Recent Developments at DG Competition: 2017/2018

The CET is a part of DG Competition. Its staff consists of 30 economists (mostly holding PhDs) with a mix of permanent and temporary positions. The CET is headed by the Chief Competition Economist, who is an external academic who is recruited for a three-year term.

The CET has both a support role and a scrutiny role. As part of its support role, the team provides guidance on methodological issues of economics and econometrics in the application of EU competition rules. It contributes to individual competition cases—in particular, the ones that involve complex economic issues and quantitative analysis—and to the development of general policy instruments, as well as assisting with cases that are pending before the Community Courts.

Members from the CET who are assigned to specific cases have a specific and independent status within case teams, and report directly to the Chief Competition Economist. As part of the scrutiny role, the Chief Competition Economist can report his opinion directly to the Director-General of DG Competition as well as to the Competition Commissioner, providing her with an independent opinion on the economic aspects of a case before she proposes a final decision to the European Commission.

The CET is active in DG Competition’s three main areas of policy: antitrust, merger control, and state aid. Historically, the CET’s main domain of activity is merger investigations—typically 50–60% of CET’s time—while CET’s work on state aid and antitrust is typically in a range of 20–25% of the CET’s time each.

The Decision identifies the following fundamental sources of Google’s dominance in the market for general search: the importance of scale, indirect network effects, and brand effects.

Scale is identified as an important factor that increases the barriers to entry of rival search engines. Search engines need to churn large volumes of queries in order quickly to update and refine the relevance of results. The larger is the number of queries that are submitted to and results that are clicked in a search engine, the faster the search engine learns to detect a change in users’ behaviour and adapts and improves the relevance of the results. This holds for both free ‘blue link’ results and for paid search ads.

The importance of scale is further amplified by the fact that queries do not all have the same frequencies of usage. As the Decision explains, there are common queries (“head” queries) and uncommon queries (“tail” queries). Tail queries make it more difficult for a smaller, rival search engine to provide relevant results. The poor performance for the tail queries affects negatively the user experience, and this reduces traffic for the smaller search engine in general. The Decision acknowledges and endorses the fact that a sufficiently large amount of data is a necessary input for the ability to compete in these markets.³³

Indirect network effects are an important characteristic of two-sided platforms, as is recognised in the economic literature.³⁴ Both general search services and comparison shopping services are two-sided platforms. The Decision relies on the economic intuition developed in the literature that there are positive feedback effects between the advertisers on one side of the platform and the users who request search services on the other side. In line with the results of this economic literature, the Decision finds that the higher is the number of users, the higher is the value for advertisers to post (and to bid for) their ads on the page that contains the results of a search.

As a result, for advertisers who typically are faced with fixed costs in developing their ad campaigns, there is little incentive to use other search engines given the expected smaller number of users who are attracted to rival general search services. Similarly, this dynamic is also present in the market for comparison shopping services, where the two groups of customers are potential online consumers and online retailers.

Brand awareness seems also to characterise the behaviour of users who consume general search services. Contrary to Google’s claim, the Decision finds that overall users are faithful to one search engine and that they do not tend to multi-home. The Decision shows evidence that users of general search services were found to use only Google for the majority of their queries, while only occasionally diverting to other search engines.

These effects make Google of significant commercial importance for users who look for or want to be directed to comparison shopping services. In other words, when shopping online and wanting to compare prices, users often search on Google and then, based on Google’s suggestion, to make a choice of where next to go. This finding is consistent with the fact that users typically have limited information and rely heavily on the suggestions of Google. There is evidence that users do not read search results in pages other than the first page and have a very high tendency to click on the results ranked in top slots in the page.³⁵

The Google Shopping decision is a leveraging abuse case. Google used its dominance in the market for general search services to give its comparison shopping service an artificial advantage and exclude competing comparison shopping services, which leads to anticompetitive foreclosure.

This approach is similar to that of the Guidance Paper where anticompetitive foreclosure is defined as “a situation where effective access of actual or potential competition to supplies or markets is hampered or eliminated as a result of the conduct of the dominant undertaking whereby the dominant undertaking is likely to be in a position to increase prices to the detriment of consumers.”³⁶

The anticompetitive foreclosure is established in the Decision by the likelihood of having negative consequences on competition. Ultimately, the foreclosure of competitors would also have negative consequences on users as it could increase advertising costs and thus eventually lead to higher product prices. What is more, from the outset, the reduction of choice could determine the inability of consumers to reach the most relevant shopping opportunity. Also, a concentrated market that results from the foreclosure could reduce the pace of innovation.

