What Constitutes Self-Preferencing and its Proliferation in Digital Markets

This is an Insight article, written by a selected partner as part of GCR's co-published content. Read more on Insight


Regulators and lawmakers have expressed concerns that large digital platforms could distort competition by pursuing strategies that favour their own downstream products or services over those of third parties that operate on their platforms. These practices are commonly referred to as ‘self-preferencing’.[2] Regulators’ concern is that downstream competition requires a ‘level playing field’, but platforms may not provide one when they can play both ‘umpire and player’. Consequently, efforts have been made to deter self-preferencing, both by investigating alleged conduct ex post under competition law and by regulating it ex ante. The latter has resulted in the Digital Markets Act (DMA), which explicitly prohibits certain types of self-preferencing by designated ‘gatekeeper’ platforms.[3]

Concern about self-preferencing has grown in the wake of the Google shopping case.[4] However, the term ‘self-preferencing’ was not actually used in the Google shopping case at all. While the term is used extensively by the practitioner community, it does not refer to a well-defined conduct or specific theory of harm under competition law. Rather, it is a catch-all phrase, encompassing a variety of conducts and concerns. Many of those behaviours are already captured by existing theories of harm, such as refusal to supply and bundling. But some self-preferencing behaviours – such as those relating to prominence and visibility – are prevalent, difficult to capture with existing theories of harm and potentially more problematic in digital markets. The precise meaning of the term, then, and the substantive test for it remain unclear.

In this chapter, we:

  • provide an overview of what might be termed self-preferencing conduct in digital markets;
  • explain why some of the concerns regarding self-preferencing are more pronounced in digital markets;
  • summarise the recent economic insights into the effects of self-preferencing; and
  • discuss the treatment of self-preferencing by competition agencies.

What is self-preferencing?

At the most basic level, ‘self-preferencing’ refers to a platform favouring its own products and services over those of third parties that operate on the platform.[5] This definition is sufficiently broad to include a wide variety of conduct, with no consensus over the exact boundaries.

In principle, the situation is no different from the case of a vertically integrated player competing downstream with third parties. The vertically integrated player may find it beneficial to use one of various types of conduct to favour its own downstream entity over third parties. Many of these are already covered by existing theories of harm in competition law. For example, a vertically integrated company may (1) sell its input only to its own downstream entity, possibly engaging in ‘refusal to deal’; (2) charge itself a price lower than it charges to third parties, possibly engaging in ‘margin squeeze’ or ‘price discrimination’; or (3) favour its downstream entity in distribution, possibly engaging in ‘tying or bundling’ or ‘discrimination’. In many cases, the company’s motivation is the same: to favour its own downstream business. Only the method differs.

In practice, recent cases that commentators have referred to as self-preferencing cases exhibit a wide variety of conduct that allegedly distorts competition downstream (see Table 1, below). Some can easily be characterised using traditional theories of harm, but not all can be.

The first four categories of behaviour listed in Table 1 are the ones most commonly referred to as self-preferencing by industry practitioners. These are all types of non-price conduct where a dominant platform refuses to grant third parties full or equal access to certain aspects of the platform (e.g., by providing inferior access to data or application programming interfaces or by downgrading rankings in search results page). Although the specifics differ, this conduct is somewhat conceptually related to ‘refusal to deal’-type theories of harm. Rather than refusing to provide access to the platform outright, the platform refuses to provide access on competitive terms and therefore places its downstream rivals at a disadvantage.

The last two categories may also be thought of as self-preferencing conduct, insofar as these categories involve the platform privileging the distribution of its own products over competitors’ products. The anticompetitive concerns around these behaviours are closely related to long-established conduct – tying and bundling and margin squeeze – so much so that it is likely that these practices can be considered under existing well-established analytical lenses. For instance, although the tying and bundling concerns arising in the European Commission’s cases against Microsoft long predate the origin of the self-preferencing label, the substantive issues are similar.[6]

Table 1: Taxonomy of self-preferencing behaviours

BehaviourDescriptionExamples of casesPrice or non-price behaviour?Related Article 102 Treaty on the Functioning of the European Union categories
Search results manipulationPreferential ranking or more prominent display in search results pagesGoogle shopping,[7] Amazon - Buy Box[8]Non-priceRefusal to deal/discrimination
Privileged access to dataUse of data gained from third parties on the platform to improve first-party offerAmazon Marketplace,[9] Google Adtech[10]Non-priceRefusal to deal/discrimination
Preferential treatment to customers that use an adjacent first-party productFirst party at one level of intermediation preferring customers that use its first-party serviceGoogle Adtech, Amazon - Buy Box, Apple ATT[11]Non-priceRefusal to deal/tying/bundling
Preferential access to platform featuresWithholding or limiting access of third parties to application programming interface and other features of an operating systemApple Mobile Payments,[12] Italian Enel v. Google case,[13] Apple ATTNon-priceRefusal to deal/discrimination/tying/bundling
Tying and bundlingTying or bundling digital products in adjacent marketsMicrosoft (Media Player, Internet Explorer, Teams),[14] Google Android,[15] Facebook Marketplace,[16] US/UK/AUS Epic litigation[17]Predominantly non-priceTying/bundling
A dual-mode platform operator charging third party for distributionCharging third-party apps a fee for using the app storeApple Music,[18] Netherlands dating apps,[19] US/UK/AUS Epic litigationPriceMargin squeeze

Why have concerns regarding self-preferencing been heightened in the digital economy?

