Contributions to Competition Economics: Introduction
This Feature relates to the economics of competition policy, and covers a wide range of topics within that including merger policy, cartels, re-sale price maintenance (RPM) and optimal fining policy.
It is broadly accepted that, when enforced effectively, competition policy has an important role to play in ensuring that economies are characterised by competitive markets, which deliver innovation, productivity growth and benefits for consumers. Effective competition enforcement is not, however, always straightforward, and many difficult and fascinating questions arise. Academic research is crucial for the refinement and improvement of competition policy, and the articles in this feature collectively mark a significant further contribution, with relevance to a number of key active debates in the area.
1. Regulation of Mergers
There are two articles in the Feature examining the application of merger policy. The article by Affeldt et al. (2013) builds on a new methodology for merger analysis, upward pricing pressure (UPP). The UPP methodology, which is summarised in Farrell and Shapiro (2010), has become influential as a screening device for identifying potentially anti-competitive mergers in recent years, alongside its sister concept the Gross Upward Pricing Pressure Index (GUPPI), also covered in the article.
The primary rationale behind the use of this new methodology is that it makes good sense for merger analysis to focus on the key likely anti-competitive effect of a merger – the unilateral change in incentives that arise from merging two firms – and that UPP/GUPPI can provide useful indicator of this. This contrasts with the more traditional assessment of mergers, which at its worst involved a mechanistic process of market definition and calculation of market shares. The new approach has been codified in the 2010 US and UK merger guidelines and applied in a variety of cases in the US and UK, and has more recently in the EU. Proponents of UPP/GUPPI consider it especially useful as an initial indicator of whether a merger might raise competition concerns, in what is known as Phase I of the two-phase merger assessment process.
One of the important lessons learned from applying these new tools in casework is that the UPP/GUPPI formulae should not be applied mechanistically. These tools attempt to measure the change in pricing incentive that will result from a merger; these will depend on the economics of the industry. Affeldt et al. (2013) provide an extension of the UPP/GUPPI methodology to two-sided markets. Two-sided markets are markets that serve two distinct groups of customers, each of whose demand depends on the number of customers of the other type. Payment card systems, which serve both retailers and consumers, is an example of a two-sided market; retailers will value the system more as additional customers use the card, and similarly consumers will value the card more when more retailers accept it. This extension of the UPP/GUPPI methodology is potentially important because mergers regularly occur in markets that can be characterised as two sided and the impact of a merger on firms' pricing incentives will depend on the nature of that two sidedness. The article uses both the conventional and the extended methodology to consider a merger between local daily newspapers in Holland, which by their nature are two sided, as newspapers serve both advertisers and readers.
Why might the impact of a merger on pricing incentives change in a two-sided market? The key intuition is that when firms raise prices in a two-sided market, this not only reduces sales on that side of the market (say, readers) but potentially also reduces sales on the other side of the market (advertisers). This will occur if there are positive indirect network effects (say, if advertisers have higher demand for articles which have more readers). The overall effect is that the competitive constraint on pricing will be greater, where these positive indirect network effects exist, than would be predicted by reviewing only one side of the market in isolation.
This in turn means that the impact of a merger on prices will tend to be greater than might be predicted from a one-sided review only and this is exactly what the algebra and example presented in the article demonstrate. The article acts as a warning that assessing UPP/GUPPI only on one side of a two-sided market would lead to measures that are biased downwards, where there are positive indirect network effects. Of course markets can also be characterised by negative network externalities – as would occur for example if readers disliked advertising – and then the forces at work would apply in the opposite direction. More generally, we may have a mix of positive and negative indirect network effects.
However, it is important to note that what has proven relatively easy to develop is the algebra. The data requirements needed for application of the tool in real-world cases are substantially larger than in the one-sided version and may prove unrealistic to collect in some instances. For example, to carry out the analysis, separate estimates are needed for marginal costs on each side of the market but there could be significant practical difficulties involved in estimating such marginal costs, not least because it may not be straightforward to allocate the firm's marginal costs between its activities on the two sides of the market.
