The recent EU Google decisions may represent a high-water mark for the use of behavioral economics in EU antitrust to date, but what do they imply for competition policy in the future? Do such cases represent the outer extremes of how far behavioral thinking can and should be taken? Or do they represent baby steps towards the more comprehensive incorporation of behavioral economics into competition policy thinking? This article highlights the widespread influence of behavioral economics across other areas of policy and discusses a number of directions in which competition policy could potentially be transformed. Noting the existence of an extensive literature in behavioral antitrust, it focuses on a number of aspects which have been given less attention to date.
PDF – Click here for the full article.
I. INTRODUCTION
The recent EU Google decisions made waves around the world. In large part, this was because they involved a huge and innovative digital economy platform and were the highest fines yet imposed by the EU. However, the decisions are also notable for their reliance on key insights from behavioral economics.
Behavioral economics may not have been mentioned explicitly in the Google Shopping decision, but the Commission’s case hangs on the fact that the “more favorable positioning” of results on the Google search page leads to increased traffic and click-throughs. As such, the case effectively relies on a behavioral tendency called saliency bias, whereby individuals typically decide on the basis of what is most obvious or prominent to them.
The EU Google Android case is more explicit still. The decision is not yet out, but the press release uses behavioral economics terminology in mentioning status quo bias as a key underlying driver of the abuse. This behavioral tendency means that users who find search and browser apps pre-installed on their devices are likely to stick to these apps.
Together, these Google decisions may represent a high-water mark for the use of behavioral economics in EU competition policy to date, but are they so novel? Not necessarily. The 2009 EU Microsoft Browser decision was effectively also related to status quo bias, even if the term was not used, and the “choice boxscreen” remedy was specifically designed to ensure consumers made an active and unbiased choice, implicitly recognizing the risk of default bias.
The bigger question is what the Google decisions imply for competition policy in the future. Are such cases as far as behavioral economics can and should be taken in this arena? Or do these cases represent baby steps towards the more comprehensive incorporation of behavioral economics into competition policy thinking?
In some senses, the use of behavioral economics in these cases is far from extreme. The biases mentioned above – status quo bias, default bias, and saliency bias – are among the most well-evidenced and least controversial of all behavioral biases. It would arguably be more extreme to ignore factors that are so obviously relevant to how consumer behavior, and therefore competition, in fact works.
Moreover, behavioral thinking has already had widespread influence in a variety of other areas of policy. In EU consumer law, the recent online ban on opt-out selling of add-on products (otherwise known as the ban on pre-ticked boxes) is specifically designed to address default bias. In developing recent EU law relating to financial services, the Commission has increasingly adopted consumer testing of new regulations, reflecting a recognition that behavioral factors are highly relevant. In the UK, the law has been changed such that firms must now provide workplace pensions on an opt-out, rather than an opt-in, basis. This was based on an understanding of behavioral biases, and in particular the consumer inertia, surrounding pension decisions.
So, does such thinking have a much wider role to play in competition policy too, reflecting the revolutionary effect it is having across both economics and policy more widely? This short article discusses four key questions, relating to ways in which behavioral thinking could potentially transform this area.
- Does effective competition policy require more than standard antitrust?
- Within standard antitrust, will behavioral economics change theories of harm?
- What does behavioral economics imply for empirical analysis in antitrust?
- Do supply-side biases need to be considered too?
Some of the ideas discussed below may be controversial. Others, however, seem almost incontrovertible if competition policy is properly to reflect real world behavior.
II. DOES EFFECTIVE COMPETITION POLICY REQUIRE MORE THAN STANDARD ANTITRUST?
In the UK, the competition policy tools available to the Competition and Markets Authority (“CMA”) include market investigations. If the CMA identifies features of a market which have an adverse effect on competition, it can impose proportionate remedies.
A key insight from these investigations, over the years, is that the identified competition problems are often related to problems on the demand-side. These are very different from the sorts of supply-side considerations that are more normally addressed by standard antitrust. However, they can be equally important, if not more, for driving effective competition which delivers for consumers.
