Pricing algorithms can be deployed as a tool to facilitate collusion and therefore cause consumer harm. Where this happens, the conduct may not necessarily involve a “contract,” “arrangement” or “understanding” which had to be proven to establish an infringement of the former section 45 of the Australian Competition and Consumer Act 2010 (Cth). In November 2017, the CCA was amended to prohibit concerted practices that have the purpose or effect of substantially lessening competition. However, Australian competition law lacks a definition of the term “concerted practices.” The judicial interpretation of that term will be decisive for the scope of the prohibition and the types of algorithmic collusion it may capture. This paper briefly analyses the extent to which the new prohibition of concerted practices may capture algorithmic collusion.
By Baskaran Balasingham1
I. INTRODUCTION
Pricing algorithms are often used by online retailers and e-commerce platforms. The data-driven innovations including price monitoring and efficient price discrimination that stem from pricing algorithms can enhance consumer welfare. On the other hand, pricing algorithms can be deployed as a means to facilitate collusion and can therefore cause consumer harm. Until not long ago it was cumbersome enough for the Australian Competition and Consumer Commission (“ACCC”) to establish an anticompetitive agreement when it was solely about conduct of human beings. How much more difficult will it be now that artificial intelligence and machine learning have entered the picture? In November 2017, Australia’s Competition and Consumer Act 2010 (Cth) (“CCA”) was amended to prohibit concerted practices. In the wake of the entry into force of the Competition and Consumer Amendment (Competition Policy Review) Act 2017, the Chairman of the ACCC, Rod Sims, expressed confidence that Australia’s competition law, particularly with the addition of the prohibition on concerted practices, can successfully deal with algorithms that harm competition.2 In this paper I will briefly assess how the reformed CCA is likely to cope with algorithmic collusion.
II. INTRODUCTION OF THE CONCERTED PRACTICES PROHIBITION TO CLOSE THE ENFORCEMENT GAP
A. The ACCC’s Struggle to Establish Collusive Conduct
Before the latest reform (the “Harper Reform”), the CCA did not apply where two or more firms engaged in a collusive conduct that had the purpose, effect, or likely effect of substantially lessening competition, unless the parties had the necessary mental element to collude in the form of a “contract, arrangement or understanding” (“CAU”). In ACCC v. Leahy, Justice Gray explained that the terms contract, arrangement and understanding encompass a “spectrum of consensual dealings.”3 These concepts are considered to be related, overlapping, and along a scale of formality – with “contract” at the higher end and “understanding” at the lower end, and “arrangement” somewhere in between.4 An “arrangement” lacks some of the essential elements of a “contract,” and if not express negotiation, it must at the very least involve express communication between the parties.5 The word “understanding” is meant to connote a less precise dealing than either a contract or arrangement.6 Unlike an arrangement, an understanding can be tacit so long as there is a meeting of minds.7 While arrangements and understandings are intended to catch dealings which are less formal and may not give rise to a legally binding contract, courts have concluded that both concepts require three elements: communication, consensus, and commitment. Communication may be implicit in some act rather than express words, and can be as informal as a wink or nod.8 There must be consensus as to what is to be done by at least one party. The most controversial and problematic requirement for establishing an arrangement or understanding is to show that at least one party made a commitment to act in a certain way. It is sufficient if the commitment involves a moral obligation, or obligation binding in honour only, but independently held hopes or beliefs will not be enough.9 The ACCC argued that the courts’ interpretation of the commitment element has made it difficult to prove the existence of a CAU.10
Provided the requisite mental element is established, section 45 requires determining whether the agreement has the purpose, effect or likely effect of substantially lessening competition (“SLC”) in a market. The SLC test is not defined in the CCA, and despite its use in various provisions of the Act, its meaning is still rather vague.11 Case law has provided limited guidance as to the concept of “substantiality.” So far, the most useful clarification is that “substantial” means “meaningful or relevant to the competitive process.”12 Apart from limited guidance, there is also inconsistency in the approaches adopted by the courts in assessing whether conduct has a substantial effect.13 The courts have given more guidance as to the meaning of “purpose,” “effect” and “likely effect.” They have interpreted “purpose” to involve an enquiry into the subjective purpose of the parties to the CAU.14 This subjective purpose needs to be ascertained from the evidence of the mental state of the parties responsible for the CAU, although inferences may also be made from a party’s conduct. Where the purpose of an agreement cannot be established it will suffice to show that it has the effect or likely effect of SLC. This lowers the threshold for satisfying the SLC test considerably, particularly as evidence about firms’ intent to collude may be hard to obtain. Yet overall, it is difficult to gauge whether the SLC test is more likely to lead to an overreach or underreach of section 45. Even though the SCL test does not balance anticompetitive harm and welfare-enhancing efficiencies, the possibility for parties to apply for an authorisation to the ACCC limits the risk of false positives. And while the SLC test may be an enforcement hurdle in cases where the coordination of conduct is manifestly anticompetitive, the ACCC’s little success in bringing cartel cases (in particular involving information-sharing and price signalling) came largely down to its inability to establish the commitment element.