There are other well-recognised limitations to the UPP/GUPPI methodology. For example, it does not allow for product re-positioning, nor does it reflect on whether other competitors have the capacity to serve the customers that the merged firm is expected to lose. These limitations are as valid in two-sided markets as in one-sided.
Importantly, however, the authors note an interesting further limitation which is specific to two-sided markets. A shortcoming of the UPP/GUPPI methodology is that it ignores pricing responses by competitors. In a one-sided market, with differentiated Bertrand competition, prices may be assumed to be strategic complements; meaning that firms in a market will raise their prices in reaction to their rivals raising their prices. In such circumstances, the pricing responses of competitors will tend to further increase the incentives of the merged firm to raise prices. In this case, UPP/GUPPI would be a conservative measure of the true merger impact on pricing incentives. In practice, this may not be true even in one-sided markets; some empirical models do not impose the assumption of strategic complementarity, allowing for the possibility that when rival firms win significant numbers of price sensitive consumers their optimal prices can fall (i.e. for prices to be strategic substitutes). In a two-sided market, however, the situation is more complex, even in the theoretical models. Prices on one side of a market may be strategic complements (as in theoretical one-sided market models) or strategic substitutes (Fahri and Hagiu, 2008). This means that UPP/GUPPI may overstate or understate post-merger incentives to raise prices in two-sided markets.
The second article by Duso et al. (2013) studies the effect of the 2004 reform of the EU merger regulation on the performance of the European Commission's Directorate General for Competition (DG Competition) in merger evaluation. A key element of this reform was that the legal standard for merger assessment was changed from a dominance test to a ‘substantial lessening of competition’ (SLC) test. The earlier test assessed whether a merger was likely to create or enhance a dominant position in the market, where dominance can be viewed as substantial individual firm market power. This test was seen as potentially failing to capture an important class of significantly anti-competitive mergers, in which competition was lessened but not eliminated, and the move to the SLC test already in use in a number of jurisdictions (including the UK and US) was broadly welcomed as being better aligned with the economic theory of mergers.
In assessing this reform, the authors build on the basic approach and intuition due to Eckbo (1983, 1985), which argues that the stock market reaction of rival firms could be informative about whether a merger was pro or anti-competitive. Broadly speaking, a merger between directly competing firms can have two types of anti-competitive effects: ‘coordinated effects’, whereby the merger increases the likelihood that firms will be able to tacitly coordinate in the market; or ‘unilateral effects’, whereby the merger enhances the merged firm's market power through the elimination of a competitive constraint on that firm. At the same time, a merger can sometimes deliver significant pro-competitive effects, either because it generates important cost efficiencies or because it otherwise enhances the ability and/or incentive of the merged firm to compete more vigorously. This could include making tacit collusion less easy to achieve or sustain.
It may seem natural in considering a merger to analyse the impact of the merger on the merging firms' profits. However, these can increase (or indeed decrease) in both anti-competitive and pro-competitive mergers. The important intuition identified by Eckbo is that a merger that is anti-competitive will tend also to be good for shareholders in rival firms to the merged firm, as market prices will increase. By contrast, a pro-competitive merger which generates efficiencies for the merged firm, or otherwise leads it to competing more vigorously, will tend to be bad for shareholders in rival firms because they now face a tougher competitor.
While not in the authors' sample, an interesting illustration is provided from the aftermath of the prohibition of NYSE/Deutsche Börse1 merger wherein shares in the London Stock Exchange (LSE) gained almost 3%.2 Why would that happen? Eckbo's identification strategy suggests one reason: prohibiting a merger which would have made NYSE/Deutsche Börse more efficient means that LSE shareholders are better off without the merger than with it.
Or course, as with every identification strategy, there are limits to its power. The assumption made is that any merger which is bad for rivals will be good for consumers because prices decrease. However, in the presence of positive network externalities, for example, such as those achieved when market liquidity is enhanced by bringing together two trading platforms, a merger may potentially mean both that the ‘winner takes all’ in a market (bad for rivals) and also that the merger generates market power (bad for consumers) as well as the liquidity efficiencies (good for consumers). The same is also true of any merger which enhances the ability of a firm to abuse a dominant position, for example, a firm using its market power to prevent actual or potential competitors from entering or expanding in the market, to the detriment of both rivals and consumers.