What is the thinking behind this? Behavioral economics tells us that consumers may not act like rational automata, choosing an optimal product that perfectly maximizes their utility. Rather, they instead exhibit all sorts of biases and these can in turn have important implications for competition.
- What is the thinking behind this? Behavioral economics tells us that consumers may not act like rational automata, choosing an optimal product that perfectly maximizes their utility. Rather, they instead exhibit all sorts of biases and these can in turn have important implications for competition.2
- Second, if consumers differ in the extent to which they exhibit such biases, we may observe market segmentation, whereby there is plenty of competition for “active” customers, but far higher prices for “inactive” customers. Overall, profits need not necessarily increase, depending on the extent to which firms compete away, in the “active” segment, the rents they make from the “inactive” segment. Nonetheless, such pricing may be of concern, both because it distorts consumption decisions and due to fairness considerations, the latter of which can also be highly political. In the UK energy market, such concerns recently culminated in the introduction of a safeguard price cap to protect inactive customers. In the meantime, there are continuing attempts to develop more competition-friendly solutions to this particular problem.
- Third, biases may result in competition occurring on “the wrong dimensions.” For example, if consumers are more likely to choose products on the basis of what is most salient, then firms will tend to compete harder on more salient dimensions and act more monopolistically on less salient dimensions. As a result, in some markets we may see plenty of competition on upfront price, which is highly salient, but firms offering poor quality or terms and conditions, which are less salient. Consumer law can help here. For example, the law on unfair contract terms can be viewed as a way of helping to ensure that competition works to deliver good consumer outcomes. However, other interventions may also be needed to nudge consumers towards the more holistic appraisal of options and reduce the impact of saliency bias.
- Third, biases may result in competition occurring on “the wrong dimensions.” For example, if consumers are more likely to choose products on the basis of what is most salient, then firms will tend to compete harder on more salient dimensions and act more monopolistically on less salient dimensions. As a result, in some markets we may see plenty of competition on upfront price, which is highly salient, but firms offering poor quality or terms and conditions, which are less salient. Consumer law can help here. For example, the law on unfair contract terms can be viewed as a way of helping to ensure that competition works to deliver good consumer outcomes. However, other interventions may also be needed to nudge consumers towards the more holistic appraisal of options and reduce the impact of saliency bias.
Indeed, in all of the above, while consumer law can clearly play a positive role, it is important to recognize that consumer law is essentially motivated by a focus on consumer protection, not competition concerns. As such, it may not always be ideally designed for the latter objective.
If competition authorities are to address this important aspect of competition policy effectively, therefore, they may require competition-focused rule-making tools – like UK market investigations – which go beyond the standard antitrust provisions. This set of concerns may also provide a rationale for combining competition and consumer enforcement powers within one authority, which is then able to address concerns from both perspectives at once.
III. WITHIN STANDARD ANTITRUST, WILL BEHAVIORAL ECONOMICS CHANGE THEORIES OF HARM?
Behavioral economics can also enhance our understanding of how firms’ actions can have anti-competitive effects. This may involve entirely new theories of harm, but it may also involve tweaks to more standard theories of harm. There is an extensive existing literature on “behavioral antitrust” which highlights a number of such potential implications.3 Rather than simply summarize those ideas, this article focuses on a few more novel, and potentially more controversial, aspects which have been given less attention to date.
A. Abuse of Dominance
First, as is shown by the Microsoft Browser and Google Android cases, the impact of tying and bundling can potentially have a more serious anti-competitive effects if one allows for default or status quo bias. If consumers can be tied into a particular related service initially, this can create long-term market power, even if they are free to switch thereafter. This is a well-recognized point.
