The digital wave has presented an opportunity for countries to leverage their potential, increase efficiencies, promote consumer welfare and create a favourable ecosystem for businesses. It is also true that the wave of innovation and destructive creations have brought the anti-competitive undercurrents in the ocean of online markets. As a result, they have presented several challenges before policymakers to deal with digital markets often characterised as having strong economies of scale, network effects, multi-sidedness, and greater transparency.
In this post, we will try to limit our discussion to algorithmic collusions and their potential anti-competitive effects. We shall try to find out the answer to the question ‘whether the competition law of India is well-equipped to scrutinize the algorithmic collusions?’.
Tacit Algorithmic Collusion: Definition and Concept
Tacit collusion refers to a form of anti-competitive collusion which can be maintained by parties by recognizing their reciprocal interdependence. For tacit collusion, there is no need for explicit agreement among undertakings. Tacit collusions generally do not attract anti-trust scrutiny as the firms have the right to adapt intelligently to their rival’s existing or anticipated conduct. However, the existence of “concurrence of wills” or “meeting of minds” between the parties can turn this intelligent adaptation or tacit collusion into a concerted practice. For the purposes of digital collusion, this “meeting of minds” can be replaced with “meeting of algorithms”.
Now, algorithms can be understood as a well-defined set of instructions that needs to be followed by a computer/processing unit to complete a task. In a digital world, pricing algorithms are used as a set of computational codes which enable it to set the prices automatically and optimize the profits by using big-data.
Now, it must be noted that concentrated markets involving homogenous products are prone to algorithmic tacit collusion as algorithms can easily monitor the collusion and retaliate effectively to any deviation from artificial equilibrium.
Types of Algorithmic Collusions
The competitiors can exchange the competitively sensitive information privately or it may be exchanged publicly, for example by making advance price announcements. This publicly exchange of information is known as “Signaling”. In Container Shipping case, the European Commission accepted the commitments offered by Container Shipping which used to announce publicly its intended future increases of shipping prices for different channels.
There could be four different scenarios wherein algorithmic collusion can take place –
- Algorithms as Enabler – When algorithms are used as a tool for implementing an agreement (fixing the prices) that has already been agreed upon between colluding parties.
- Algorithms as Administrator – When parties offer their product or services on a common platform (administrator) which is governed by algorithms. For example, hub and spoke cartels wherein the ‘spokes’ (manufacturer or supplier) engage in collusion which is facilitated by a common ‘hub’ (service provider or retailer) without any direct contact between the spokes.
- Algorithms as Predictable Agent – When each competitor unilaterally designs its price algorithms in a certain way with the anticipation of similar changes being made by other competitors in the price algorithms. This predictable outcome would lead to tacit collusion or conscious parallelism. For example, if petroleum companies are using the same price-algorithms, a change in the prices of one company may lead to a real-time change in the competitors’ prices as well. Such change in pricing algorithms may depend upon market factors such as changes in input cost, demand, inflation, etc.
- Autonomous Collusion – When independent algorithms used by different companies collude themselves in order to determine the prices or offer a discount.
- Algorithms and Big Data
Firms often collect data on a voluntary basis or they extract it by observing the user’s journey from one website to another or even within a website. The collection of big-data has become indispensable for such websites whose products or services are related to data only e.g., online dating platforms or matrimony platforms. Business models such as Google search engine, are wholly dependent on the collection of big-data. On the other hand, for some businesses data is ancillary but essential for their sustainability e.g., Ola, Uber, and Netflix etc.
The European Commission’s e-commerce sector inquiry (2017) revealed that “53% of the respondent retailers track the online prices of competitors, out of which 67% use automatic software programs for that purpose”. Pricing algorithms when combined with big-data collected by firms may allow them to enter into a tacit understanding with their competitors to charge similar prices and gain supra-competitive profits.
Companies are using big data to provide better services, quality product and ample choices to the consumers. With the technological advancement, the idea of perfect competition, to some extent, is becoming reality with low or no switching cost, no search cost, and increased market transparency, etc. However, when we look at the underlying intentions behind providing these benefits, an algorithmic pattern in the simultaneous rise and fall of prices across the platforms can be observed. This occurs when artificial intelligence itself learns to stabilize the prices with the market situation and when all the firms use the same algorithms. In the Antitrust domain, it is referred to as tacit collusion amongst competitors.
Digital Collusion under Indian Competition Law
Recently, the Competition Commission of India (CCI), vide its order dated 3 June 2021 in the case of Ms. Shikha Roy v Jet Airways (India) Limited and others, has dismissed allegations of cartelisation against five airlines Jet Airways, Spice Jet, Indigo, Go Air (India) and Air India for skyrocketing prices between Delhi-Chandigarh and Delhi-Amritsar routes during Jat community’s agitation. The investigation conducted by Director General (‘DG’) neither reveals any price parallelism/identical pricing of tickets by the airlines nor any concurrence of wills among the airlines. Further, since all the airlines are using different software for pricing of tickets, it seems the CCI might have correctly dismissed all the allegations of cartelization against above-mentioned airlines.
Nonetheless, the CCI has highlighted the potential anti-competitive effects of algorithms by noting that the “widespread usage of algorithms in price determination by individual firms could pose possible anti-competitive effects by making it easier for firms to achieve and sustain collusion without any formal agreement or human interaction”.
Section 3(3) of the Competition Act, 2002 (‘the Act’) explicitly prohibits cartelization or collusion. For a collusion to exist, the existence of an agreement between the parties is sine-qua-non. Notably, the definition of an ‘agreement’ as defined under section 2(b) of the Act is an inclusive one and includes inter alia any arrangement or understanding or action in concert. Therefore, all the cases where human involvement is possible, the provisions in the Act are sufficient to deal with such cases of algorithmic collusion.