Although recent concern about self-preferencing focuses on the conduct of digital platforms, in principle, self-preferencing is neither new nor specific to digital markets.

In many sectors, vertically integrated intermediaries have long offered their own products as well as those of competitors. Many intermediaries favour the sale of their own products over that of their rivals. For example, supermarkets and department stores must decide how to allocate shelf space and often place their own-brand or private label products in preferential locations compared with manufacturers’ rival brands. In the digital world, we can view demoting third parties’ products and services to the second or third page of search results as an analogous form of self-preferential product placement.

Although self-preferencing exists in the brick and mortar world,[20] it has not triggered actions by regulatory authorities and lawmakers to the level seen in digital markets. The reason for this may be that the scope for harmful brick and mortar self-preferencing is limited by competition between retailers (i.e., competition upstream). If a supermarket or department store had a 95 per cent market share of the upstream market, then we might expect authorities to have more concerns about how that company chooses where to display its own products compared with rivals’ products.

Concern about self-preferencing in digital markets is greater, generally speaking, because it is assumed that upstream competition (i.e., between platforms) is weak and therefore it is less able to discipline a platform’s incentive to self-preference downstream products or the (negative) impact of it doing so. The reason for that presumption is that the specific nature of many digital markets makes them prone to ‘tipping’, weakening competition between platforms and leaving one upstream provider with a very large market share. For instance, many platforms exhibit network effects, economies of scale and endogenous sunk costs and benefit from mass data. These make it more likely that consumers opt for the same provider, and once the market tips, others will find it hard to challenge the platform’s advantages even if they offer superior functionality.

However, the specific nature of a digital market has to be considered. Although it may exhibit features that limit upstream competition and make self-preferencing downstream harmful, it may not. Many digital markets are characterised by high product differentiation and multi-homing – meaning that consumers have multiple routes to downstream products, such that no single ‘gatekeeper’ can unduly influence consumers’ decisions. In the presence of these features, we would not expect the market to automatically tip in favour of a single platform. The impact of self-preferencing in such markets could be similar to its impact in brick and mortar commerce.[21]

Concern around self-preferencing also reflects an additional and growing concern around unfairness. While unfairness has – in some form – always been a part of the assessment of abuse of dominance, it has been generally limited to cases where it reduced the efficiency of the market or had a detrimental effect on contestability. However, given the size and scale of the large digital platforms and the concern that normal competition may not be possible in these markets, there is an additional concern about fairness for its own sake, even when it does not directly reduce consumer welfare.[22]

What does economics tell us about self-preferencing?

Vertical relations have been explored extensively in the economic literature.[23] The consensus is that the effects of favouring the downstream entity depend on the specifics of the case. Where the conduct softens downstream competition and further entrenches the upstream position, the effects are likely going to be anticompetitive. However, where the conduct results in greater efficiency, the effects are likely procompetitive. This is notably likely to be the case when competition is vibrant in the upstream market and therefore disciplines the platform. But this is not a necessary condition. Even with limited upstream competition – which is more likely to be the case in the context of an Article 102 Treaty on the Functioning of the European Union (TFEU) investigation – favouring a downstream affiliate may still prove to be procompetitive (e.g., through the removal of double marginalisation).

These questions have been explored extensively as regards refusal to supply, price discrimination, margin squeeze and tying or bundling.[24] In this respect, the similarity between self-preferencing and existing theories of harm is instructive: lessons learnt can be applied relatively straightforwardly.

In the context of digital markets, economic literature continues to find that the effects of self-preferencing are nuanced and, crucially, whether they are procompetitive or anticompetitive depends on specific circumstances in a predictable way. It is therefore imperative to consider the actual or potential anticompetitive effects of self-preferencing alongside any procompetitive benefits on a case-by-case assessment of the relevant circumstances. The very restraints that are deemed to limit rivals’ ability and incentive to compete may be the same that create competition and improve the functioning of markets (e.g., to build an ecosystem and to increase competitive pressure on competing platforms).

That said, it remains the case that procompetitive effects are more likely when competition is vibrant in the upstream market. They are also more likely when the platform has prospects for further growth, as its primary incentive is to attract more users even absent upstream competition. In that growth phase, the platform may have limited incentive to self-preference its affiliated downstream services to better exploit the base of current (or ‘installed’) users, if that would also deter new users from adopting its intermediation services and slow the growth of the platform.