Perhaps less obvious is that, in the digital arena, this effect may potentially be amplified by the fact that services are ostensibly free, albeit effectively paid for with consumer data. Another behavioral bias may be relevant here. If services are apparently free, then consumers may be disinclined to focus on the less salient price they are paying in terms of their data. Moreover, even if they did, consumers find it very hard to value this data, and their revealed preferences may be very different from their stated preferences. Such factors may make consumers even less likely to move away from the default or status quo choice, thus exacerbating the risk of anti-competitive tying and bundling in this digital environment.
Second, as discussed above, the Google Shopping case essentially relies on saliency bias, such that consumers tend to make choices on the basis of what is most prominent to them, rather than assessing information more holistically. While that case involves a platform giving undue prominence to its own vertically integrated offering, and thereby leveraging its market position from one activity to another, the strong impact that rankings can have on sales by platform users could potentially have wider anti-competitive effects.
Suppose, for example, that a platform provides seller rankings to consumers which appear to reflect their interests but in fact depend on the level of commission paid to the platform by the sellers. This means that competition ends up occurring on the basis of which sellers can pay the most to the platform for the ranking, rather than which sellers actually offer consumers the best product offering. Such behavior is therefore potentially misleading for consumers, which could breach consumer law. However, it is also arguably exploitative of the platform’s “competitive bottleneck” position in reaching those consumers. This is especially likely to be true where consumers are “single-homing” in that they do not search around across platforms, perhaps due to status quo bias. If the platform also gives better rankings to sellers which are willing to list exclusively on that platform, this could potentially also be exclusionary.
Moreover, requiring the platform to provide information to consumers on how the ranking is in fact derived is likely to be of little use in correcting this situation, given that real consumers, who exhibit bounded rationality, are unlikely to know how to adjust their choices on the basis of this information. Given this market context, then, might such ranking rules constitute an abuse of dominance?
Third, another area which has raised renewed interest in recent years has been personalized pricing. Where price discrimination was based on consumers’ willingness to pay, authorities typically took a fairly sanguine view. However, it is not obvious that this laissez faire approach is still justified when price discrimination reflects consumer biases.4 Again, this has been identified as a particular issue in a digital environment. In this context there are also concerns that such price discrimination will be much easier for an incumbent which has masses of data about consumer behavior, and much harder for an entrant without access to such data. As such, in a digital environment, personalized price discrimination could potentially be exclusionary, as well as exploitative.
Fourth, the impact of consumer myopia on firms’ incentives to protect and exploit their own proprietary aftermarkets are fairly well understood. The less weight that consumers give to future prices, the more they will tend to opt for low upfront prices, even if these are to be followed by “rip-off” aftermarket prices. As such, a greater integration of behavioral economics within antitrust could thus lead to a renewed interest in aftermarket cases.
Perhaps less well recognized is that similar “consumer tie-in” effects can result from other behavioral biases too. For example, saliency bias can lead consumers to focus on prominent upfront prices and ignore the less prominent after-market prices. If firms are able to reduce the prominence of the latter through deliberate “shrouding,” could this be seen as an abuse of dominance?
Likewise, as discussed above, status quo bias can mean that, once a firm has won a consumer in one period, it is more likely to keep that consumer in later periods. Firms may be able to exploit this position by charging these inactive customers a higher price than they offer to active customers. They may also be able to exacerbate the effect of the status quo bias by making it harder for consumers to search or switch away, perhaps playing on behavioral factors such as forgetfulness (which may limit a consumer’s ability to cancel the contract during a defined termination window) or dislike of conflict (which may limit a consumer’s willingness to switch if it requires calling up the original supplier).
As such, just as is the case for aftermarkets, firms can potentially behave in an exploitative and exclusionary manner in respect of their inactive customers. Should this ever be considered an abuse of dominance?
B. Anti-competitive Agreements
Consumer behavioral biases may also have implications for anti-competitive agreements.
