Introduction

1Gig platforms are a modern economy enterprise structure characterized by a firm matching service providers with consumers – prominent examples include ride-sharing platforms, like Uber; delivery platforms, like Wolt; and lodging rental platforms, like Airbnb. As are all online platforms, gig platforms are data-driven business models that employ and develop algorithms and AI tools that learn from user behavior and adapt to make interactions increasingly efficient.¹ In contrast to other online platforms, such as advertising exchanges or online marketplaces for goods, gig platforms enable users to sell their labor or services to other users via the platform.²

2Scholarship has shown enterprises that contract with their service providers, who are then by necessity operating as independent enterprises, are best analyzed as agreements implicating Art. 101 TFEU and Section 1 of the Sherman Act.³ Currently, the dominant legal treatment of service providers on platforms including Uber (ride-sharing) and Wolt (food delivery) is as contractors rather than employees.⁴ We employ here the lens of a hub-and-spoke arrangement,⁵ with the platform as the hub and the service providers as the spokes, and the algorithmically-established price terms representing a collection of parallel vertical agreements. We then engage in a comparative study of the legal implications under antitrust law in the US and the EU of hub-and-spoke arrangements.

3The chapter proceeds to outline the hub-and-spoke structure of the service provider-platform agreements in a gig economy enterprise, including the universal agreement to abide by prices set by algorithm in contracting for services. It covers various design options for pricing algorithms that can be used by platforms to coordinate the transaction between its users. Next, the chapter considers the EU caselaw on hub-and-spoke arrangements, analyzing authorities from across the EU, and identifies the probable treatment of the gig economy agreements in the light of these authorities. The chapter then conducts a similar analysis of leading recent authorities in the US and likewise concludes the most probable treatment under US law. In the conclusion, the chapter compares and explains the likely legal treatment of an algorithmically defined hub-and-spoke agreement and suggests areas for change.

I. A Taxonomy of Algorithmic Pricing on Gig Platforms

4Among legal scholars, the algorithm has been made an object of broad concern, raising concerns for decision-making “black boxes”; discrimination; privacy abuses; and anti-competitive outcomes, among others.⁶ In this section we only focus on algorithms with one specific function, namely to determine the price of services offered on gig platforms. At its core, the concept of algorithmic pricing raises a question of the speed of information processing and the breadth of the information that can be included in a pricing decision – as one commercial provider of pricing algorithms advertises, “if a manager can deal with only 3 or 4 factors while crafting a single price, the advanced pricing software… processes 60 factors at once.”⁷

5As an intermediary enabling transactions between various user sides, a gig platform has the power to influence transaction prices, and thereby also the distribution of surplus between various user groups.⁸ Together with the platforms’ search and matching algorithms, the pricing algorithms play an important role in generating the efficiencies that platforms provide. There are three different ways in which prices can be set on a platform: they can be set by the service providers themselves (e.g. Airbnb, Upwork), the platform can assist the service provider in setting the price (e.g. Airbnb)⁹ or the platform itself can set the transaction price (e.g. Uber).¹⁰

6Platforms where service providers set their own prices are least likely to present competition law concerns. As long as individual economic operators determine their market conduct independently, no anti-competitive coordination issues arise.¹¹ Competition issues could potentially arise as soon as the platform steps in and coordinates prices, either by suggesting prices or by fixing prices centrally. We discuss potential competition problems from platforms getting involved in price setting from an US antitrust and EU competition law perspective below.

7Broadly speaking, gig platforms can choose between using two categories of algorithms for suggesting or setting transaction prices: simple if-then algorithms or more complex algorithms that allow for dynamic pricing. Simple computerized algorithms are a programmed set of “if-then” commands that define a path to producing outcomes based on an observed set of events. Also referred to as deterministic algorithms, these are the original and most basic technological intervention in computerized decisions on contract terms and other business decisions.¹² In the context of a ride sharing platform, for example, a simple if-then algorithm might set the tariff per km, depending on the time of the day and weekday. It will always yield the same price for a certain route, at a certain weekday, at a certain time. Cachon et al. (2017) modeled various scenarios that show that simple algorithms setting prices can be detrimental to both the welfare of the service providers and consumers, in addition to not maximizing the platform’s revenue.¹³

8More complex algorithms, such as machine learning (ML) algorithms, incorporate results from earlier runs and continual updating of decision factors both to ensure currency of decisions and to improve their quality.¹⁴ In the context of gig platforms, ML algorithms deployed by the platform can learn from patterns in the data of past transaction in order to predict prices for future transactions and set prices to make transactions more efficient.¹⁵ Furthermore, ML algorithms can be used to run experiments with a variety of prices in order to improve pricing.¹⁶ One example of this is Uber’s surge pricing. There is no consensus on whether pricing based on more complex pricing algorithms is beneficial for the welfare of consumers and service providers, as its effects appear to be ambiguous. Nonetheless, under a not unlikely assumption of high opportunity cost providers, dynamic pricing is found to benefit both providers and consumers.¹⁷