In addition, the stock market valuation of firms will sometimes reflect the likely future merger prospects of a firm. If a merger is blocked, this may lead the stock market to downgrade its expectations that a rival firm will be successful in getting its own prospective merger through. In this case, its share price may fall, even if its future earnings prospects are unaffected (or even increased) by the prohibition.3 In such situations, we cannot immediately equate ‘good’ mergers with those whose rival stock market valuations decline on their announcement.
Nonetheless, and given the various limitations inherent to the identification strategy, the article is able to reach some apparently robust conclusions about the application of EU merger law, before and after the 2004 EU reforms. The primary focus of the article lies in examining the stock market reactions of rivals in respect of a sample of 368 EU mergers from 1990 to 2007, 216 of which were notified during the pre-reform period. Around 250 of the mergers were cleared in phase I of the two-phase merger assessment process.
As a preliminary test, not involving stock market movements, the authors test for the ex ante predictability of merger policy based on observable characteristics of the merger. They find that predictability increased post-reform. Given that predictability of merger policy is generally seen as a good thing, this is presumably a positive reflection on the reform. The authors then go on to apply a series of tests based on stock market movements. First, they test for the extent of type I discrepancies (apparently pro-competitive mergers that were blocked/remedied) and type II discrepancies (apparently anti-competitive mergers that were unconditionally cleared). They find that the frequency of type I discrepancies decreased significantly post-reform. Again, this is a positive finding. Second, they examine the degree of rent reversion resulting from intervention, where it occurred. They find that prohibitions substantially reverse rents, but that their results suggest that the effectiveness of remedies (i.e. commitments provided by merging parties to allay DG competition's concerns about anti-competitive effects of a merger) is more limited.
Finally, the authors look at the deterrence effect of merger policy interventions on the likelihood of firms proposing anti-competitive deals. They find that past prohibitions, remedy decisions and withdrawn mergers have the combined effect of reducing the probability of anti-competitive mergers being notified, while having no effect on the probability of pro-competitive mergers being notified. They interpret these results as a sign of the regime achieving effective deterrence of anti-competitive mergers but not over-deterrence.
Overall, then, the article paints a mixed picture of the 2004 EU Merger Policy Reform and, indeed, of current EU merger policy. While there are some strong positive aspects, the article also raises potentially important concerns about remedy effectiveness, which would merit further investigation. The authors argue that, given these concerns about remedy effectiveness, there may be an argument for DG competition making greater use of the more intrusive tool of full prohibitions. In that context, it is noteworthy that there has indeed been an apparent rise in the proportion of EU prohibitions in recent years, albeit after the period studied in this article.
In the context of cartels, ‘ringleaders’ are cartel members who initiate, facilitate or through coercion enforce collusive behaviour. Davis and De (2013) examine the role of cartel ringleaders in 89 European cartel decisions during the period 1990–2008, building on earlier work on the role of ringleaders by Ganslandt et al. (2012) and Bos and Wandschneider (2011). They find ringleaders in 19 of these cartels, or around 1 in 5 of the cartels studied.
Strikingly, the authors find that ringleaders tend to appear where there are relatively large numbers of firms in the cartel and also there is a relatively significant degree of asymmetry in the cartel. This corresponds to situations in which cartels are traditionally believed to be more difficult to sustain. As such, Davis and De (2013) argue that the evidence is consistent with a view of ringleaders as having the role of facilitating stable cartel behaviour where it might not otherwise occur. Also noteworthy is the fact that formal compensation schemes and price wars are significantly more prevalent in cartels with a ringleader than those without, which again fits with ringleaders existing where a cartel is otherwise harder to sustain. Likewise, for this sample at least, ringleaders are not observed for cartels that involve agreements not to compete in certain geographic areas or over certain categories of customers rather than explicit agreements about prices (known as market sharing cartels), which is consistent with the view that such cartels are typically easier to sustain than price fixing or bid rigging cartels.