Some of the recent theoretical research into self-preferencing has focused on the positive and negative welfare implications of (1) first-party selling and (2) self-preferencing by large digital platforms. These are two very different questions with different implications. The first informs the question whether large digital platforms should be prohibited from first-party selling, while the second informs the question whether self-preferencing should be per se prohibited. Consideration has also been given to the introduction of ex ante regulation and how that may have an impact on a platform’s decision to operate as either an open or a closed environment. Research to date offers mixed findings, suggesting the need for a careful and case-by-case approach.[25]

As regards first-party selling, the competitive and consumer welfare impact of a platform’s entry in competition with third-party sellers distributing through that platform is discussed by Hagiu et al. (2022);[26] Dryden, Khodjamirian and Padilla (2020);[27] and Etro (2020).[28] These studies find that entry aims to prod third-party sellers to compete more aggressively. It also allows the platform to improve the quality of its offerings by adding some missing varieties. This, in turn, makes the platform more competitive with other platforms and helps to grow the platform. The studies find that such entry typically enhances consumer surplus, with benefits including lower prices, higher quality and greater variety. The same applies for a platform striving to grow demand for its intermediation services, even in the absence of platform competition. Hagiu and Spulber (2013)[29] discussed the extreme case of a nascent platform, which can face a ‘chicken and egg’ problem – the platform must have attractive downstream products to attract users, but it needs to be a popular platform to attract developers of downstream services. First-party selling can resolve that problem. By offering both upstream and downstream services, it overcomes buyers’ and sellers’ initial scepticism about the potential success of the platform and triggers a virtuous circle: its downstream services attract users to the platform, which attracts developers of other downstream services, which attracts more users to the platform, and so on.

Empirical evidence on the effect of a platform’s competition with third-party sellers distributing through that platform is ambiguous. For instance, using data from the mobile app market, Wen and Zhu (2019)[30] find that after Google’s entry threat increases on Android, affected app developers reduce innovation and raise the prices for the affected apps. However, their incentives to innovate are not completely suppressed; rather, they shift innovation to unaffected and new apps. Similarly, based on data collected in the United States, Zhu and Liu (2018)[31] find evidence that the entry of Amazon into the seller’s product space on Amazon.com reduces innovation incentives of third-party sellers. However, the opposite result is found in Dryden, Khodjamirian, Rovegno and Small (2020).[32] Using a richer data set than Zhu and Liu (2018), covering all products across a wider range of categories and all sellers of those products for France and Germany between June 2016 and June 2018, they find that Amazon’s entry positively affects the size of catalogues of the third-party sellers operating on Amazon, their probability of innovation and the number of new-to-market products they introduce. Foerderer et al. (2018)[33] and Li and Agarwal (2017)[34] also find that platform entry increases innovation and benefits consumers with respect to Google’s entry into the market for photography apps on Android and Facebook’s integration of Instagram, respectively.

The mixed findings of the empirical literature reflect the need for a careful case-by-case approach, taking into account the specifics of the digital market and platform of interest.

As regards self-preferencing, the first question may be to identify whether the conduct constitutes self-preferencing to begin with. In many cases, it may not be straightforward to identify whether there is, for example, a bias resulting from a particular conduct. Aguiar et al. (2021)[35] identify a methodology to assess whether there is a bias in the rankings in the context of the music industry.

Regarding the impact of self-preferencing, Hagiu et al. (2022) discuss the impact of self-preferencing on competition and consumer welfare in the context of competition with third-party sellers distributing through a platform. They conclude that banning self-preferencing intensifies price competition between first-party and third-party offers, resulting in lower prices and higher consumer surplus. Zennyo (2022)[36] finds somewhat opposite results. Self-preferencing allows the platform to steer customers to its own downstream offering and thus earn greater profits from sales of its own products. However, this increases the platform’s expected profit per consumer, so the platform acts in a way to attract more consumers to its marketplace. To this end, it will find it optimal to charge a lower commission to the third parties so that the final prices to the consumers are also lower, which attracts more users to the platform. This study concludes that own-content biases are not necessarily anticompetitive. Rather, they can be regarded as procompetitive in certain circumstances. These results were not obtained in the model of Hagiu et al. (2022), because they assume that the total number of consumers and sellers is fixed. The platform’s ability to self-preference therefore allows it to act in a way that its interests are aligned with those of end consumers, and this is also demonstrated by Anderson and Bedre-Defolie (2021)[37] and Etro (2023).[38]

Again, the important consideration is how self-preferencing affects demand for the platform, whether that is competition between rival platforms for market share or the prospects for a platform to grow the total size of the market. In this respect, the incentive and impact of first-party selling can change with a platform life cycle. Padilla, Perkins and Piccolo (2023)[39] posit that the impact of first-party selling depends on the growth and, hence, the expected future profitability of the platform. When demand growth is slow, the platform may not have the incentive to grow the platform but instead to choose to exploit the installed user base of consumers and foreclose downstream competitors to better extract downstream rents. This is the case even if the platform offers a worse product than rival products on that platform, thereby hurting consumers.

