandstalige Rechtbank van Koophandel, Brussel, Vonnis in de zaak van Europese Unie tegen Otis en anderen, A.R. A/08/06816, 24 November 2014, pp. 23 and 27.

18  Mors SA v Labinal SA, Cour d’Appel de Paris, 1ère chambre, section A, arrêt no 334, 30 September 1998.

19  INAZ Paghe srl v Associazione Nazionale dei Consulenti del Lavoro (Corte d’Appello di Milano, 10 December 2004).

20  Crehan v Inntrepreneur Pub Company (CPC) & Anor [2003] EWHC 1510 (Ch); [2004] EWCA Civ 637; [2006] UKHL 38; Case C-453/99 Courage Ltd v Bernard Crehan [2001] ECR I–6297.

21  2 Travel Group Plc (in Liquidation) v Cardiff City Transport Services Limited, [2012] CAT 19, 5 July 2012. We acted as experts for the defendant in this case.

22  Albion Water Limited v Dŵr Cymru Cyfyngedig, [2013] CAT 6, 28 March. We advised the defendants on this matter.

23  Ibid., at [69].

24  Ibid., at [71].

25  Conwood Co LP v US Tobacco Co 290 F 3d 768, 793 (6th Cir. 2002).

26  These are explained in econometrics textbooks, such as Wooldridge (2013).

27  Re Aluminum Phosphide Antitrust Litigation, 893 F Supp 1497 (D. Kan. 1995).

28  Stelwagon Mfg Co v Tarmac Roofing 63 F 3d 1267 (3d Cir. 1995).

29  Conwood Co v US Tobacco Co 290 F 3d 768 (6th Cir. 2002).

30  See, for example, Hendry and Clements (2004) and Timmerman (2006).

31  Tullis Russell Papermakers Limited v Inveresk Limited, 2010 CSOH 148, 10 November 2010. We acted for the claimant in this case.

32  Ibid., at [183].

33  Conwood Co v US Tobacco Co 290 F 3d 768 (6th Cir. 2002).

34  German Paper Wholesale Cartel (German Federal Court of Justice, 19 June 2007).

35  Fishman v Estate of Wirtz 594 F. Supp. 853 (ND Ill. 1984); and 807 F 2d 520 (7th Cir. 1986).

36  Conduit Europe SA v Telefó‎nica de Españ‎a SAU (Madrid Commercial Court, Judgment of 11 November 2005).

37  LePage’s Inc. v 3M Co., 324 F 3d (3d Cir. 2003).

38  Bundesarbeitskammer v Powerdrive Fahrschule Andritz GmbH, Landesgericht für Zivilrechtssachen Graz (Regional Civil Court of Graz), 17 August 2007.

39  Note that simple interpolation and ARIMA discussed above are also forms of forecasting (or backcasting), as they determine the counterfactual based on data from outside the infringement period, not during it.

40  In RE: TFT-LCD (Flat panel) antitrust litigation, Nokia Corporation and Nokia Inc v AU Optronics Corporation et al, Case No. C 09-5609 SI, Order granting in part defendants’ joint motion for partial summary judgment as to (1) claims based on inferred invoices; and (2) state law claims (ND Cal, 2012). We acted as experts for the claimants in this case.

41  See, for example, Krum et al. (1994), and Card and Krueger (1994).

42  SA Verimedia v SA Mediametrie, SA Secodip, GIE Audipub, Cour d’Appel de Versailles, 12ème chambre, section 2, arrêt no 319, 24 June 2004.

43  Telystem SpA v SIP SpA (now Telecom Italia SpA) (Milan Court of Appeal, December 1996).

44  Fishman v Estate of Wirtz 594 F. Supp. 853 (ND Ill. 1984); and 807 F 2d 520 (7th Cir. 1986).

45  GT Linien A/S (under bankruptcy—subsequently GT Link A/S) v De Danske Statsbaner DSB and Scandlines A/Sn (formerly DSB Rederi A/S) UFR 2005.2171H (Danish Supreme Court).

46  OLG Düsseldorf, Urteil vom 16.4.2008 VI-2 U(Kart) 8/06—Stadtwerk (Higher Regional Court of Düsseldorf).

47  Albion Water Limited v Dŵr Cymru Cyfyngedig, [2013] CAT 6, 28 March.

48  Ibid., at [77–8].

49  Concord Boat Corporation v Brunswick Corporation 207 F 3d 1039 (8th Cir. 2000), 24 March 2000.

50  German Paper Wholesale Cartel (German Federal Court of Justice, 19 June 2007).

51  2 Travel Group Plc (in Liquidation) v Cardiff City Transport Services Limited, [2012] CAT 19, 5 July 2012. We acted as experts for the defendant in this case.

52  Ibid., at [223].

53  The CAT also awarded £60,000 in exemplary damages, which is relatively rare. It emerged that Cardiff Bus had received legal advice at the time that its actions might be seen as anti-competitive, but knowingly disregarded this risk. The way the £60,000 was determined is not specified in the judgment, but it appears to be proportionate to the lost profit award.

54  Illinois Brick Co v Illinois 431 US 720 (1977); and Hanover Shoe Inc v United Shoe Machinery Corp 392 US 481 (1968).

55  The cost increase could of course lead to a reduction in downstream output and the exit of a number of downstream suppliers. This may give rise to a lost profit claim separate from the overcharge claim. However, it is usually harder to prove such lost profits, especially for suppliers who are no longer in the market.

56  Audiencia Provincial de Valladolid, Sentencia num. 261/2009, Judgment of 9 October 2009.

57  Tribunal Supremo, STS 5819/2013, Judgment of 7 November 2014.

58  Paris Commercial Court, Juva v Hoffmann La Roche, Decision of 26 January 2007.

59  See van Dijk and Verboven (2009).

61  See, for example, Bacon (1991).

62  Joined Cases C-192/95 to C-218/95 Société Comateb and others v Directeur Général des Douanes et Droits Indirects [1997] OJ C74/3.

63  Arkopharma v Roche and Hoffmann-La Roche, no RG2004F02643, Tribunal de Commerce de Nanterre (Nanterre Commercial Court, 11 May 2006).

64  Directive 2014/104/EU of the European Parliament and of the Council of the European Union of 26 November 2014 on certain rules governing actions for damages under national law for infringements of the competition law provisions of the Member States and of the European Union, OJ L 349/1, 5 December 2014, [recital 12].

65  Plaintiffs in the USA are entitled to treble damages. In Europe the payment of interest can also triple the original amount if the damage arose a long time ago.

66  National Grid Electricity Transmission Plc v ABB Ltd and Others, [2012] EWHC 869 (Ch). The case settled out of court in 2014 for an undisclosed amount. We acted as experts for National Grid on this matter.

67  For simplicity, this example assumes that the cash flows occur on 1 January of each year. Another assumption is that the interest is compounded—that is, the calculation includes interest on accumulated interest from prior periods (see below).

68  Bürgerliches Gesetzbuch, Section 288 Book 2; and Burgerlijk Wetboek, Article 120 Book 6.

69  Fishman v Estate of Wirtz 594 F. Supp. 853 (ND Ill. 1984); and 807 F 2d 520 (7th Cir. 1986).

70  Burgerlijk Wetboek, Artikel 119a Boek 6.

71  Case T-171/99 Corus UK Ltd v Commission [2002] OJ C3/23.

72  Sempra Metals Ltd v Revenue & Anor [2007] UKHL 34, 18 July 2007.

73  Claymore v Nautilus [2007] EWHC 805 (TCC).