The authors distinguish between two possible broad roles of ringleaders. The first is a ‘facilitating’ or ‘organisational’ role, which could be seen as providing a delegated administrative function on behalf of the set of cartel members. The second is a more ‘aggressive’ role, dictating price and coercing cartel members. The article categorises the 19 ringleader cartels into these two broad categories. In all 11 ‘aggressive’ cases, it finds that the number 1 ranked firm in the cartel was a ringleader – either alone or together with the number 2 firm. As the authors state, this suggests that if a ringleader is to exert an aggressive role it needs to be able to wield some power within the cartel. The authors also find that aggressive ringleaders are more likely to be found in more asymmetric cartels. By contrast, in the eight ‘organisation’-only cases, it is more common to find multiple leadership and not always involving the number 1 ranked firm.
As the authors state, analysis of this sort can never paint a complete picture. The evidence is based on published European Commission cartel cases, which necessarily reflect detected cartels only. We do not know the extent to which these provide a fair picture of the wider set of undetected cartels. In particular, if the existence of ringleaders tends to provide greater stability within cartels, it may be that ringleaders are far more prevalent amongst undetected cartels than these figures suggest. Likewise, the analysis is heavily dependent on what the Commission was able to prove as part of its investigation. There may well have been greater prevalence of ringleaders even within these 89 detected cases, which is not recorded. It is also noteworthy that nearly half (8 of the 19) of the Commission decisions referring to ringleaders occur in just 2 (2000–2) of the 18 years covered by the sample, 1990–2008. This raises the question of whether there is something peculiar about the Commission's cartel decision making in that period. Finally, this is clearly a limited sample and we need to be cautious about drawing overly strong conclusions on the basis of just 19 ringleader cartels.
Even so, the article is very welcome as a contribution to the literature. Evidence from actual cartel investigations consistently suggests that results from simple economic theory about when we will find collusion sustainable are overly stylised.4 It is important, therefore, to understand more about mechanisms, such as ringleaders, that can be used to overcome the difficulties inherent in agreeing to collude, monitoring rivals and ultimately ensuring at least some degree of cartel stability.
The article does not endeavour to draw policy conclusions but it does mention that ringleaders are treated differently under the US and EC cartel leniency programmes. In the US, ringleaders are unable to claim immunity, whereas in the EC they can do so if they have acted as an ‘instigator or played a determining role’ in a cartel but not if they have ‘coerced’ others to join the cartel. Assessing which of these approaches is most effective in practice in deterring cartels would be a useful area for future research.
3. Re-sale Price Maintenance
Re-sale price maintenance occurs when an upstream supplier sets the re-sale price (or alternatively the minimum re-sale price) at which a downstream distributor or retailer can re-sell its goods. From an economic perspective, RPM is well understood to have both potentially significant anti-competitive effects but also important possible efficiency benefits. For this reason, there has been substantial debate about the appropriate legal treatment of RPM in both the US and EU, with the US recently moving from a per se legal standard to a ‘rule of reason’ standard for assessing RPM.5 Under a per se standard, an RPM agreement will constitute an infringement irrespective of the economic facts of the case, while a demonstrably pro-competitive RPM agreement would be allowed under a ‘rule of reason’ standard. Under EU and UK competition law, however, RPM remains a ‘hardcore’ infringement. This is akin to a per se infringement, albeit efficiency arguments in favour of the practice are not absolutely barred from consideration as they are for per se infringements in the US.
Because RPM is viewed as a hardcore infringement in the UK and EU, the OFT's standard practice over the relevant period, on receiving a complaint about alleged RPM, was to send a warning letter requiring that any RPM be terminated. Giovannetti and Magazzini (2013) analyse a novel data set of 72 RPM complaints received by the UK Office of Fair Trading (OFT) between 2007 and 2009. Each complaint was followed by a warning letter, which not only asked the firm to desist from the RPM but also requested a reply to the OFT to confirm its compliance. The data set records whether a reply was received, whether it confirmed compliance and any other substantive points made in its response.
The authors have supplemented the OFT data set with a variety of information about characteristics of the upstream and downstream markets and firms involved. (The complainants are invariably downstream firms, who wish to be able to price freely, complaining about RPM imposed by upstream firms.) From this wider database, the authors derive several conclusions of which the following are the most striking.
