I INTRODUCTION

Consider a typical cartel in an intermediate goods market, say the market for widgets. Widgets are sold to manufacturers of blodgets, which then sell them to end-consumers. The goal of the widget cartel was to raise prices. So, if successful, the cartel would have caused widget prices, and most likely blodgets prices, to increase hurting blodgets manufacturers (i.e., the direct customers of widgets) and end-consumers of blodgets (i.e., the indirect customers of the widget cartelists). Both the direct and indirect customers of a cartelised product can claim compensation for damages in the European Union.

In this chapter, we discuss the methods commonly used to estimate damages in the Member States of the European Union. We first explain the effects of a cartel on the profits of direct customers as well as on the surplus of end-consumers. We then discuss the different economic methods used to estimate those effects. We conclude with a few practical considerations.

II CARTEL EFFECTS

The damages caused by the widget cartel to a blodgets manufacturer will equal the loss of profits suffered by that direct customer as a result of the cartel. A cartel may have three effects on the profits of a direct customer: (1) the price paid by the customer may increase (the ‘price overcharge’); (2) the retail price which the direct customer charges to the final consumers of blodgets may also increase (the ‘pass-on effect’); and (3) sales volumes may fall (the ‘output effect’). The widget cartel will hurt the end-consumers of blodgets only if the pass-on effect is positive (i.e., only if the retail price of the blodgets increases in response to the higher widget price set out by the cartelists).

The magnitude of the damage inflicted on both direct and indirect customers can be computed as follows.

Absent the cartel, a blodgets manufacturer pays a price c to the widget supplier and sells q units at a retail price p to the consumers of blodgets. The blodgets manufacturer thus earns a profit equal to (p - c) x q. In the cartel scenario, the blodgets manufacturer pays an input price c’ to the widget supplier and sells q’ units at a retail price p’. The profits under the cartel equal (p’ - c’) x q’. The change in profits for the blodgets manufacturer caused by the cartel can be decomposed in three terms. First, α = (c’ - c) x q’ is the reduction in profits stemming from the increase in the price of widgets caused by the cartel. Second, β = p x (q’ - q) is the reduction in profits resulting from the loss of volume caused by the increase in the retail blodget price. Third, γ = (p’ - p) x q’ represents the additional profits accruing to the blodgets manufacturer as a result of the higher retail blodget price. Therefore, the loss in profit to the blodget manufacturer resulting from the widgets cartel equals the loss of profits α + β minus the additional profits γ.

Quantifying the damages suffered by a blodgets manufacturer thus requires comparing widget prices with and without the cartel (c and c’, respectively), blodget prices with and without the cartel (p and p’, respectively), and blodget sales with and without the cartel (q and q’, respectively). We can observe widget prices, blodget prices and volumes during the cartel, but can only estimate these magnitudes in the absence of the cartel. The main challenge in the quantification of cartel damages is indeed determining the likely evolution of the market absent the infringement (i.e., absent the cartel). This hypothetical scenario, commonly known as the counterfactual or but-for scenario, is not observable and, therefore, must be approximated, possibly using econometric tools. This is, needless to say, a complex task, since it requires simulating an economic scenario that did not occur.

III ESTIMATION METHODS FOR THE PRICE OVERCHARGE

In 2013 the European Commission published a Practical Guide offering assistance to courts and parties involved in actions for damages on how to quantify the harm caused by antitrust infringements, including price-setting cartels.2 The European Commission is also preparing additional guidance for national courts on the quantification of the passing-on effect and has published an expert study on this topic.3

The Commission’s Practical Guide sets out a number of methods and techniques to estimate the value of the economic variable or variables of interest in the non-infringement or counterfactual scenario. In a price-setting cartel the main economic variable of interest is the price that would have been paid by the cartelists’ customers absent the cartel: the widget price. Estimating this price for the counterfactual scenario and comparing the counterfactual price to the observed price provides us with an estimate of the price overcharge.4

i Comparator methods

The discussion in the Commission’s Practical Guide focuses in particular on the methods most widely used by parties and courts in EU antitrust damages cases: the so-called comparator methods, which in a cartel case estimate the (widget) prices that would have been observed in the market absent the infringement by reference to (1) the (widget) prices observed at time periods before or after the infringement, (2) the prices observed in other geographic (widget) markets which were unaffected by the infringement, or (3) the prices of other product (non-widget) markets which bear a relationship with the (widget) markets affected by the cartel but which were effectively competitive.