Others have considered the impact of business models on the incentives of platforms to self-preference. Scott-Morton et al. (2020)[40] have argued that a device-funded platform is more likely to internalise the benefits of ecosystem effects, so is less likely to have the incentive to self-preference compared with platforms that are funded by advertisements. While business model may have some relevance, in and of itself, this is unlikely to give a definitive indication of where self-preferencing may be procompetitive versus anticompetitive. Further, Padilla, Perkins and Piccolo (2023) note that business models change over time, so the incentive to self-preference and its impact on competition and consumers can change through a platform’s growth cycle. Again, as growth slows, even platforms that have historically been device funded may benefit more from self-preferencing their own services to earn a greater proportion of the services sold to ‘installed’ users.

Some authors have considered practices by specific platforms. However, the literature is still scarce. Chen and Tsai (2019)[41] focus on product recommendations by Amazon. They highlight self-preferencing in the selection of products recommended by Amazon as ‘frequently bought together’ (FBT) in the product pages viewed by users. They conclude that the steering via Amazon’s FBT algorithm is driven by seller identity rather than consumer preferences. Lee and Musolff (2021)[42] also focus on product recommendations by Amazon, although not FBT. Their results are consistent with those of Chen and Tsai (2019); however, they find that while Amazon displays its own products above those of third-party sellers in its marketplace, doing so results in greater consumer welfare because consumers prefer those products.

Ex ante regulation of self-preferencing

Ex ante regulation of self-preferencing practices has been introduced for large platforms that are designated as gatekeepers.

In the European Union, Article 6(5) of the DMA prohibits self-preferencing, albeit in a narrow sense since it prohibits only practices related to ranking, indexing and crawling services:

The gatekeeper shall not treat more favourably, in ranking and related indexing and crawling, services and products offered by the gatekeeper itself than similar services or products of a third party. The gatekeeper shall apply transparent, fair and non-discriminatory conditions to such ranking and related indexing and crawling.[43]

Given the procompetitive implications arising from certain kinds of self-preferencing, as discussed in ‘What does economics tell us about self-preferencing?’, above, a narrow definition of ex ante regulated behaviour may allow for positive and procompetitive differentiation between first- and third-party products. Still, the per se prohibition established may come at a cost: muting some of these benefits. The question is therefore whether this cost is outweighed by potential benefits of a ‘level playing field’ and curbing the market power of digital platforms.

The DMA’s self-preferencing prohibition is subject to further specification by the Commission and does not allow gatekeepers to provide any objective justifications should the practice generate procompetitive effects. Therefore, there does not appear to be scope for economic considerations of whether the specific self-preferencing behaviour is welfare enhancing or welfare reducing in the narrow scope set out in the DMA.

Rather, economic considerations may thus be relevant only in relation to the proper identification and detection by the Commission of the practices addressed by the DMA, and then for any enforcement actions the Commission may take if a breach of the DMA is suspected.[44] However, it is not straightforward to define what neutral rankings or recommendations, or both, might look like in specific instances. In particular, it can be challenging to detect adverse self-preferencing conduct as opposed to legitimate differential treatment.[45] For instance, there are several legitimate reasons that may cause algorithms to rank the platform’s products higher than those of competitors: they can be cheaper, have vertically differentiated features that consumers value (e.g., priority shipping) or have horizontally differentiated features that are preferred by consumers.

This is a risk to the approach taken by the DMA. Failing to properly identify and detect problematic self-preferencing will lead to costs for gatekeepers, third parties and consumers in the case of over-enforcement, and to consumers and third parties in the case of under-enforcement.

Other parts of the DMA address some of the types of conduct that may be considered self-preferencing in a broader sense. For example, Article 6(10) concerns equal access and use of data between first and third parties, to prevent more favourable treatment of first- versus third-party services, and Article 6(12) concerns fair, reasonable and non-discriminatory terms of access for a subset of core platform services.

In addition to the DMA, ex ante regulation of self-preferencing has also appeared in several other jurisdictions.

  • A new competition tool in Section 19a of the German Competition Act came into force at the beginning of 2021. It opens the way to prohibit self-preferencing practices by platforms, considered to be of ‘paramount significance for competition across markets’ on a case-by-case basis.[46]
  • In the United Kingdom, the Competition and Markets Authority’s (CMA) Digital Markets Unit (DMU) has been set up to investigate and indicate remedies for potentially problematic conduct in digital markets. The United Kingdom’s competition regime also allows entire industries to be investigated on competition grounds using market studies and market investigations. Through these tools, the DMU can then specify bespoke codes of conduct for specific gatekeepers. This could include requirements to not engage in undue self-preferencing of its own services and, if necessary, impose procompetitive interventions such as functional separation remedies to remove self-preferencing incentives altogether.[47],[48] Additionally, draft legislation – the Digital Markets, Competition and Consumers Bill – has been proposed in the United Kingdom to give the CMA further powers to set rules for and regulate digital firms.[49]

Ex post enforcement of self-preferencing

New ex ante regulation aside, there is still ongoing debate around whether self-preferencing constitutes a new and independent theory of harm under Article 102 TFEU and, ultimately, which legal test to use to assess the legality of it.[50]

At present, self-preferencing does not have a firm definition within the context of Article 102 TFEU. The term has been used by policymakers and commentators, but the self-preferencing label itself does not appear to be crucial to decisions in competition law enforcement cases to date.