In order to cover the limitations of the CCA, price signalling laws were introduced in 2011. They comprised a per se prohibition on the private disclosure of pricing information to competitors, as well as a prohibition on the public price disclosure that were made to substantially lessen competition. The price signalling legislation was only applied to banking services and included a number of exceptions. Apart from this rather limited scope, the effectiveness of this legislation was further constrained by the complex drafting of the provisions, which were highly prescriptive, in an attempt to exclude pro-competitive or benign price disclosure from the ambit of the prohibitions.15 These shortcomings eventually led to the failure of the price signalling regime which was demonstrated by the fact that no proceedings were ever instituted by the ACCC under these laws.
B. New Prohibition of Concerted Practices
In 2013, the Harper Committee launched a root and branch reform of the CCA where it among other things recommended that the price signalling laws should be repealed and that section 45 should be expanded to apply to contracts, arrangements, understandings and concerted practices. The government accepted these recommendations in the Competition and Consumer Amendment (Competition Policy Review) Act 2017. Section 45(1)(c) of the CCA now provides that “a corporation must not … engage with one or more persons in a concerted practice that has the purpose, or has or is likely to have the effect, of substantially lessening competition.”
Unfortunately, the amended CCA does not define the term “concerted practices.” In the Guidelines on concerted practices the ACCC sets out how it currently proposes to interpret section 45(1)(c) but they also lack a definition. The interpretation of the term is ultimately the responsibility of the Australian courts.16 The Explanatory Memorandum to the Competition and Consumer Amendment (Competition Policy Review) Bill 2017 explains that a legislative definition was contemplated, but ultimately not adopted. It reasoned that a legislative definition would potentially increase certainty as to what does and does not constitute a concerted practice, but this would require a focus on determining whether a certain type of conduct fits within detailed technical provisions, rather than making a principled assessment of the conduct as a whole. The Committee was concerned that this approach might inadvertently exclude conduct that should be treated as a “concerted practice.” Instead, it decided to provide additional guidance as to the typical features of a concerted practice whilst also leaving scope within section 45 to apply the concept to new forms of anticompetitive conduct as they arise. According to the Committee this option increases certainty and at the same time lowers the risk of inappropriately limiting the scope, thus retaining a “flexible and principled application of the concept.” On the one hand, this approach leaves businesses with some uncertainty until the courts come up with an interpretation. On the other hand, it provides the ACCC with flexibility as to how to approach the concerted practices prohibition.17
The Explanatory Memorandum characterises a “concerted practice” as “any form of cooperation between two or more firms (or people) or conduct that would be likely to establish such cooperation, where this conduct substitutes, or would be likely to substitute, cooperation in place of the uncertainty of competition.”18 This description resembles the prohibition on concerted practices under EU competition law. While the European concept has been criticised for being “nebulous,”19 there is at least some useful interpretation and guidance by the European courts. The explanatory material for the Exposure Draft version of the Competition and Consumer Amendment (Competition Policy Review) Bill 2017 expressly states that the interpretation of the concept should be “informed by international approaches to the same concept, where appropriate.”20 As discussed below, some interpretations from EU case law may be helpful in the Australian context.