Comparisons over time

This method requires data to be available both during the cartel period and outside the cartel period. It also requires information about the beginning and end of the cartel period. As the Commission’s Practical Guide notes:

An advantage of all methods comparing, over time, data from the same geographic and product market is that market characteristics such as the degree of competition, market structure, costs and demand characteristics may be more comparable than in a comparison with different product or geographic markets.5

However, this should not be interpreted as saying that a simple comparison of the prices of widgets before, during and after the cartel will provide an accurate estimate of the price effect of the cartel: c’ – c. If important market factors, such as the level of demand or costs, change over time, a simple time comparison may be misleading and should be treated with caution.6 Suppose, for example, that the cost of producing widgets was higher during the cartel period than before or after the cartel period. In that case, we should expect widget prices to be higher during the cartel period even in the absence of the cartel and, therefore, a simple comparison of widget prices over time would produce a biased estimate of the cartel overcharge.

Unless all factors that may have influenced prices during the cartel period other than the cartel itself (the so-called ‘confounding factors’) are appropriately taken into account, the use of time comparison methods could overestimate or underestimate the price effect of the cartel. Econometric techniques, such as regression analysis, are typically used to account for those confounding factors, such as changes in fuel prices and/or the economic cycle, when conducting these comparisons.7

Comparisons across different geographic markets

A second common approach to estimate the price overcharge is to use data from a different geographic market for the same product (e.g., widgets) during the period of the infringement to compare the (widget) prices observed during the infringement period in the relevant geographic market with the corresponding prices during the same period but in the comparator market.8 The (widget) prices in the comparator market provide an estimate of the counterfactual prices (i.e., the (widget) prices that would have obtained absent the cartel).

For this approach to be valid, the comparator market would need to be one where the infringement had no effect. The Commission’s Practical Guide states that:9

The more a geographic market is similar (except for the infringement effects) to the market affected by the infringement, the more it is likely to be suitable as a comparator market.

This approach is not used very often because it is generally difficult to find a comparator market that is (1) sufficiently similar to the market affected by the infringement and (2) not affected, directly or indirectly, by the infringement. In addition, even if a comparator market can be found, the comparison needs to take into account possible differences across markets that may explain a price differential such as disparities in labour costs and income levels.

Comparisons across different product markets

A third approach to estimate the price overcharge is to compare the (widget) price during the cartel period with the price of a different, yet related, product market that is not affected by the infringement.10 In this approach, the price for the comparator product is used to estimate the counterfactual price. The main challenge in using this approach is to identify a product market that is sufficiently similar to the market in which the infringement took place but was not itself affected by it. As the Commission’s Practical Guide states:11

[…] the comparator product should be carefully chosen with a view to the nature of the products compared, the way they are traded and the characteristics of the market e.g. in terms of number of competitors, their cost structure and the buying power of customers.

ii Other methods

Other methods discussed in the Practical Guide are simulation models and cost-based models. Simulation models draw on economic theory to simulate market outcomes with and without the infringement, while the cost-based models use accounting information on production costs and a reasonable estimate of the profit margin to estimate the prices in the hypothetical non-infringement scenario.12 As the Commission’s Practical Guide explains, no single approach is always better than the others so the choice of approach must vary from case to case.

Simulation models

Economic models can be constructed that simulate prices, output, costs and profits under alternative market scenarios.13 For example, in a cartel case they can be used to simulate market outcomes in a counterfactual scenario where competitors set prices non-cooperatively.

Such models are adjusted (or ‘calibrated’) so that they produce the price levels actually observed during the cartel period. Having calibrated the model in this way, one can then change the relevant behavioural assumptions and assume that there was no cartel to simulate the prices that would have prevailed absent the cartel.

One difficulty with this approach is ensuring that the model assumptions and model outcomes accurately reflect the real world, so that it can be trusted to provide a reliable estimate of the counterfactual prices. This is extremely complex and may prove practically impossible. It is difficult to determine, for example, how the nature of the competitive interaction between firms would have been absent the infringement. In addition, as the Commission’s Practical Guide explains:14

[…] the development of complex simulation models can be technically demanding and may require significant amounts of data that may not always be accessible to the party concerned or possible to be estimated with sufficient reliability.