First, the upstream market shares of the firms complained about are typically well below the level at which any market power might be inferred (as indeed are the market shares of the downstream firms making the complaints). This suggests that any RPM is not likely to be related to upstream firms exploiting or enhancing their individual market power.
Second, around 70% of the complaints related to markets where there was a multiplicity of complaints, raising the possibility of network effects arising from the widespread use of RPM across a market.
Third, the data are consistent with a view that much of the RPM was focused on limiting price cutting in an online environment; 40% of the complaints came from downstream firms that operated only online, whereas the remaining 60% operated both online and on the high street.
Fourth, over 30% of the firms that received warning letters responded to explain that they had engaged in RPM to protect their brand or to overcome free riding issues (especially around customer support). This fits well with some of the economic efficiency rationales proposed in the theoretical literature.
Fifth, over 20% of firms who received warning letters stated that they had put RPM in place following complaints from other downstream firms (presumably rivals of the firm that (later) complained to the OFT). This is potentially consistent with a picture of RPM instigated by downstream firms with buyer power. In this light, however, the database used for the analysis does not include information on the market shares of other downstream retailers in the markets involved. As such, the authors are not able to test for whether RPM is more prevalent in such circumstances. This may be a useful avenue for further research.
The authors then carry out an econometric study to assess the factors that tend to drive firms to respond to the OFT confirming that they would comply with the request to terminate their RPM. Although this is a small data set, they nevertheless find some apparently strong results. In particular, compliance is more likely to be confirmed by firms for which similar complaints had already been recorded with OFT (i.e. ‘recidivists’), more likely when the downstream complainant is a wholesaler rather than a retailer, and also more likely where the downstream complainant is a purely online business. This latter point is perhaps somewhat counter-intuitive. While the authors suggest one possible rationale, this again may merit future research.
4. Optimal Fining Policy
Finally, the collection includes two interesting articles studying the effect of fining policy rules on consumer and firm welfare. While fines are levied on firms, after the fact, for specific breaches of the law, a key aim of fining policy is to deter firms more generally from breaching competition law in the first place. This deterrence effect is designed to reduce the extent of anti-competitive behaviour in this economy, with consequent benefits for consumers.
However, the welfare effects of a given fining policy include not only the benefits that result from this deterrence effect but also any costs associated with that policy, including those that arise from unintended consequences such as any changes in firms' incentives to compete that result from the policy. An optimal fining policy involves striking the socially optimal balance between such costs and benefits.
Many existing fining policies, including that in the UK, relate fines imposed on any firm to that firm's revenues. Under such policies, the core fine will typically be related to the firm's revenues in the market affected by the anti-competitive behaviour and there may then be a cap on total fine that is related to the firm's total revenues across all markets. Bageri et al. (2013) study the welfare properties of fines imposed for cartel behaviour, where these involve simple rules of this sort, based on firms' revenues.
They argue that such simple rules may be actively harmful. First, they consider fining policies that incorporate a cap on fines which is set as a proportion of total firm revenues. These, they argue, can introduce distortions as those firms that are diversified, with activities across a range of different markets, will end up having a looser cap in respect of cartel behaviour in a given market than those firms that are active only in that market. This could in turn encourage firms to under-diversify to limit their legal liability.
Second, they consider fining policies which set fines as a proportion of the firm's revenues in the market affected by the cartel behaviour. They argue that this can introduce perverse incentives for non-deterred cartelists to restrict output even below monopoly levels.6
Third, also in respect of such fining policies, they highlight that there will be unjustified cross-sectional variation in the relationship between cartel profits and potential fines, depending on where a firm is in the value chain. Firms at the end of a supply chain, such as retailers, will tend to have low profit/revenue ratios, due to their high input cost of materials relative to their own value added. Such firms will face larger fines, relative to their likely collusive profits, than firms that have a larger profit/revenue ratio, for example, due to their position at the beginning of the supply chain. The authors conclude by discussing the ‘obvious need to depart from these distortive rules-of-thumbs that appear to have the potential to substantially reduce social welfare’.