The finding of Google’s anticompetitive foreclosure in the market for comparison shopping services is also dependent on the Decision’s assessment of the role of “merchant platforms” (e.g., Amazon, eBay). The Decision reaches two conclusions on merchant platforms. First, it concludes that comparison shopping services and merchant platforms are in different product markets. They serve different purposes for customers, and merchant platforms have business models that are closer to online retailers.³⁷ This is also complemented by the evidence that merchant platforms and comparison shopping services serve different groups of online merchants.

Second, it explains that, even if the merchant platforms were to be included in the analysis of effects, the importance of comparison shopping services would still be relevant and thus the coverage of Google’s anticompetitive foreclosure would still be non-negligible.

Google has contested these two findings of the Commission. Google states that, on the one hand, Google is not the only way that users have to reach comparison shopping services and online retailers; and, on the other hand, that alternative mechanisms, such as merchant platforms—including Amazon and eBay—are driving competition in this (broader) sector of directing and enabling users to shop online.³⁸

While the Decision does not discuss directly Google’s incentive to foreclose, the structure of Google’s well-known business model—which is also described in the Decision—contains an inherent monetisation problem and allows a detailed discussion of foreclosure incentives.³⁹

The strategic decision (and commitment) of Google to distribute free general search services is a core element for Google’s incentive to foreclose in adjacent markets, such as the comparison shopping services market. Only clicks on Google’s ads or Google’s comparison shopping service trigger a payment; clicks on the general search results do not. The free Google general search results generate traffic for a wide range of business that offer paid products and services. However, Google cannot directly monetise this traffic.

This feature of the general search services market makes the classical Chicago School critique not applicable. The Chicago School critique argues that a dominant company should have no incentive to foreclose competitors from adjacent markets because the dominant company could make as much profit (if not more) by appropriately pricing its dominant product/service and extracting all the value because of the “one monopoly profit” theory. Google’s strategic decision not to price general search services puts Google in a situation where the only way for Google to appropriate the value that is created by users’ clicks that lead to rival comparison shopping platforms is to appropriate the clicks of those users by directing them to its own comparison shopping service. Sending users to competing comparison shopping services would amount to a net profit loss and, over time, to an increasing risk of users’ bypassing Google and going directly to the competing comparison shopping services.

The static loss of profits that are derived from users who are directed to competing comparison shopping services provides incentives for Google to foreclose competitors in the comparison shopping services market. This static incentive to foreclose is also complemented by the risk of consumers bypassing Google over time if rival comparison shopping services become too successful. The static incentive is thus also reinforced by the dynamic risk for Google to lose its gatekeeper role for consumers who look for (or want to be directed to) comparison shopping services.

In order to demonstrate Google’s potential to foreclose competitors, the Decision presents several pieces of empirical evidence. First, it shows that the introduction of specific algorithm changes (in particular the launch of the “Panda” algorithm) led to the demotions of competitors’ ranking.⁴⁰ The Sistrix Visibility Index⁴¹ indicates a substantial decrease in competitors’ visibility and therefore a deterioration of display after the introduction of the Panda algorithm. The Decision also presents evidence that this reduction in visibility and ranking reduces traffic. Data on clicks by search result rank and positioning show that there is a clear link between visibility, the lay-out of Google’s general search results pages and click-through behaviour: results that are displayed higher and in a more visible format attract significantly more clicks than those that are displayed lower or beyond the first page.

Second, the Decision shows that the evolution of the visibility index of competitors was indeed following a similar trend to the generic search traffic of competitors. Moreover, the non-confidential graphs in the Decision show a trend break in the traffic (clicks) to competitors: a stable increase turned into a stable decrease, although the extent of this break varies across countries. The Decision offers documentary evidence that indicates that comparison shopping services suffered a short-term and long-term traffic decreases due to the introduction of systematic demotions.