Three views have emerged when considering how self-preferencing should be treated in ex post enforcement.

  • Refusal to deal or supply only: Proponents of a narrow view of self-preferencing suggest that conduct by a dominant firm should be classified as abusive under Article 102 TFEU only insofar as it amounts to a refusal to deal or supply. In other words, the assessment of self-preferencing depends on the refusal to supply test, including the indispensability criterion, as established in Bronner and later jurisprudence.[51]
  • Broader Article 102 applications: Proponents of a broader view of self-preferencing argue that aside from a refusal to deal, there may be several legal bases under which self-preferencing conduct may have relevance in competition law (e.g., traditional leveraging theories including discrimination, tying, margin squeeze or even unfair pricing).[52]
  • New, stand-alone conduct: Finally, there have also been suggestions that self-preferencing in the specific form that has arisen in the case of digital markets (e.g., in relation to prominence and rankings) should be considered as a stand-alone abuse with its own legal test for ex post enforcement.[53]

Each of the views above is not without its challenges, especially as the various forms of self-preferencing are far from identical, as seen in Table 1, above. Focusing solely on a refusal to deal or supply test may miss a range of conduct that could have detrimental effects on consumers. Broadening the meaning of self-preferencing to include other forms of conduct may instead beg the question whether the self-preferencing label is meaningful if it covers such a wide range of more traditional conduct. There may be some merit in using the label for new forms of problematic conduct relating to ranking and prominence, which may not have been as problematic in the brick and mortar world. Having clarity on the meaning of self-preferencing is particularly important because it will determine whether it is appropriate to use a uniform legal test. It could very well be problematic to use the same legal test for the assessment of practices that are fundamentally different or having several legal tests for similar practices.[54] Therefore self-preferencing may be misleading as a stand-alone abuse as it may obscure the true issues underlying a case.

One criterion to distinguish self-preferencing from broader Article 102 conduct could be the degree to which the dominant firm favours its own products over those of competitors. In this respect, self-preferencing may be distinct from broader ‘tying’ and ‘refusal to deal’ concerns, in that it would refer to conduct where the integrated digital platform allows some form of access to rivals in the downstream market. This would be in contrast to, for example, an outright refusal to deal, where there would be no interaction between the integrated player and the unintegrated competitors, and competitors would not have the opportunity to sell via the integrated platform. This would echo the traditional distinction between partial and total foreclosure, yet classifying self-preferencing this way would have the counterintuitive outcome that the most extreme manifestations of favouring (whereby the firm does not deal in any way with rivals) could be subject to a less demanding legal test than less extreme forms thereof. This is arguably the case if one assesses the refusal to deal through the lens of the Bronner test and imposes meeting the ‘indispensability’ criterion outside the very narrow set of circumstances where it would be justified (i.e., where there are investment incentives to protect).[55] This could reduce the incentive to open platforms to begin with, which would likely not be in the interest of consumers.

One of the reasons for the ambiguity is that the term has been developed and popularised largely by the practitioner community. To date, competition agencies’ and courts’ decisions have been very cautious about referring to self-preferencing as the specific abuse in question: very few do use this term, and, in fact, it does not even appear in the Google shopping decision.[56] In Google shopping, the Commission argued that Article 102 TFEU prohibits abusive leveraging but, nonetheless, it did not qualify Google’s conduct as any known leveraging practice. The Commission did not label this a self-preferencing behaviour, and, indeed, no novel abuse was suggested by the Commission at all.

The General Court shared the Commission’s conclusions in its judgment of November 2021, confirming the Commission’s Decision.[57] Given that leveraging practices of a dominant firm are not prohibited as such by Article 102 TFEU and that the scope of the special responsibility imposed on a dominant undertaking must be considered in light of the specific circumstances of each case, practices may be found abusive that are not included among those already assessed in the decisional practice of the Commission and in the case law of the European Court of Justice. Based on this, the General Court did not rely on a specific test to find Google’s conduct abusive but rather found that the analysis of the specific circumstances of the case is fundamental to assess whether the dominant undertaking’s conduct deviates from competition on the merits, without any objective justification or proof of efficiencies. As such, the Court’s judgment highlights two conditions for self-preferencing to constitute an abuse of dominance: (1) the conduct must have actual or potential anticompetitive effects; and (2) the conduct must depart from what would be expected under normal competition (even if ‘normal’ is not a very clear concept under competition law).