Also highly relevant is the distinction between a concerted practice and the other concepts of section 45(1), as well as the object and purpose of this new concept. It is clear that a concerted practice is now the lowest form of consensual dealing. The Explanatory Memorandum sets out that a concerted practice is intended to capture conduct that falls short of a CAU.21 Emphasising the distinction between a concerted practice and an understanding, the Explanatory Memorandum notes that the former does not require any commitment. It states that a “concerted practice may exist even if none of the parties is obliged, either legally or morally, to act in any particular way.”22 The Guidelines on concerted practices confirm this point.23
Absent the commitment element to establish a concerted practice the ACCC’s pre-amendment difficulties in proving anticompetitive collusion will be significantly alleviated. Without the commitment element there is no need for reciprocal obligations between the parties to a concerted practice. Although the Full Federal Court in Morphett Arms Hotel v. TPC found that an understanding does not require reciprocity,24 there is still uncertainty whether a unilateral obligation is sufficient for finding an understanding. In the Australian Egg Cartel proceedings between 2016 and 2017, i.e. prior to the amendment entering into force, the ACCC failed to establish an arrangement or understanding between an association of egg producers and two of its members. The trial judge held he was not persuaded, at least to the requisite degree of persuasion, that the respondents intended the conduct in question come about pursuant to an agreement or understanding involving reciprocal obligations.25 Courts have expressed doubt that a case which involves only unilateral conduct would be difficult to envisage. Since no case to date has established an understanding without reciprocity between the parties, it has been suggested that an understanding cannot be unilateral and must involve mutual obligations.26 Before the Harper Reform mutual commitment was considered crucial for the distinction between conscious parallelism and an understanding.27 Under the current law a concerted practice needs to be distinguished from parallel conduct.28 The distinguishing factor is not commitment but consensus between the parties. This becomes evident from Justice Lockhart’s opinion in TPC v. Email where he highlighted that in the absence of commitment, communications were not enough to give rise to the requisite meeting of minds. His Honour found that a credible explanation from a defendant may be sufficient to negate an inference from circumstantial evidence of an arrangement or understanding.29 Following the case law of the European courts parallel conduct may amount to a concerted practice where there is positive contact between two or more parties; it involves cooperation contrary to normal competitive processes; and the effect is to alter or maintain the parties’ commercial conduct. Moreover, according to EU case law, a defendant may provide a plausible explanation to refute a concerted practice. Finally, while it may indeed be difficult to envisage an understanding that only involves unilateral obligations, EU case law as well as the examples of a concerted practice in the Explanatory Memorandum and the Guidelines on concerted practices illustrate that there is no requirement for mutual obligations to establish a concerted practice.
III. ALGORITHMIC COLLUSION UNDER THE CCA
In their seminal work Ezrachi and Stucke identify four potential scenarios in which the use of algorithms may raise antitrust concerns.30 The level of sophistication of algorithms and their use in the market, as well as the challenges for competition law enforcement between these four scenarios differ largely. In the first two scenarios algorithms are used to facilitate humanly agreed upon collusion, whereas in the last two scenarios the algorithms are more advanced and their use may result in tacit collusion. In the first scenario, the “Messenger” scenario, humans agree on colluding and use algorithms to monitor and enforce the agreement. This scenario is the least problematic from an antitrust enforcement perspective. Even before the Harper Reform this scenario could have been dealt with under the CCA as the requirement of a CAU is likely met. Weaker forms of communication that do not involve a CAU, can now be caught by the concerted practice prohibition. In the second scenario, the “Hub and Spoke,” the parties to the collusive conduct exchange sensitive information using a single pricing algorithm to determine their market prices. In the third scenario, the “Predictable Agent,” each firm unilaterally programs its algorithm to monitor and react to price alterations. Finally, in the fourth scenario, the “Digital Eye,” the algorithms are self-learning and able to optimise their performance by reflecting on past decisions in order to maximise profits. This scenario falls outside the scope of antitrust enforcement as the algorithms engage in conscious parallelism without the intent or knowledge of corporate actors. In order to illustrate the scope of the new concerted practice provision, along with some of the above-mentioned pre-amendment enforcement difficulties, the discussion on the application of the CCA to algorithmic collusion in this section is limited to the second and third scenarios.