In terms of horizontal agreements, it is increasingly well understood that the presence of behavioral biases may potentially alter the likelihood of standard price or market-sharing collusion. For example, Bos et al. (2011) show that if consumers exhibit strong inertia, then price collusion is easier to achieve.5
Less attention has been given to the idea that, in the presence of demand-side behavioral biases, firms may be able to engage in an alternative form of collusion: collusion to dampen competition. For example, it may be in the joint interest of two rival firms to agree to set their price structures very differently, or make their pricing highly complex, in order to limit comparability between them. By dis-incentivizing consumer search, this can dampen competition and enhance firm profitability.6 Likewise, firms may agree to make quality far more salient to consumers than price. With saliency bias, this will tend to lead to competition occurring on quality rather than price, which in turn may again be rather weaker, and thus generate higher profits.
In terms of vertical agreements, much of the behavioral antitrust literature to date has focused on how behavioral biases may provide additional support for efficiency rationales frequently given for vertical agreements.7 However, there are potential negative implications of behavioral biases for vertical agreements too, which have so far been less well developed.
As an example, there has been much discussion of the role of retail MFNs in changing platforms’ incentives when setting their commission rates. There has been less focus on their potential impact on consumer behavior. If retail MFNs are accompanied by credible “best price” claims, then consumers who are anyway disinclined to search across different platforms may be even more inclined to stick to a single platform. This is important because any increase in “single homing” on the consumer side of the market will in turn tend to increase the market power of the platform in relation to the seller side of the market, by making the platform more of a competitive bottleneck to consumers. As such, retail MFNs can potentially increase platform market power, through their effects on consumer behavior.
C. Mergers
Demand-side biases may also have implications for merger analysis. For example, if consumers find it hard to think about absolute quality, then they may focus their decision-making on the relative quality of different products. In this situation, we may expect firms to compete more vigorously on quality, since there is an added incentive to achieve a higher quality than rivals. This in turn means that the impact of a merger in reducing quality might be greater than would be the case in the absence of this behavioral bias.
Likewise, we know that consumer behavior is affected by how the decision facing them is framed. Consumers may be more likely to buy a particular flight ticket, if they are informed that there are only a few tickets still available at the current price. They may be more likely buy a £5 bottle of wine that was £10 yesterday than one which has always been £5, even if the £10 was never a real price.
This in turn means that such framing behavior can affect competitive outcomes. Should merger assessment therefore include consideration of the impact of merger on firms’ incentives when framing consumer choices? For example, how should authorities consider a takeover by one firm, which is expert in framing their offering in a misleading way, of a second firm which is more scrupulous?
IV. WHAT DOES BEHAVIORAL ECONOMICS IMPLY FOR EMPIRICAL ANALYSIS IN ANTITRUST
Behavioral biases may also have implications for the empirical evidence and analysis typically carried out in antitrust cases.
First, there may be a need for changes to existing analytical tools. For example, standard demand estimation techniques do not typically allow for the fact that consumer purchasing behavior may be strongly affected by both their past purchasing behavior and framing effects. For example, a price reduction from £2 to £1.50 may have a very different impact on sales if the price label specifically states “Was £2, now £1.50,” as opposed to the price simply changing without such labelling.
Meanwhile, consumers who exhibit loss aversion may have very different reactions to a price change depending on the direction of the change, with many more switching away on the basis of a price rise from £2 to £2.20 than would switch to the product on the basis of a price reduction from £2.20 to £2. Again, most demand estimation techniques implicitly assume symmetric reactions.
Second, behavioral biases have important implications for the effectiveness of remedies, where these are reliant on consumer behavior. For example, offering consumers a new option may have little impact on competition if they exhibit strong default or status quo bias. In some cases, remedies will only work well if they change the choice architecture facing consumers, not just the choice options. A thoughtful example was the remedy in the Microsoft Browser case; a “boxchoice screen” which forced consumers to make an active and unbiased choice. Following the introduction of this remedy in the EU, Internet Explorer’s market share in the browser market fell significantly more rapidly in the EU than it did in the U.S., which was not subject to the remedy.8
Consumer reactions can, however, be hard to predict, and competition authorities can easily get this wrong. A key implication, therefore, is that authorities should carry out consumer testing of any such remedies, ideally through the use of randomized controlled trials. This is a relatively new technique for antitrust, but has become increasingly commonplace in sector regulation, at least in the UK, when putting in place new consumer-focused regulatory interventions. It has shown clear benefits in terms of helping to identify the most effective remedies.