Third, beyond the impact on traffic of demotions, the Decision assessed an experiment by Google; this assessment addressed the impact on traffic from the rich and prominent display of Google’s own comparison service. In this “ablation” experiment, Google removed the Shopping Unit from its general result pages for a randomly selected group of users. Comparing the click-through rate in this selected group to the click-through rate of the unchanged display allowed an estimation of the causal effect of the more prominent placement of the Shopping Unit with respect to competitors’ traffic. The Decision concluded that the experimental evidence points to an effect on competitors’ traffic. The Decision illustrates this impact on both the generic and text ad (AdWords) traffic of competitors. The Shopping Unit’s effect on text ad traffic (AdWords) suggests that competitors found it difficult to replace lost free generic search traffic by paid text ad traffic.

Finally, in the same period of time, based on the same two metrics (visibility and clicks), Google’s comparison shopping service experienced a sharp increase in its traffic on a lasting basis. Thus, the Decision reaches the conclusion that Google’s conduct had the effect of promoting Google’s comparison shopping service—which was not a successful service prior to the conduct.⁴²

Ready made abuse categories might not fit new and quickly changing markets. Still, the Google Shopping case demonstrates that the antitrust framework of Article 102 of the TFEU can accommodate a foreclosure theory of harm based on the well established and understood notion of leveraging. The Decision relies on detailed and comprehensive economic evidence which underpins such a foreclosure theory: the evidence on the search habits of users; the visibility and traffic data trends; the correlation of click through rates and ranks; and the critique of Google’s ablation experiment.

Price discrimination in the context of state aid control can distort competition. In state aid cases price discrimination typically results from the intervention of the State, for example by applying different tax rates to various groups of companies. The core economic question in such cases is whether such price discrimination can be justified by economic reasons.

Varian (1989) offers an extensive definition of price discrimination: “The conventional definition is that price discrimination is present when the same commodity is sold at different prices to different consumers. However, this definition fails on two counts: different prices charged to different consumers could simply reflect transportation costs, or similar costs of selling the good; and price discrimination could be present even when all consumers are charged the same price - consider the case of a uniform delivered price. We prefer Stigler’s (1987) definition: price discrimination is present when two or more similar goods are sold at prices that are in different ratios to marginal costs.”⁴⁶ Economic literature thus links price discrimination to price differences that cannot be attributed to differences in costs. If serving different users or groups of users causes different costs, it may be economically reasonable to apply different prices to these users, to reflect marginal cost differences. In such case, differential pricing does not constitute price discrimination in the economic sense.

Beyond differences in marginal costs of serving different users, variations in demand sensitivity may also justify differential prices. In regulated industries—such as electricity or telecommunications—where high fixed costs need to be recovered, Ramsey pricing is typically considered to be an efficient pricing strategy. With Ramsey pricing—a form of third-degree price discrimination—groups of users are charged prices that are inversely related to their demand elasticity, so that price-sensitive customers pay lower prices (and vice versa). Such pricing enables utilities to recover fixed costs while minimizing consumption distortions.

Hence, preferential pricing can constitute price discrimination, though not necessarily, and price discrimination is not necessarily distortive. These two economic principles appear to underpin the application of state aid rules to differential pricing.

The principles described above were applied in two recent cases where the Commission investigations focused on the economic rationales that motivated differential tariffs.

As is illustrated by the waste water case in Denmark and the network charges case in Germany, economic principles underpin the assessment of state measures that potentially involving illegal subsidies in Europe.

In order to determine whether differential pricing is at all covered by state aid rules, it has to be assessed whether lower prices that are paid to public undertakings (or reduced through public intervention) can be justified by lower costs of production. Whilst full cost coverage is not necessary, pricing needs to be related to the costs of serving specific firms or groups of firms (e.g., at least variable costs plus a portion of fixed costs or incremental costs). Demand-side considerations and optimal fixed cost recovery are equally relevant. Differential pricing may constitute illegal state aid if it leads to price discrimination that cannot be justified by commercial strategies that are consistent with optimal pricing decisions and if it does not generate economic benefits outweighing competition distortions.

For state aid control, the concern is that specific firms or groups of firms benefit—through public funds—from preferential tariffs relative to their competitors. More favourable prices that are not justified either by cost differences or by efficient pricing strategies may be distortive if they artificially reduce the price of inputs for beneficiaries of aid (or increase relative input prices for rivals). This could lead to competition distortions unless justified by other objectives with positive welfare effects (e.g., regional development, or social objectives).