In addition to Google shopping, insights can be drawn from several other recent decisions regarding self-preferencing or closely related conduct.

  • In 2021, the Autorita garante della concorrenza e del mercato (AGCM) fined Amazon for abusing its dominant position. According to the AGCM, Amazon tied the allocation of the Prime badge to third-party merchants to the use of Amazon’s own logistics service, Fulfilment by Amazon (FBA). Sellers with a Prime badge may appear more favourably in product search results on Amazon; Amazon also allows customers to filter for product results from Prime sellers only. The AGCM found that Amazon’s conduct had a dual effect. First, it increased Amazon’s dominance in the market for intermediation services on online marketplaces to the market for e-commerce logistics. Second, it strengthened Amazon’s dominant position by discouraging third parties from selling their goods in other online marketplaces. It is worth noting that, in addition to ‘tying’, the AGCM defined the anticompetitive conduct as ‘self-preferencing’, with explicit reference to Google shopping. The conduct was found to be detrimental to competition in the e-commerce logistics services market, as it was harmful to rival logistics operators, thereby widening the gap between Amazon and its competitors. The AGCM also found that this strategy unduly favoured revenues generated by Amazon with FBA and its growing share of the relevant market, not based on the quality and price of the logistics services but mainly on the benefits offered on Amazon.[58] The European Commission reached a similar conclusion when securing commitments from Amazon in Germany, France and Spain.[59]
  • In 2021, France’s Autorité de la concurrence (FCA) found that Google had implemented two distinct practices aimed at ensuring that its ad server, DFP, favoured its own platform for selling advertising space and, conversely, that its advertising space platform favoured its own ad server, DFP. According to the FCA, these practices may have affected the ability of competitors to develop in the (still emerging) online advertising market. They limited the attractiveness of ad servers and third-party supply side platforms (SSPs) from a publisher’s perspective and have enabled Google to significantly increase its market share and its revenues. They affected not only competitors but also publishers who have been deprived of the possibility of making full use of the competition between the various SSPs.[60]
  • In 2021, Apple was fined by the Dutch Competition Authority (ACM) for abusing its dominant position in the Dutch App Store concerning dating apps. This abuse was based on the imposition of what the ACM deemed ‘unreasonable conditions on dating-app providers’.[61] The ACM argued that Apple’s terms and conditions had significant detrimental effects. They countered Apple’s assertions, highlighting that these terms severely constrained the freedom of choice for dating app providers and users. The ACM asserted that these terms effectively forced dating app providers into a position of dependence on Apple, with no practical means of escape, thus resembling a ‘take it or leave it’ ultimatum. Furthermore, the ACM underscored that Apple’s terms and conditions shifted the customer relationship, making dating app users customers of Apple rather than the dating app providers themselves. This shift had adverse consequences for both customers and dating app providers, hindering app providers’ ability to directly engage with users for customer service, security checks and age verification. Without these terms and conditions, app developers could have managed their own payments for in-app purchases or choose their own payment service provider. Additionally, Apple’s terms and conditions disrupted the direct customer relationship between dating app providers and their users, placing consumers at a disadvantage.[62]

It is clear from these decisions that, although self-preferencing is systematically sanctioned, no unique legal or economic test seems to be systematically implemented in decisions regarding these cases. In addition to these cases, there are currently a number of open cases,[63] and these may shed further light on how consideration of self-preferencing may evolve in the context of enforcement cases. In particular, as the European Commission has provided a (narrow) definition of self-preferencing in the DMA, this may, in turn, have a soft influence on how self-preferencing is considered in ex post cases as well.


From an economic perspective, the self-preferencing label is not well defined. It overlaps with a variety of conduct that is already well established in competition law. The popularity of the label does appear to stem from the wider concerns about digital platforms, in particular that their size and scale mean that it would be unrealistic to always expect ‘normal competition’ to constrain them. This means that there is an additional concern about fairness or ‘levelling the playing field’, even when it does not directly improve consumer welfare.

Self-preferencing has been sufficiently concerning in digital markets that ex ante regulation of some of these practices has recently been introduced. The new ex ante regulation aside, there is still ongoing debate revolving around whether self-preferencing constitutes a new and independent abuse under Article 102 TFEU. That said, from an economic point of view, competition risks associated with self-preferencing do not appear significantly different from those that emerge in other scenarios of vertical and conglomerate integration, although some concerns may be more problematic, given the nature of digital markets. These risks require detailed case-by-case analysis.


1 Guillaume Duquesne is a senior vice president, Thibaut de Bernard and Kadambari Prasad are vice presidents, and Paul Armstrong and Thomas Bowman are senior economists at Compass Lexecon.