Simulation models may be used when the data is insufficient to implement a comparator approach (e.g., because there is no reliable data outside the infringement period), however, in general they are less preferred to the simple comparator models described above. In the few cases where we are aware that simulation models have been used, they were used in conjunction with a comparator approach.

Cost-based assessments

The cost-based approach relies on the assumption that under competitive conditions (i.e., absent the cartel), prices would have equalled costs plus a reasonable mark-up.15 Under this assumption, which presumes heroically that the market would have been perfectly competitive absent the cartel, counterfactual (widget) prices could be estimated by applying an appropriate mark-up to the costs of producing widgets.

Implementing this approach requires information on production costs plus an estimate of the ‘mark-up’ that allows for a reasonable profit in the absence of the infringement. The relevant cost measures depend on the industry and the time horizon under consideration. The cost of production is combined with assumptions on counterfactual margins to obtain a counterfactual price.

This approach raises a number of potentially insurmountable challenges. For example, identifying the relevant cost when the infringement relates to one product of a multi-product firm is a difficult task. Estimating the profit margin in the counterfactual is even more complex. Both these tasks may require making assumptions, which are not always easy to test and verify. For example, one needs to make assumptions regarding the level of competition in the absence of the infringement, and hence on the degree of product differentiation, the existence of capacity constraints and the cost structure of producers absent the infringement. This may explain why these methods are less commonly used that the comparator methods described above.

IV ESTIMATING THE PASS-ON EFFECT

As we explained above, direct customers of the cartelists paying an overcharge for the products they use as input in their own production of other goods or services may decide to raise the prices for their own goods or services, thereby passing on some or the entire initial overcharge to their own customers.

The decision to pass on all or part of an input cost increase normally entails a negative volume effect. In other words, an increase in the retail blodget price will reduce blodget sales. The reduction in the volume of sales will cause a loss of profit. The pass-on and volume effects are, therefore, two sides of the same coin. Not surprisingly, their magnitudes are determined by the same factors: most importantly, the elasticity of end-consumer demand.16

Pass-on rates can vary between zero and 100 per cent and, in some cases, can even exceed 100 per cent. As the Commission’s Practical Guide points out:17

It is not possible to establish a typical pass-on rate that would apply in most situations. Rather, careful examination of all the characteristics of the market in question will be necessary to assess pass-on rates. In a specific case, the existence and degree of pass-on is determined by a range of different criteria and can therefore only be assessed having regard to the conditions of the market in question.

The extent to which the direct customers of a cartel can pass on the price overcharge to their own consumers depends on a series of factors which include (1) the size of the affected market; (2) the intensity of competition in the downstream market where they compete for the demand of the end consumers; (3) the sensitivity of final demand to prices; (4) the variation of marginal cost with output changes; (5) the importance of the cartelised input in their total costs; and (6) the duration of the infringement and the frequency of business exchanges.18

Customers of cartelised products will not be able to fully pass on a cartel price overcharge to their customers if most of their competitors in the downstream market purchase their inputs outside of the cartel. On the contrary, if all competitors in the downstream market purchase their inputs from the cartel and are therefore similarly exposed to the overcharge, then the pass-on rate is likely to be high.

The intensity of competition in the downstream market also affects the pass-on rate. If the downstream market is perfectly competitive and the cartel affected most players in the market, the pass-on rate will tend to be equal to 100 per cent. If the downstream market is less than perfectly competitive, then it is likely that the direct customers of the cartelised product will pass on at least part of the overcharge, though often significantly less than 100 per cent. The sensitivity of end consumers to price changes (i.e., the price elasticity of end-consumer demand) will also influence the pass-on rate. The pass-on rate will generally be higher when end consumers are not highly price sensitive, that is, if demand is inelastic.

A substantial pass-on is less likely if marginal costs significantly decrease following a reduction in output, because the lower output would become less costly to produce (e.g., in the presence of capacity constraints). Conversely, a substantial pass-on is more likely if marginal costs do not significantly decrease following a reduction in output (e.g., due to the absence of capacity constraints). When the cartelised inputs are essential and are a large part of the total cost of the product, the pass-on rate is likely to be higher because failure to pass on input price increases would have a marked impact on profitability. In addition, if the cartel has an impact on variable costs, then pass-on is more likely than if the impact is on fixed costs.