In 2017, three transactions were notified to the Commission that raised complex questions about the nature of the competitive counterfactual:

In Knorr-Bremse/Haldex, Knorr-Bremse’s last minute offer for Haldex had outbid a prior takeover offer by automotive supplier ZF Friedrichshafen, which had already secured antitrust clearance from all relevant authorities. ZF had been keen to acquire Haldex’s braking expertise to complement its own autonomous driving technology. It had announced the intention to make significant investments to expand Haldex’s activities, effectively trying to turn it from a niche operator into a full-service provider that would compete more effectively with market leaders Knorr-Bremse and Wabco.⁶⁶ Eventually, that transaction did not take place, since Knorr-Bremse pre-empted it through a higher bid.

From an economic perspective, the Commission therefore had to assess whether Knorr-Bremse’s takeover proposal had the object or effect of blocking potentially pro-competitive entry by ZF into the brake market. More specifically, it had to consider whether the correct counterfactual to the transaction was the status quo ante or the counterfactual of a takeover of Haldex by ZF Friedrichshafen. Accordingly, a key element of the assessment was the content of Knorr-Bremse’s internal documents and business plans, to obtain a realistic perspective of expected industry evolutions, commercial motivations, and likely competitive consequences.

Yet, for a substantial part of Air Berlin’s former operations (specifically, the Austrian subsidiary NIKI), the Commission contested the view that these assets would not have been commercially viable absent a takeover by Lufthansa. In particular, several other carriers had made offers in the public takeover process for NIKI, but those bids had been discarded by the insolvency administrator in favor of initiating exclusive negotiations with Lufthansa.⁶⁷ As in Knorr-Bremse/Haldex, the Commission therefore had to determine whether an alternative takeover of the target company was a more appropriate counterfactual against which the proposed transaction should be assessed. After all, it would prima facie not appear implausible that a strong incumbent airline might have an incentive to block entry of a competitor by acquiring a struggling rival instead of allowing that competitor to acquire the struggling firm.

Finally, the Italian steel maker Ilva, which runs the largest integrated steel mill in Europe, had entered into insolvency proceedings in 2015 and was being sold through a competitive tender. In ArcelorMittal/Ilva the Commission therefore had to assess whether or not there would have been a viable alternative purchaser for Ilva in the absence of the acquisition by ArcelorMittal. In the case at hand, a parallel state aid investigation had found that there were indeed such viable alternatives.⁶⁸ Even so, the case involved thorny questions on the degree to which these alternatives would have materialized and would have been comparable, since they involved different business models and planned approaches toward a necessary restructuring.

There is a widespread (if inaccurate) perception that is sometimes also promoted by merging parties that merger control intrinsically involves a comparison of pre- and post-merger situations. In reality, merger assessments always compare the expected future situation with and without a merger.⁶⁹ Merger control, thus, is generally forward-looking, as it compares two unknown future outcomes.

Within this framework, the pre-merger situation is of importance, in so far as it is an informative predictor of how the market would look in the absence of the proposed transaction. Very often, this is a reasonable assumption, which makes the status quo a sensible counterfactual.

In other situations, however—particularly in markets that are undergoing material changes—it is not reasonable to assume that the future will look like the past.⁷⁰ Possible reasons for a foreseeable divergence of future market conditions from the status quo—and, thus, for employing a non-standard counterfactual—can for instance include the following: a structural decline of demand, which will lead to overcapacities in the future; a failing firm defense; the impending entry or exit of a significant potential rival; or the likely ascent of a smaller competitor that will grow into an important competitive force.

These examples show that non-standard counterfactuals can make mergers either more or less problematic than a comparison to the status quo ante. E.g., in Knorr-Bremse/Haldex, the existence of an alternative bidder with complementary assets and fewer competitive overlaps made the proposed transactions more problematic than would otherwise have been the case. Conversely, in the non-NIKI part of Lufthansa/Air Berlin the diminishing standalone viability of the respective targets made the proposed transaction less problematic than would otherwise have been the case. Finally, in ArcelorMittal/Ilva and in the NIKI part of Lufthansa/Air Berlin, the Commission maintained the pre-merger situation as the relevant counterfactual and rejected the notion that the targets would continue to decline absent the transaction because of the existence of alternative viable purchasers at the time of the tender awards.

While merger counterfactuals may deviate substantially from the pre-merger situation, it is also clear that such scenarios cannot simply be conjured based on speculative possibilities. Instead, credible evidence has to be produced if either the Commission or merging parties propose to deviate from the pre-merger situation as the appropriate counterfactual.