4 Case AT.39740 Google Search (Shopping), 27 June 2017.

5 We note that the term ‘self-preferencing’ appears in only one decision from the European Commission (Case AT.40558 and AT.40571 regarding Polish airports) and is not defined. This decision refers to the Google shopping case for more information regarding this theory of harm, although the term ‘self-preferencing’ is not used in the Google shopping case.

6 Case AT.39530 Microsoft (Tying), December 2009, available at https://competition-cases.ec.europa.eu/cases/AT.39530.

7 Case AT.39740 Google Search (Shopping), 27 June 2017, available at https://ec.europa.eu/competition/antitrust/cases/dec_docs/39740/39740_14996_3.pdf.

8 Case AT.40703 Amazon - Buy Box, March 2023, available at https://competition-cases.ec.europa.eu/cases/AT.40703.

12 Case AT.40452 Apple - Mobile payments, June 2020, available at https://competition-cases.ec.europa.eu/cases/AT.40452.

13 Case A529 Google/compatibilità app Enel X Italia con sistema Android Auto, May 2021, available at https://www.agcm.it/dotcmsdoc/allegati-news/A529_chiusura.pdf.

14 Case AT.40721 Microsoft Teams, July 2023, available at https://ec.europa.eu/commission/presscorner/detail/en/ip_23_3991.

15 Case AT.40099 Google Android, July 2018, available at https://competition-cases.ec.europa.eu/cases/AT.40099.

16 Case AT.40684 Facebook Marketplace, June 2021, available at https://ec.europa.eu/commission/presscorner/detail/en/ip_22_7728.

18 Case AT.40437 Apple Music, June 2020, available at https://ec.europa.eu/commission/presscorner/detail/es/ip_21_2061.

20 Dubé, J P (2022). Amazon Private Brands: Self-Preferencing vs Traditional Retailing. Available at SSRN 4205988 and https://www.zbw.eu/econis-archiv/bitstream/11159/524836/1/EBP085836427_0.pdf.

21 Jullien, B and Sand-Zantman, W (2021). ‘The Economics of Platforms: A Theory Guide for Competition Policy’. Information Economics and Policy, 54, 100880.

23 Rey, P and Tirole, J (2007). ‘A Primer on Foreclosure’. Handbook of Industrial Organization, 3, pp. 2145–2220. O’Donogue KC, R and Padilla, J (2020). The Law and Economics of Article 102 TFEU. Bloomsbury Publishing.

24 Fumagalli, C, Motta, M and Calcagno, C (2018). Exclusionary Practices: The Economics of Monopolisation and Abuse of Dominance. Cambridge University Press. Motta, M (2023). ‘Self-Preferencing and Foreclosure in Digital Markets: Theories of Harm for Abuse Cases’. International Journal of Industrial Organization: 102974. https://bse.eu/sites/default/files/working_paper_pdfs/1374_0.pdf. Bougette, P, Budzinski, O and Marty, F (2022). ‘Self-Preferencing and Competitive Damages: A Focus on Exploitative Abuses’. The Antitrust Bulletin, 67.2: pp. 190–207. https://www.zbw.eu/econis-archiv/bitstream/11159/499213/1/EBP080385354_0.pdf.

25 Kittaka, Sato and Zennyo (2023) conducted a survey of the economics literature regarding dual-role platforms and their practice of self-preferencing. Self-preferencing, defined as the act of platforms favouring their own products or services over those offered by third-party sellers within their marketplaces, was the central focus of the paper. The authors reviewed both theoretical and empirical studies, delving into various aspects of self-preferencing, including its forms, incentives and effects. Notably, they found that the impact of self-preferencing on consumers and overall welfare heavily hinges on factors such as market structure, differentiation and the nature of competition. Kittaka, Y, Sato, S and Zennyo, Y (2023). ‘Self-Preferencing by Platforms: A Literature Review’. Japan and the World Economy: 101191, available at https://cms03.jftc.go.jp/cprc/reports/disucussionpapers/r4/index_files/CPDP-89-2-E-R.pdf.

26 Hagiu, A, Teh, T-H and Wright, J (2022),‘Should Platforms Be Allowed to Sell on Their Own Marketplaces?’, RAND Journal of Economics, 53(2), pp. 297–237.

27 Dryden, N, Khodjamirian, S and Padilla, J. (2020), ‘The simple economics of hybrid marketplaces’, Competition, 23(2), pp. 85–99.

28 Etro, F (2020), Product Selection in Online Marketplaces. Available at SSRN 3641307.

29 Hagiu, A and Spulber, D (2013). ‘First-party content and coordination in two-sided markets’. Management Science, 59.4: pp. 933–949.

30 Wen, W and Zhu, F (2019), ‘Threat of platform-owner entry and complementor responses: Evidence from the mobile app market’, Strategic Management Journal, 40(9), pp. 1336–1367.

31 Zhu, F and Liu, Q (2018), ‘Competing with complementors: An empirical look at Amazon’, Strategic Management Journal, 39(10), pp. 2618–2642.