While they make a compelling case in terms of their theoretical arguments, the questions that our legal colleagues would legitimately raise ultimately, when designing practical fining policy, are
- whether the theory is robust, or whether more qualifications within the models would alter the findings;
- whether there is any empirical evidence that these distortions are material in practice; and
- whether there are positive practical reasons to consider in setting policy, such as the fact that revenue is typically far easier to measure than profits.7
All these are fair challenges. The article by Katsoulacos and Ulph (2013) starts to address the first, in that it extends and refines the basic fining model which underlies the findings of Bageri et al. (2013). The authors also view the core model as potentially applicable more generally than just to cartel cases, including not only non-cartel competition cases but also regulatory cases.
The first major point the article considers is how optimal fining policy might be affected by the timing of the authority's intervention. This could occur early in the life of an infringement, later in the life of an infringement or after an infringement has ended. The timing of intervention affects how much profit the infringement has already generated, how much prospective overcharge is avoided through intervention and also how likely the infringement is to be detected in the end.
The authors draw on a database of 32 abuse of dominance cases investigated by European national competition authorities from 1990 to 2010 to derive estimates of decision timing. On the basis of these, they conclude that allowing for timing of decisions reduces optimal fines to around 75% of those calculated without allowing for timing. Clearly, questions could be raised about whether evidence on timing from abuse of dominance cases (which involve firms abusing their individual market power) will also be applicable to cartels (which involve firms coordinating their behaviour to exploit their joint market power), or indeed other forms of infringement. Nevertheless, the derived figures seem broadly plausible and the key contribution of the article is in providing a methodology that allows for timing of interventions. Estimates specific to other forms of infringement could potentially easily be used for analogous calculations, if developed through separate research.
The authors then apply their analysis to derive an optimal penalty rate, drawing on earlier analysis by Boyer and Kotchoni (2011) who found that cartel activity raised prices by 17.5% on average. This article draws on data collected by Connor and Bolotova (2006) but refines their estimate of a 30% average cartel overcharge, having argued that this previous figure suffers from a significant upward bias. On this basis, and assuming an annual likelihood of intervention for any infringement of 14%, Katsoulacos and Ulph (2013) calculate an optimal penalty rate of 25–31% of revenue. They highlight that this is broadly consistent with current fining levels.
It is noteworthy, though, that the optimal fining rate depends heavily on the estimated cartel overcharge. Moreover, while the discussion is presented in terms of the average overcharge, an important property of the Connor and Bolotova (2006) database is that overcharges are highly positively skewed (Boyer and Kotchoni, 2011). While the median overcharge estimated by Boyer and Kotchoni (2011) is just 13.6%, there are examples in the long tail of overcharges amounting to 100%, 200%, 300% or even up to 1,800% in the underlying Connor and Bolotova (2006) data set. This raises a problem. If cartels are both harmful and can potentially occur in a fairly wide range of circumstances, then we need policy rules to deal with deterrence in that fairly wide range of circumstances. But the data would appear to suggest that the optimal fine rate would vary absolutely enormously across cases, possibly far more than is currently accounted for by relatively limited adjustments from a base case. This is clearly a difficult issue and we believe it would in itself merit further research. It may be that it provides further support for giving more weight to profits, rather than (or in addition to) revenues for the purpose of setting fines. In the UK, it is noteworthy that, while penalties are initially set in proportion to turnover, they are then adjusted for ‘proportionality’ which in practice involves considering their relationship with a number of other measures including profitability.
In addition, if the authors are serious about ensuring the applicability of their model to a wider set of infringements than cartels only, a further extension to the model may be usefully considered – to allow for changes in the profits and revenues associated with infringement over time. This may be especially important in considering exclusionary abuse of dominance cases, where profits (and perhaps revenues) will often be low during the exclusionary period, but then high during the ‘recoupment’ period.
The second part of the Katsoulacos and Ulph (2013) article relates more specifically to the potential for distortions in the case of cartels arising from setting penalties proportional to revenue rather than profit. This draws on the modelling and findings in Bageri et al. (2013) but considers two key extensions to the basic model (summarised in footnote 7).
First, they allow for the possibility that intervention may be more likely when the cartel overcharge is higher, perhaps because more egregious price levels are more likely to result in authority scrutiny. This potential effect is important because it means that the incentive to set high cartel overcharges is more muted than the simple model would suggest and it becomes ambiguous whether these overcharges are distorted upwards where penalties are set as a proportion of revenues.