Arguably, there is a sliding scale in this regard that depends on the degree of deviation from the pre-merger scenario.⁷¹ This precludes the Commission from judging a proposed merger against arbitrary alternative combinations that involve one of the parties with potentially better competitive outcomes. When reliable evidence exists that a specific alternative transaction is the most plausible alternative course of events, however, it would make little sense to evaluate the proposed merger against a status quo ante that may no longer be relevant for the future or against other counterfactuals that ignore the existence of the alternative transaction. For instance, in a previous case from 2009 (Lufthansa/Austrian Airlines), the Commission found that—absent the proposed transaction—Austrian would likely have been acquired by Air France-KLM, which made the latter a reasonable counterfactual.⁷²

An important question when applying non-standard counterfactuals, however, is what precisely is meant by considering market outcomes “absent the proposed transaction”. In particular, it can make a large difference whether one means “absent the merger proposal” (the ex-ante perspective) or “absent the clearance and implementation of the proposed merger” (the ex-post perspective). For instance, in all three of the recent cases the issue arose whether the rival bidders for the respective targets would still have an interest in acquiring them at the time of the assessment by the Commission, since their bids formally proved interest only at the time of their respective offers (which was several months prior to the eventual merger notifications).

A potential divergence between the ex-ante and ex-post approach towards counterfactuals can pose complex trade-offs between deterrence and efficiency considerations. For instance, consider the situation where the interest of an alternative purchaser has become less certain ex-post due to a deterioration of the ailing target’s commercial situation. Pursuing the ex-post approach may then imply having to permit acquisitions that have a negative effect on competition relative to the situation where that transaction had not been proposed. This may go as far as having to clear transactions with demonstrably exclusionary intent, whose purpose is to prevent other purchasers from entering the market.⁷³ Conversely, pursuing the ex-ante approach may imply having to block a proposed merger even though the applied counterfactual will not in fact materialize.

In our view, the ex-ante approach is a more sensible policy approach, since it is directed at preventing anticompetitive outcomes at their root. Arguably, it also comes closest in spirit to the stated aim of comparing “the competitive conditions that would result from the notified merger with the conditions that would have prevailed without the merger.”⁷⁴

It is important to realize that the ex-ante approach does not ignore observable developments post-announcement of the transaction. Rather, post-notification developments that are unrelated to the merger are taken into account in both the merger scenario and the counterfactual. However, post-notification developments that are causally related to the transaction can only be attributed to the merger scenario. For instance, in GE/Alstom, the Commission found that a “recent deterioration of Alstom’s financial situation in so far as it would not have occurred in the absence of the proposed merger cannot be taken into account”.⁷⁵ The post-notification decline of the target in that case was therefore causally attributed to the merger.

Eventually, none of the proposed transactions proceeded as originally notified by the parties. Both Knorr-Bremse and Lufthansa ultimately decided to withdraw their respective takeover bids for Haldex and NIKI.⁷⁶ ArcelorMittal/Ilva, on the other hand, was cleared in the second phase, subject to extensive divestments that consisted of integrated steelworks and finishing lines in five different EEA countries (Italy, Romania, the Czech Republic, Belgium, and Luxembourg).

Conversely, the withdrawal of Knorr-Bremse/Haldex did not lead the alternative bidder ZF Friedrichshafen to seize the opportunity and proceed with its earlier intention of taking over Haldex. It has been reported that the main reason for this change of heart was that ZF’s supervisory board became increasingly concerned about management’s costly M&A aspirations (which included an abandoned takeover attempt of Knorr-Bremse’s other competitor, Wabco, subsequent to the failed Haldex bid).⁷⁷

This divergence of ex-ante and ex-post outcomes illustrates the potential challenges that are associated with non-standard counterfactuals. First, considerations of ex-ante deterrence do not necessarily coincide with ex-post efficiency in terms of competitive outcomes. Nevertheless, as discussed above, we believe that prioritizing incentive effects through an ex-ante approach is a more sensible policy.

Second, if the intent of a merger proposal is to block the entry of a rival, the purchaser may succeed with its aim even if the acquisition proposal is ultimately prohibited as alternative bidders may have lost its interest in the target during the merger proceedings. In practice, the latter case is not an unlikely outcome, since the length of takeover proceedings may well imply further deterioration of the commercial viability of ailing targets. In certain cases, deterrence of anticompetitive takeover attempts may therefore require considering additional antitrust instruments (such as Article 102) to be fully incentive-compatible.