32 Dryden, N, Khodjamirian, S, Rovegno, L and Small, I (2020). ‘Another look at the impact of Amazon Retail’s entry on third-party innovation at Amazon’s Marketplace’. On file with authors.

33 Foerderer, J, Kude, T, Mithas, S and Heinzl, A (2018), ‘Does platform owner’s entry crowd out innovation? Evidence from Google photos’, Information Systems Research, 29(2), pp. 444–460.

34 Li, Z and Agarwal, A (2017), ‘Platform integration and demand spillovers in complementary markets: Evidence from Facebook’s integration of Instagram’, Management Science, 63(10), pp. 3438–3458.

35 Aguiar, L, Waldfogel, J and Waldfogel, S (2021), ‘Playlisting favorites: Measuring platform bias in the music industry’, International Journal of Industrial Organization, 78.

36 Zennyo, Y (2022). ‘Platform encroachment and own-content bias’. The Journal of Industrial Economics, 70(3): pp. 684–710.

37 Anderson, S P and Bedre-Defolie, O. (2021). ‘DP16243 Hybrid Platform Model’. CEPR Discussion Paper.

38 Etro, F(2023). ‘Hybrid Marketplaces with Free Entry of Sellers’. Review of Industrial Organization, 62, pp. 119–148.

39 Padilla, J, Perkins, J and Piccolo, S (2022), ‘Self-Preferencing in Markets with Vertically Integrated Gatekeeper Platforms’, The Journal of Industrial Economics, 70(2), pp. 371–395.

40 Caffarra, C et al. ‘Designing regulation for digital platforms: Why economists need to work on business models’. EUVox, https://voxeu. org/article/designing-regulation-digital-platforms (2020).

41 Chen, N and Hsin-Tien, T (2019). ‘Steering via Algorithmic Recommendations’. RAND Journal of Economics, Forthcoming. Available at SSRN: https://ssrn.com/abstract=3500407 or http://dx.doi.org/10.2139/ssrn.3500407. Available at SSRN 3500407.

42 Lee, K H and Musolff, L (2021). ‘Entry Into Two-Sided Markets Shaped By Platform-Guided Search’. Job Market Paper, Princeton University.

43 Regulation (EU) 2022/1925 of the European Parliament and of the Council of 14 September 2022 on contestable and fair markets in the digital sector and amending Directives (EU) 2019/1937 and (EU) 2020/1828 (Digital Markets Act).

44 Peitz, M. ‘The Prohibition of Self-Preferencing in the DMA’. CERRE Issue Paper (2022), available at https://cerre.eu/wp-content/uploads/2022/11/DMA_SelfPreferencing.pdf.

45 Carugati, C (2022). ‘How to implement the self-preferencing ban in the European Union’s Digital Markets Act’. Bruegel Policy Contribution, Issue 22/22. https://www.bruegel.org/sites/default/files/2022-12/PC%2022%202022_3.pdf.

46 10th Amendment to the German Act against Restraint on Competition, Section 19(a), Paragraph 2.1.

50 Colangelo, G (2023). ‘Antitrust Unchained: the EU’s case against self-preferencing’. GRUR International ,72.6: pp. 538–556. https://reason.com/wp-content/uploads/2022/09/EU-self-preferencing.pdf.

51 Vesterdorf, B (2015), ‘Theories of Self-Preferencing and Duty to Deal - Two Sides of the Same Coin?’, Competition Law & Policy Debate, 1(1), pp. 4–9.

52 Petit, N (2015) ‘Theories of Self-Preferencing Under Article 102 TFEU: A Reply to Bo Vesterdorf, ORBi.

53 AT.40670 - Google - Adtech and Data-related practices.

54 Colomo, P I (2020). ‘Self-Preferencing: Yet Another Epithet in Need of Limiting Principles’. World Competition, 43. https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3654083.

55 Opinion of Mr Advocate General Jacobs delivered on 28 May 1998, available at https://eur-lex.europa.eu/legal-content/EN/TXT/PDF/?uri=ecli:ECLI%3AEU%3AC%3A1998%3A264.

56 Case AT.39740 Google Search (Shopping),27 June 2017.

57 T-612/17 Google and Alphabet v. Commission (Google shopping).

61 The order subject to periodic penalty payments mandates Apple to change the conditions regarding access to the Dutch storefront of the App Store in such a way that dating app providers are also able to use an alternative payment system other than Apple’s (IAP service). The existing conditions prohibit this. In addition, dating app providers must be given the opportunity to refer in their apps to payment options outside the app.

63 AT.40670 Google - Adtech and Data-related practices, AT.40716 Apple App Store Practices (music streaming), AT.40684 Facebook Marketplace, AT.40721 Microsoft Teams, AT.40452 Apple Mobile Payments, US/UK/AUS Epic l itigation.

Unlock unlimited access to all Global Competition Review content