As far as any legislative intervention concerning autonomous algorithmic collusion is concerned, the CLRC report refused to suggest any amendment in the Act to deal with the cases of “autonomous algorithmic collusion” by citing the lack of evidence demonstrating the anti-competitive concerns associated with them. However, there are sufficient economic studies suggesting the potential anti-competitive use of algorithms and big-data.
The leading case concerning algorithms is European Union’s (‘EU’) Eturas case. In this case, Eturas was the administrator of the E-TURAS system, a common online travel booking system, on which Lithuanian travel agencies sold their products. Eturas informed the travel agencies that a technical restriction would be placed on the E-TURAS system according to which the discounts offered by travel agencies would be capped at 3% for online bookings.
The Lithuanian Competition Council (LCC) held that Eturas and 30 travel agencies had engaged in a concerted practice of limiting discounts at 3%, by way of implied or tacit assent, and therefore are liable for infringement of Article 101 TFEU. The decision of LCC was challenged before the Lithuanian Administrative Supreme Court, which referred the matter to the European Court of Justice (‘ECJ’) for a preliminary ruling.
The ECJ held that it can reasonably be presumed that travel agencies have participated in a concerted practice as they had failed to distance themselves from the anti-competitive practice. Here, the ECJ followed the Treuhand principle which renders the undertakings liable under Article 101 TFEU for their “passive modes of participation in the infringement”. Therefore, it is prudent for an undertaking to publicly distance itself from the infringement.
The ECJ also pointed out that the criterion of passive participation should not be used as the sole factor rather it should be used in combination with other objective and consistent indicia to establish a rebuttable presumption of awareness. Otherwise, the presumption of innocence will prevail as the mere dispatch of a message cannot be in itself enough to infer awareness of its content. The ECJ also pointed out that the presumption of a participant can be rebutted by “evidence of a systematic application of a discount exceeding the cap in question”.
It is true that algorithms save a lot of time for consumers while finding the desired product from the ocean of asymmetric information available online. Nonetheless, the collection of a large amount of consumer data is a potential threat to consumer’s privacy and can be used to manipulate the consumers’ choices. Therefore, the question which needs to be answered is – where to draw a line between pro-competitive efficiencies and anti-competitive effects of algorithmic calculations? Banning the algorithms cannot be an option instead identifying the solutions compatible with incentives to innovate along with avoiding any anti-competitive practice seems to be a better idea. This approach could help the regulators to sustain the competition as well as to protect the consumers’ interests.
Though the ability of algorithms to make market transparent and maximise the profits could be lucrative for new firms to enter into the market, at the same time the exploitation of big data would raise concerns regarding driving out the existing small firms and in an oligopolistic environment, it may also raise the entry barriers for the new firms and that would make it difficult for these new firms to access the big data and without having access to the big data, new firms will not be in a position to offer desired products and services to the consumers.
Humans enter into ‘agreements’ with the ‘intention’ to collude with competitors to fix or stable the prices but in cases of automatic algorithmic collusion, it is nearly impossible to extract ‘intention’ of the parties as none is present. This algorithmic collusion prima facie seems to be unilateral actions based on systematic and dynamic price algorithms but in essence, are the result of ‘same-minded’ artificial intelligence using consumers’ preferences and behaviours to determine prices to offer a product as per the willingness to pay of consumers which led to maximization in profits through collusion escaping route of anti-trust scrutiny.
When artificial intelligence starts to determine prices via algorithms by self-learning, a pattern in rising or falling of prices can be observed across the platforms. It can be characterized as self-learning in a transparent market occupied by similar-minded agents to optimize profits which may lead to collusion. A computer machine can easily co-operate with other self-learning machines to react in a similar way towards changing market situation, therefore, creating a non-competitive environment for companies which would ultimately be detrimental to consumers.
Keeping the above discussion in mind, the digital markets, if left without any ex-ante regulatory mechanism, may unleash various new forms of digital collusions which would be untraceable for competition authorities.
With the emergence of the Internet of things, Artificial Intelligence, and reduction in the cost of data collection, businesses have been undergoing a digital transition and are relying on algorithms to determine prices for products and services. The ability of artificial intelligence and algorithms to self-learn while determining the prices of products can dilute the anti-trust liability as it would be difficult to trace the collusion.
Some scholars have argued that the algorithmic collusions are nothing but old wine in a new bottle. However, in view of the author, a ‘digital collusion’ or ‘tacit algorithmic collusion’ cannot be evaluated on the same anti-trust pedestal on which traditional agreement-based collusions are assessed. The existing competition law framework of India is even though capable of countering the anti-competitive effects of algorithmic collusions where human involvement is possible, there is a need to change the approach. The CCI should focus on economic-based approach rather than preponderance of probabilities to tackle the algorithmic collusions. The CCI can use the tools such as algorithm auditing, collusion incubators to trace algorithmic collusion.
Furthermore, regulatory measures in the form of ex-ante regulations are needed to tackle the autonomous algorithmic collusions. The CCI has recently released a discussion paper concerning ‘blockchain technology and competition’ which highlights the potential anti-competitive effects of blockchain in the form of collusion and abuse of dominant position. A similar study on the potential abuse of algorithms is the need of the hour to solve this anti-trust riddle wrapped up in an algorithmic enigma.
Preferred Citation: Dhiman, Vanshaj., “ANTI-TRUST RIDDLE WRAPPED UP IN ALGORITHMIC ENIGMA”, The Law Culture (2021)