A. The Hub and Spoke
In a classic hub and spoke arrangement multiple suppliers collude by using a common downstream firm as distributor to facilitate their alignment of their behaviour. In e-commerce, cartel participants can achieve a similar result by using a single algorithm. The algorithm adopted by the parties functions as the “hub” to facilitate collusion amongst its common users, the “spokes.” Although no court in Australia has ever explicitly confirmed the existence of a hub and spoke cartel, the potential for competitors to form an anticompetitive agreement through a third party was recognised in News Ltd v. Australian Rugby League Ltd (No 2). In this case the Full Federal Court found support for a hub and spoke agreement, suggesting that a horizontal arrangement or understanding between the spokes can be established indirectly by demonstrating the vertical contact between the spokes and the hub.31 Similarly the Explanatory Memorandum clarifies that a concerted practice “may be established in the absence of any direct contact between the firms, for example where firms communicate indirectly through an intermediary such as a peak industry body.”32
In the Australian Egg Cartel case the trial judge drew a distinction between two situations that might be relevant in the near future for the interpretation of the concepts of concerted practices and hub and spoke arrangements. In the first situation industry participants are being brought to an appreciation of what is in their interests, independently of what others are doing, in order to act in a certain manner. At the time of the proceedings this situation did not contravene the CCA. In the second situation industry participants are being invited to agree to act in a certain way in the expectation of reciprocal conduct by others. Such conduct was and still is in breach of the CCA.33 Post-amendment the first situation could potentially be caught under the new section 45(1)(c). A hub and spoke arrangement would exist if the association were to communicate with each egg producer individually to exchange sensitive commercial information rather than the latter directly communicating with each other. Even if the egg producers only agreed to using a common third party and did not necessarily agree to fix prices themselves, there would still be a concerted practice as their behaviour amounts to cooperation. Despite the lack of a commitment, the exchange of sensitive commercial information via the third party suffices to reduce uncertainty and may facilitate alignment of the companies’ competitive behaviour. Likewise, in the context of automated systems agreeing to use a single algorithm that performs the tasks of the third party may give rise to a concerted practice.
Apart from establishing the necessary mental element, the ACCC needs to show that the conduct has the purpose, effect or likely effect of substantially lessening competition. This may be another obstacle as became clear in ACCC v. Air New Zealand, which so far is the only case that addressed in detail whether an arrangement or understanding involving information sharing would have the effect or likely effect of SLC. In this case the ACCC alleged that Air New Zealand exchanged information about future surcharge pricing intentions with other airlines through surveys and meetings conducted by an industry association. The trail court stated that the exchange of future pricing intentions would not always substantially lessen competition, noting that “something needs to be shown aligning the exchange of future pricing intentions with a likely drop in competition.”34 In cases where the use of a single algorithm among various parties is manifestly anticompetitive the threshold of the SLC test may raise an issue of underreach of the concerted practices prohibition.35
B. The Predictable Agent
The Predictable Agent scenario poses even greater challenges for antitrust enforcement than the Hub and Spoke. In this scenario, each firm develops its own algorithm and programs it to monitor market prices and maximize the firm’s profits. The algorithms are also likely to engage in predictive analytics to build forecasts by studying real-time, historical and third-party data. Under certain market conditions, the industry-wide use of such algorithms may transform the market dynamic to effectively enable conscious parallelism.36 This may bring about coordination in one of two ways. The algorithms can swiftly react to competitor’s price reductions reaching a common understanding that arises not as a result of explicit negotiation but because each algorithm is programmed to quickly detect price cutting and adjust prices in order for the firm to remain competitive. Alternatively, the algorithms may also explore price increases as long as they are sustainable, i.e. when others follow in a timely manner so that no competitor benefits from keeping prices lower.
Even though the corporate actors in this scenario do not explicitly agree to adopt a profit-maximising algorithm nor to fix prices, they know that price coordination is the likely outcome if each firm adopts that type of algorithm.37 This scenario may therefore be treated as tacit collusion rather than conscious parallelism. Without any commitment to adopt a profit-maximising algorithm there cannot be a CAU. However, the conduct could now be considered a concerted practice. There is arguably a meeting of the minds between the firms as they know that collusion will only be effective if profit-maximising algorithms are adopted industry-wide or at least by a large number of firms in the market. Otherwise, the algorithms would match the prices of those firms that have not adopted a profit-maximising algorithm, potentially leading to competitive pricing in the market.
As mentioned above, not having to prove commitment nor reciprocity lowers the threshold for establishing a concerted practice. On the other hand, this could lead to an overreach where the concerted practices provision erroneously captures innocent parallel conduct. A safeguard against the overreach might be the SLC test. As mentioned above, satisfying this test is not a straightforward task. Provided the ACCC can show an effect or likely effect of a substantial lessening of competition defendants should be afforded the possibility to adduce a plausible alternative explanation for their parallel conduct.
