Introduction
The past forty years of the Anthropocene have witnessed exponential growth in the complexity and deployment of algorithms across various fields of human interest. With
specific reference to digital markets, algorithms exert influence on two aspects of human behaviour: first, choice architecture, which is primarily concerned with optimising the manner in which products and services are presented to potential consumers; and second, pricing, through which entities attempt to strike a balance between, maximised
profits and assured gains with methods such as discriminatory pricing. While the usage of algorithms can certainly be beneficial (giving rise to pro-competitive effects and efficiency gains), they could also encourage anticompetitive behaviours in manifold ways: direct consumer harm, entrenchment of market dominance, and price collusion. The implications of algorithmic collusion and its anti-competitive effects have
been explored comprehensively- however, the abuse of dominance engendered and abetted by algorithms has failed to receive as much attention. As a subset of algorithmic abuse of dominance, the relationship between algorithms and self-preferencing behaviours exhibited by digital platforms with significant market power is one that has not been sufficiently considered by antitrust regimes across the world. Given that self-preferencing has only recently come into the limelight vis-à-vis
digital platforms, regulators are now attempting to govern the usage of self-preferencing (mostly by gatekeepers or their substantive equivalents in other jurisdictions) through existing legislations, as well as newer legislations such as the Digital Markets Act and the American Innovation and Choice Online Act (“Bill”).
Self-preferencing can have exclusionary and/or exploitative market effects. By employing various algorithms such as search algorithms, recommendation algorithms, ranking algorithms etc., firms are now able to transition to data-driven positions of dominance partly with the aid of self-preferencing. This is particularly important as, in digital markets, network effects further entrench the impact of abuse of dominance as they often create self-reinforcing feedback loops, which enable algorithms to be more
efficient in their self-preferencing behaviours. This article therefore sets out an argument for why algorithmic abuse of dominance through self-preferencing demands more attention from policymakers and regulators. The first section of this article (Part-I)
provides an overview of self-preferencing, which is followed by a brief description
of cross-jurisdictional responses to self-preferencing (focusing on the EU,USA, and India). The subsequent section (Part-II) explicates the problems posed specifically by the exploitative effects of self-preferencing, and highlights legal and economic factors that need to be taken into consideration in the process of delineating dynamic and effective regulatory responses to the anticompetitive effects of self-preferencing. Since it is inevitable that the modalities of self-preferencing will continue to evolve in the years to come, this article also makes the argument that regulatory responses to self-preferencing must endeavour to account for pro-competitive effects in their analyses.
I. An Overview of Self-Preferencing
Self-preferencing, as the name suggests, is a term used to refer to behaviours whereby an entity treats its own products and/or services in a more favourable manner than those of others. Through manipulation of choice architecture– which in most cases focuses on ranking of products/services- the algorithm is able to nudge consumers towards its own products rather than those which would objectively be of more use to consumers. As a form of algorithmic unilateral conduct, self-preferencing cases comprise the majority of enforcement cases relating to algorithms; but as an absolute number, they are still not numerous.
Self-preferencing can be exclusionary or exploitative. In its exclusionary form,
self-preferencing can impede competitors’ access to business ecosystems through
various means. For instance, the accumulation of data pertaining to the products/services offered by complementors, and the corresponding consumer data,
allows retail platforms such as Amazon to impose an asymmetrical relationship onto competitors (on the retail front) who seek access to consumers (on the marketplace front). This is then acted upon through the skewed usage of algorithms such as ranking
algorithms and those which facilitate price discrimination. For example, most Amazon
users would notice the ‘Frequently Bought Together’ feature which is exclusive to Amazon. With the aid of large datasets and the consumer’s current choice, Amazon’s
algorithms have been shown to direct consumer attention towards paired products not based on consumer choice, but based on economic viability for Amazon. Even if there
are products (complimenting products) sold by others (complementors) which would be more useful to consumers, Amazon effectively promotes its own products to the detriment of these third-party retailers.