Finally, where infringements last for a long time, it is more likely that pass-on occurs because companies will be unable to sustain the negative impact on profitability indefinitely. The same applies to sectors where business exchanges and price adjustments are frequent.

V PRACTICAL CONSIDERATIONS

For the reasons explained above, given the data typically available in cartel cases, the price overcharge is commonly estimated by comparing prices during the infringement period with prices before and/or after the infringement period, taking into account that other relevant factors may have influenced prices during these periods. In our professional experience, there seems to be a wide consensus among plaintiffs’ and defendants’ economists in the European Union about the superiority of comparator methods to estimate the price overcharge. While we have witnessed many disagreements about the precise implementation of those methods in actual cases, we have not been involved in any case where one of the experts relied exclusively on a simulation model and/or a cost-based approach.

Similarly, most plaintiffs’ and defendants’ economists agree that there is no typical overcharge. The evidence reported by Connor and others show that estimated cartel overcharges vary enormously, from zero to 100 per cent.19 This wide dispersion is explained, inter alia, by the fact that the effect of a cartel on prices depends on numerous factors including the characteristics of the products involved, the degree of buyer power, and the economic context under which the cartel operates. All these factors need to be addressed on a case-by-case basis when estimating the effects of the cartel, otherwise overcharge estimates will under- or overestimate the effect on prices.

There is also no controversy regarding the right analytical approach for the estimation of the pass-on and volume effects. And yet, in practice, the assessment of these effects is typically conducted using theoretical arguments and/or relying on anecdotal evidence or back-of-the-envelope calculations. The reason for this is that plaintiffs rarely conduct these studies, relying on the presumption of zero pass-on that is common in the European Union, and defendants usually find themselves unable to obtain the data needed to perform these estimations themselves in order to rebut that presumption.


1 Enrique Andreu, Jorge Padilla, Nadine Watson and Elena Zoido are economists at Compass Lexecon.

2 European Commission (2013), Practical Guide: Quantifying Harm in Actions for Damages Based on Breaches of Article 101 or 102 of the Treaty on the Functioning of the European Union, 11 June 2013, SWD (2013) 205.

3 RBB and Cuatrecasas, Gonçalves Pereira (2016), Study on the Passing-on of Overcharges, Study prepared for the European Commission, available at http://ec.europa.eu/competition/publications/reports/KD0216916ENN.pdf.

4 An estimate of the overcharge can also be obtained from the comparison of observed and counterfactual widget profits although, as we explain below, the comparison of profits may in certain industries be more complex than the comparison of prices.

5 Commission’s Practical Guide, para. 41.

6 Commission’s Practical Guide, para. 42.

7 Regression analysis is a standard statistical tool used to test and estimate economic relationships: for example, the relationship between prices and costs.

8 Commission’s Practical Guide, para. 49.

9 Commission’s Practical Guide, para. 50.

10 Commission’s Practical Guide, para. 54.

11 Commission’s Practical Guide, para. 55.

12 Commission’s Practical Guide, para. 28.

13 Commission’s Practical Guide, para. 97.

14 Commission’s Practical Guide, para. 104.

15 Commission’s Practical Guide, para. 107.

16 The elasticity of end-consumer demand which determines the profit maximising response (i.e., the price and corresponding sales volume) of the direct purchaser to an input cost increase.

17 Commission’s Practical Guide, para. 168.

18 Commission’s Practical Guide, paras. 169–171.

19 Connor, J M and Y Bolotova (2006), ‘Cartel Overcharges: Survey and Meta-Analysis’, International Journal of Industrial Organization, 25(6), pp. 1109–1137; Connor, J M (2007), ‘Price-Fixing Overcharges: Legal and Economic Evidence’, in: J B Kirkwood (ed.), Research in Law and Economics, vol. 23, Elsevier, Amsterdam, pp. 59–154; and Bolotova, Y (2009), ‘Cartel Overcharges: An Empirical Analysis’, Journal of Economic Behaviour & Organization, 70(12), pp. 321–341; Boyer, M and R Kotchoni (2015), ‘How Much Do Cartel Overcharge?’, Review of Industrial Organization, 47(2), pp. 119–153.