V. DO SUPPLY-SIDE BIASES NEED TO BE CONSIDERED TOO?
Finally, and perhaps most controversially, it has to be recognized that behavioral biases may not be restricted to the demand-side of markets. Firms can exhibit them too. There is an extensive and growing literature on the tendencies of executives within firms to engage in a variety of behaviors that are not necessarily profit-maximizing, such as empire-building, maximization of stock market valuation, focus on the relative performance of the firm (rather than its absolute performance), seeking admiration through taking big risks, and even protection of market share in order to protect jobs (perhaps to avoid difficult conversations with staff being made redundant).
In some cases, these behaviors may in fact be individually rational for the executives involved, given the reward structures they face and the perceptions of shareholders and wider capital markets, which may themselves be hard to fully rationalize. However, whether individually rational or not, these apparent biases can potentially lead to anti-competitive behavior (or indeed pro-competitive behavior) which is not apparently profitable for the firm.
As has been highlighted in the existing behavioral antitrust literature, this has clear implications for antitrust, which has traditionally included a strong focus on considering the profit incentive of firms to engage in the behavior in question. This emphasis has perhaps been stronger in the U.S., where the influence of the Chicago School has been stronger. However, such thinking is present in many EU cases too, with authorities often seeking to demonstrate in their decisions that the dominant firm is likely to profit from the abusive behavior.
If we take supply-side behavioral biases seriously, however, it is far from obvious that this is still a sensible question to ask. For example, a firm may engage in a course of abusive conduct simply because its CEO wants to preserve market share for personal reasons, and irrespective of whether the behavior will be profitable. Would it be so outlandish for competition authorities to allow for such possible motivations in abuse cases?
Likewise, in the context of mergers, if some mergers are driven not by pure profitability motives but by executive reward, empire-building incentives, or potentially over-optimism or over-confidence bias, then this may be relevant to assessing their likely effects and also their efficiency justifications. Such factors may also be relevant to assessing the likelihood of successful entry post-merger, in that potential entrants may well be over-confident about their likely success in the market. Indeed, recent ex -post evaluation work, carried out for the CMA, found entry to have been less successful than expected in constraining post-merger competitive outcomes in four out of the eight cases reviewed.9 Again, to what extent should the authorities take such considerations into account?
Similar supply-side biases may also justify additional scrutiny around purchaser approval for assets divested in order to gain merger clearance. The UK merger regime has sadly overseen a number of failed merger divestment remedies, most notably in a series of three grocery mergers (Co-op/Somerfield (2008), Co-op/Lothian (2009), and Asda/Netto (2010)) which led to the divestment, in total, of 52-54 stores to an apparently dynamic and aggressive new grocery retailer called Haldanes. The company turned out to have been unrealistically over-confident about its own chances of success. It failed in 2011, within a year of the final divestment, resulting in most of the divested stores closing down.
In the area of cartels, supply-side biases may affect the likelihood of anti-competitive agreements being formed and remaining stable. For example, collusion may be facilitated by strong trust and social links across cartel members,10 but may be hampered by a non profit-focused human desire to be a law abiding member of society. This suggests that a successful strategy to reduce collusion may need to focus on changing culture and social norms, not just penalizing illegal cartels.
Finally, it should be noted that allowing for behavioral biases on the supply-side is not the same as accepting that firms might engage in any sort of unexplained non-rational behavior. For example, it is sometimes argued that mergers will not lead to price rises or foreclosure effects, on the basis that managers of the different divisions of the merged firm will not take each others’ profits into account. Such arguments are unlikely to be credible unless there is a clear behavioral rationale, and supporting evidence, for such inaction.