Another pivotal example of such exclusionary self-preferencing behaviour can be seen in Google and Alphabet v. Commission (“Google Shopping”). As the first instance of digital
platforms coming under legal scrutiny exclusively for self-preferencing and its anticompetitive effects, Google was held liable for abuse of dominance in light of its usage of self-preferencing algorithms to favour its own comparison service. In Google Shopping, the General Court laid down two criteria for self-preferencing to constitute an abuse of dominance: first, anticompetitive effects (actual or potential) must arise as a result of the conduct in question; and second, this conduct must not be the same as
what can be expected under conditions of normal competition. The contours of Article 102 of the Treaty on the Functioning of the European Union were thus broadened significantly with this holding of Google’s self-preferencing as a standalone abuse of dominance. However, in Servizio Elettrico Nazionale, it was reaffirmed that an effects-based analysis is more central to competition law analyses of whether any conduct is abusive.
II. Regulatory Responses to Self-Preferencing
In the European Union, the Digital Markets Act (“DMA”) has adopted a rather stringent approach towards self-preferencing by way of Article 6(5). Gatekeepers are precluded from treating their own products/services more favourably than those of third parties. As an outright ban on self-preferencing by gatekeepers, this is problematic for three reasons: first, the range of behaviours which will be considered self-preferencing has not been delineated by the DMA; second, there is no scope to provide relevant justifications in favour of why self-preferencing was undertaken; and third, it does not address the issue of self-preferencing by entities who are not designated gatekeepers but nevertheless possess market power sufficient to have a detrimental impact on consumer welfare. The DMA also does not require disclosure of algorithms used in self-preferencing- for instance, in Google Shopping, Google used adjustment algorithms to promote the output of competitors’ search services less favourably than its own. For the purposes of ascertaining whether there has been a demotion of competitors’ search results, Article 6(5) seems to be inadequate vis-à-vis equipping regulators with the tools they need to accurately conduct an in-depth inquiry.
In the USA, self-preferencing has typically been addressed by existing antitrust legislations which apply to platforms both digital and traditional. The existing framework imposes various restrictions on self-preferencing, but they are applicable only if the firm has market dominance vis-à-vis the particular good/services in question, and if competitive harms arise as a result of this refusal to extend equal treatment. However, these changes with the proposed American Innovation and Choice Online Act (“AICOA”) which seeks to prohibit digital gatekeepers (termed as “covered platforms”) from implementing self-preferencing tactics. These covered platforms are identified on the basis of the overall size (given by number of users). Moreover, AICOA does not generally require proof of competitive harm or market power.[1] As a result of the focus on digital platforms, firms with lesser online presence such as Walmart are not subjected to the same extent of scrutiny as firms like Amazon- even if they might be engaging in self-preferencing behaviours. With the lack of product-specific inquiries required by AICOA, there is an increased likelihood of false positives arising as a result of general assessments.
In India, while there is no specific regulatory focus on self-preferencing, a recent report by the Standing Committee on Finance has highlighted the need to tackle, amongst other issues, self-preferencing by systemically important digital intermediaries (“SIDIs”). Similar to the P2B Regulations in the EU, the Committee has recommended that “query, click and view data” generated by users of SIDIs must be made available on request by third-party undertakings. It has also made an expansive recommendation that “must not favour its own offers over the offers of its competitors when mediating access to supply and sales markets”-again, this does not provide the scope for undertaking an effective analysis of the resultant procompetitive and anticompetitive effects. The Digital Competition Bill, 2024 also focuses on self-preferencing by Systematically Significant Digital Enterprises on two counts: with related parties, and third parties with whom arrangements have been made. With SSDEs being identified on the basis of user base and turnover thresholds, it is apparent that there is more attention devoted to structuralconcerns rather than market effects. No test has been established to ascertain whether the anticompetitive effects areoutweighed by the pro-competitive effects.
[1] It is to be noted, however, that as an affirmative defence, it can be shown that
there has been a failure to generate material harm to competition.
This article is authored by Mathanki Narayanan and Srinithi Shankari , 4th year and 5th year law students at OP Jindal Global Law School and Symbiosis Law School, Hyderabad respectively.
Part II of this article can be accessed here