IV. CONCLUDING REMARKS
Before the Harper Reform, the CCA only prohibited contracts, arrangements or understandings that had the purpose, effect or likely effect of substantially lessening competition. The ACCC was often unsuccessful in proving anticompetitive agreements involving information sharing and price signalling as it failed to establish the commitment element, i.e. that the parties were legally or morally obliged to act in a particular way. The prohibition on concerted practices which was introduced in the Competition and Consumer Amendment (Competition Policy Review) Act 2017 does not require the commitment element. As a result, the necessary mental element for establishing collusive conduct has been lowered significantly which should also benefit the ACCC in proving certain instances of algorithmic collusion. However, even where the ACCC manages to establish the mental element, the SLC test could potentially be a hurdle for successful enforcement.
1 Lecturer, Macquarie University, Macquarie Law School, Sydney, Australia; email: baskaran.balasingham@mq.edu.au.
2 https://www.accc.gov.au/speech/the-accc%E2%80%99s-approach-to-colluding-robots.
3 ACCC v. Leahy Petroleum Pty Ltd [2007] FCA 794, para. 24.
4 Caron Beaton-Wells & Brent Fisse, Australian Cartel Regulation: Law, policy and practice in an international context (CUP 2011), 39.
5 ACCC v. Leahy, para. 26.
6 Ibid. para. 27.
7 Ibid. para. 28.
8 Ibid. para. 26.
9 Ibid. para. 940.
10 ACCC, Petrol prices and Australian consumers: Report of the ACCC inquiry into the price of unleaded petrol (2007), 228.
11 Rob Nicholls & Brent Fisse, “Concerted practices and algorithmic coordination: Does the new Australian law compute?” (2018) 26 Competition & Consumer Law Journal 82, 89.
12 Rural Press Ltd v. ACCC [2003] HCA 75, para. 41.
13 Kathrine Kemp, Misuse of Market Power: Rationale and Reform (CUP 2018), 181.
14 News Ltd v. South Sydney District Rugby League Football Club Ltd [2003] HCA 45, paras. 18; 46; 65; 216.
15 Lindsay Foster & Hanna Kaci, “Concerted practices: A contravention without a definition” (2018) 26 Competition & Consumer Law Journal 1, 2-3.
16 Guidelines on concerted practices, p. 2.
17 Rob Nicholls & Deniz Kayis, “Concerted practices contested: Evidentiary thresholds” (2017) 25 Competition & Consumer Law Journal 125, 128.
18 Explanatory Memorandum, para. 3.19.
19 Ioannis Lianos & Valentine Korah, Competition Law: Analysis, Cases, & Materials (OUP 2019), 388.
20 Commonwealth of Australia, Explanatory Materials, Exposure Draft – Competition and Consumer Amendment (Competition Policy Review) Bill 2016, para. 3.18.
21 Explanatory Memorandum, para. 3.21.
22 Ibid. para. 3.22.
23 Guidelines on concerted practices, paras. 1.2 and 3.3.
24 Morphett Arms Hotel Pty Ltd v. TPC (1980) 30 ALR 88, para. 91.
25 ACCC v. Australian Egg Corporation Ltd, para. 403.
26 Michael Gvozdenovic, “Concerted practices and statutory interpretation: An affirmation of the jurisprudence on ‘contracts. arrangements and understands’” (2019) 26 Competition & Consumer Law Journal 213, 224.
27 Caitlin Davies & Luke Wainscoat, “Not quite a cartel: Applying the new concerted practices prohibition” (2017) 25 Competition & Consumer Law Journal 173, 177.
28 Explanatory Memorandum, para. 3.25; Guidelines on concerted practices, para. 3.6.
29 TPC v. Email Ltd (1980) 31 ALR 53, 56-57.
30 See Ariel Ezrachi & Maurice E. Stucke, Virtual Competition: The Promise and Perils of the Algorithm-driven Economy (HUP 2016), 36-37 and 39-81.
31 News Ltd v. Australian Rugby League Ltd (No 2) (1996) 64 FCR 410, para. 610.
32 Explanatory Memorandum, para. 3.22.
33 Ibid. para. 75.
34 ACCC v. Air New Zealand Ltd (2014) 319 ALR 388, para. 1107.
35 Nicholls & Fisse (2018), 96.
36 Ezrachi & Stucke (2016), 61.
37 Ibid. 77.