Second, the authors allow for a deterrence effect, whereby firms make an initial decision whether to engage in cartel behaviour at all. In this case, higher penalty rates (set as a proportion of revenue) will tend to reduce cartel formation but will tend to increase cartel overcharges in those cartels that do form. Interestingly, the article shows that incorporating a fixed element into the fine can enhance deterrence and also reduce the overcharge (albeit it is not clear how such a fixed charge could be put into practice in a proportionate way, given the very wide range of sizes of actual cartels observed).
These extensions to the basic model are welcome. However, the model remains a fairly simple one. Certainly it does not allow for the fact that cartels can be difficult to form and complex to keep together, as emphasised in Davis and De (2013). In the context of real cartel situations, it may be unrealistic to expect cartels to adjust their prices and output to reflect fining policy, as if they were a single monopolist, in the way described. Moreover, in respect of the first of the distortions highlighted in Bageri et al. (2013), the authors themselves state that ‘we don't believe this is commonly happening’, taking the view that antitrust liability concerns are likely to be of secondary importance for firms when considering, for example, which markets to enter.
Nevertheless, these two articles provide a useful start to the integration of the theoretical and empirical literatures in this area. They raise issues that certainly need to be considered seriously in designing an optimal fining policy but also raise further questions which would merit future research.
The articles in the Feature each make helpful contributions to the growing and important literature on the economics of competition policy – a literature which no doubt will continue to grow in both scale and, in time, influence practitioners and policy makers alike.
Going back to the announcement date for NYSE/Deutsche Börse also nicely illustrates the difficulty that arises when other events affect the rival companies. Specifically, the LSE also announced that it would merge with Toronto Stock Exchange on the same day, making the rise in LSE's stock price in early trading difficult to interpret. See http://www.guardian.co.uk/business/2011/feb/09/lse-tmx-merger-canada-stock-exchange.
For example, authors tend to find that it is only the biggest firm or the smallest firm which matters – but such results only occur in very stylised models. More generally, there is large ambiguity coming from the theory literature about the degree to which firm asymmetry will genuinely act to prevent cartel formation. The reason can be thought about in simple full-information contexts in terms of firm-specific incentive compatibility constraints (ICC). If every firm has an ICC which is basically a non-linear function of prices – wherein some price vector p is a feasible collusive price if the value to that firm of colluding is greater than the value of defecting –, then the ICCs will define a potentially highly non-linear set of prices where collusion is feasible . This non-linear set will take on a different shape depending on the strategies being followed (e.g. grim strategies or other) and also on the detail of the game – firms discount factors, the degree of differentiation, and so forth. Thus, the conventional economic theory from full-information models of coordination does not in reality say that asymmetry will immediately kill off possibilities for coordination; its prediction is far more ambiguous – asymmetry whether created or not may or may not prove harmful to cartel formation. Of course that is not to say that there are also no reasons to believe that asymmetry can indeed be harmful to cartel creation.
See the US Supreme Court Decision, Leegin Creative Leather Products Inc. v. PSKS Inc. 551 U.S. 877 (2007).
To see why, consider a cartel's optimisation problem, when β denotes the likelihood of the cartel being discovered and φ denotes the proportion of revenue paid in fines if so. Relative to a benchmark of perfect competition, we have:
so that the first-order conditions can be written as: . If we evaluate this marginal incentive at the normal cartel outcome Q* where R′(Q*)−c = 0, it is clear that , so that if marginal revenue is positive R′(Q) > 0, the cartel has an incentive to increase its profits by further cutting output. High fine levels effectively decrease the expected marginal revenue associated with any given level of output. Thus, the theory provides a clear prediction that a fine based on a proportion of revenue applied to detected cartels provides an additional incentive – for those cartels that are not deterred from colluding – to decrease output or analogously increase cartel prices. Moreover, the greater the fining level φ the larger the distortion introduced.
We understand that in 2005 Germany, for example, actively moved from cartel fines based on illicit profits (three times the additional proceeds obtained as a result of the violation) to fines based on revenues partly because of perceived issues around practicality of using profits rather than revenues.