VI. CONCLUSIONS
Overall, while the sum of the ideas above may be controversial, it is clear that behavioral economics is here to stay. It has already had huge influence across wide swathes of policy and law, and it seems unlikely that competition policy will be immune. Competition policy is, after all, essentially about making markets work well for consumers. It will only therefore be effective if it allows for the behavior of real markets with real consumers and real firms.
There is an extensive behavioral antitrust literature which sets out a number of potential developments. Some additional aspects are presented here. However, there may also be implications that no one has identified yet, even in theory. The academic field of behavioral economics is itself developing very quickly, and new thinking – for example, in respect of theories of harm – is highly likely to emerge. It would be reasonable to bet that we have only really seen the tip of the behavioral iceberg so far.
1 Centre for Competition Policy, University of East Anglia. This article was supported by the Economic and Social Research Council [grant number ES/P008976/1]. Amelia is also a Non-Executive Director on the Boards of the UK Financial Conduct Authority, Competition and Markets Authority, and Payment Systems Regulator, and a decision-maker on enforcement cases at Ofgem. Amelia is grateful for comments from colleagues Sean Ennis, Morten Hviid, and Bruce Lyons. However, all views are her own and do not necessarily reflect the views of anyone or any organization with which she is associated.
2 For a detailed description and evaluation of the “demand-side remedies” described in this and the following bullets, see Amelia Fletcher (2016), “The Role of Demand-Side Remedies in Driving Effective Competition: A Review for Which?,” available at https://www.staticwhich.co.uk/documents/pdf/the-role-of-demand-side-remedies-in-driving-effective-competition-456067.pdf.
3 See, for example, Matthew Bennett, Amelia Fletcher, Liz Hurley & David Ruck, “What Does Behavioral Economics Mean for Competition Policy?,” Competition Policy International, Spring 2010, 120-32, https://www.competitionpolicyinternational.com/what-does-behavioral-economics-mean-for-competition-policy/; Maurice E. Stucke, “Behavioral Antitrust and Monopolization,” Journal of Competition Law & Economics, Volume 8, Issue 3, 1 September 2012, Pages 545–574, https://doi.org/10.1093/joclec/nhs018; Avishalom Tor, “Understanding Behavioral Antitrust,” 92 Tex. L. Rev. 573 (2013-2014). https://scholarship.law.nd.edu/law_faculty_scholarship/296.
4 Such naïveté-based discrimination is discussed in Paul Heidhues & Botond Kőszegi, “Naïveté-Based Discrimination,” The Quarterly Journal of Economics, Volume 132, Issue 2, 1 May 2017, Pages 1019–1054, https://doi.org/10.1093/qje/qjw042.
5 Ivan Bos, Ronald Peeters & Erik Pot (2017), “Competition versus collusion: The impact of consumer inertia,” Int Jnl of Economic Theory, 13: 387-400. doi:10.1111/ijet.12136.
6 This possibility is discussed in Robert Edwards, “Pricing and obfuscation with complexity averse consumers,” Oxford Economic Papers, gpy053, https://doi.org/10.1093/oep/ gpy053.
7 For example, an interesting behavioral justification of RPM is provided in Roman Inderst & Sebastian Pfeil, “An Image Theory of RPM,” MPRA Paper 54139, March 2014, https://mpra.ub.uni-muenchen.de/54139/1/MPRA_paper_54139.pdf.
8 See Fletcher (2016). Footnote 2.
9 KPMG LLP, “Entry and expansion in UK merger cases: An ex-post evaluation,”, April 2017, Report For Competition and Markets Authority, https://www.gov.uk/government/publications/evaluation-of-entry-and-expansion-in-uk-merger-cases.
10 See Andreas Stephan, “Cartel Laws Undermined: Corruption, Social Norms, and Collectivist Business Cultures,” Journal of Law and Society, Vol. 37, Issue 2, pp. 345-367, June 2010, http://dx.doi.org/10.1111/j.1467-6478.2010.00507.x.