II. Hub-and-Spoke Arrangements under EU Competition Rules

9Article 101 (1) TFEU prohibits “all agreements between undertakings, decisions by associations of undertakings and concerted practices which may affect trade between Member States and which have as their object or effect the prevention, restriction or distortion of competition within the internal market”. The provision distinguishes between by-object and by-effect restrictions of competition. The category of by-object restrictions of competition is reserved to those anti-competitive agreements and concerted practices that “by their very nature have the potential of restricting competition.”¹⁸ They are presumed to have such high likelihood of producing negative effects on competition that it is not necessary to show actual anti-competitive effects to establish a violation. If an agreement or concerted practice is not considered to be anti-competitive by object, it might still violate Article 101 (1) TFEU if it produces anti-competitive effects. Hub-and-spoke arrangements can entail different factual patterns that attract EU competition law scrutiny under Article 101 TFEU. They can amount to outright cartels, they can constitute a form of non-cartel horizontal coordination, or they can entail vertical restraints.¹⁹

A. Horizontal Coordination

10Horizontal coordination in the form of a cartel is usually formed and coordinated between competitors, i.e. undertakings at the same level of the supply chain. The recent ECN+ Directive 2019/1 defines a cartel as “an agreement or concerted practice between two or more competitors aimed at coordinating their competitive behaviour on the market or influencing the relevant parameters of competition through practices such as, but not limited to, the fixing or coordination of purchase or selling prices or other trading conditions [...]”. ²⁰

11Under EU competition law, hub-and-spoke arrangements facilitating price fixing between competitors are as illegal as purely horizontal cartels. In AC Treuhand²¹ for example, a consultancy had facilitated meetings and had enabled indirect exchanges of sensitive business information between various of its clients leading to price-fixing, market-sharing, and other coordinated conduct on various markets. The European Court of Justice held that not only the spokes had to face liability under EU competition law, but that the hub that had facilitated the unlawful conduct was equally considered part of the agreement and thus liable²².

12In order to find a by-object violation by a hub-and-spoke arrangement under Article 101 (1) TFEU, however, it must be shown that the spokes have expressed their concurrent intention to adopt a certain conduct on the market.²³ In a case like AC Treuhand, where the spokes had also communicated between themselves regarding prices, market sharing, and customer allocation, the concurrent intention was easy to establish. In the context of an online platform, however, concurrent intention is more difficult to prove. In the ETURAS case,²⁴ this question arose in the context of an online travel booking platform that messaged travel agencies selling via the platform about a new policy on capping maximum retail discounts. The platform also implemented this policy in through technological restrictions. The question was whether mere receipt of the message was sufficient to establish a concurrent intention to engage in anti-competitive conduct by the travel agencies. The Court held that it was necessary to show that the agencies were aware of the content of the message, that they had acted accordingly on the market, and that there was a causal link between the message and subsequent conduct on the market to establish a violation of Article 101 (1) TFEU.²⁵ The mere fact that an agency continued to use the platform where the cap on the maximum retail discount policy was implemented through an algorithm was insufficient to establish the intent to participate in a concerted practice contrary to Article 101 (1) TFEU.²⁶

13Applying the standards established in the case law on Article 101 (1) TFEU to gig platforms, it is thus unlikely that a gig platform that has algorithms to calculate the prices charged for services on its platform would be considered a cartel without further evidence of the existence of a concerted practice among the service providers using the platform. This would be due to the service providers’ lack of awareness and acceptance of the workings of the platforms’ algorithms as a mechanism that coordinates their conduct with other self-employed on the platform.²⁷

14Apart from the evidentiary obstacle to establish a concurrent intention to engage in a concerted practice by the service providers of a gig platform, the coordination of transaction parameters by the platform is likely to have pro-competitive effects.²⁸ These include the reduction of search costs and the easier and faster conclusion of transactions between various sides of the platform.²⁹ In more recent case law by the Court of Justice, there has been an increased emphasis that agreements or concerted practices that can have pro-competitive effects should not be considered by-object restrictions.³⁰ The requirement to construe the by-object category narrowly together with the efficiencies brought by platforms militates against classifying the hub-and-spoke arrangements implemented algorithmically by gig platforms as cartels, absent other evidence of a concerted practice. This would also facilitate the application of Article 101 (3) TFEU that allows an exception from the application of Article 101 (1) TFEU on efficiency grounds. The criteria for applying Article 101 (3) are four-fold: the agreement in question must (i) improve the production or distribution of goods or promote economic or technical progress, (ii) allow consumers a fair share of the resulting benefit, (iii) be indispensable, and (iv) avoid eliminating all competition in respect of a substantial part of the products in question.

15A case from Luxembourg can serve as an illustration.³¹ It concerned a B2B platform, Webtaxi, that allowed businesses to book taxis via telephone, email, or app. Webtaxi determined the tariffs for various types of taxi services (standard, executive, etc.) and had an algorithm deciding on the allocation of various rides to the participating taxi services. The Luxembourgish Competition Council came to the conclusion that this arrangement was a horizontal price-fixing agreement implemented by Webtaxi and thus a restriction of competition by object according to Article 3 of the Luxembourgish Competition Law.³² Nevertheless, the Competition Council found that an individual exemption applied. It found that the arrangement implemented by Webtaxi reduced the number of empty rides, which resulted in a better allocation of resources and reduced the pollution from the circulation of taxis. In addition, it offered Webtaxi’s clients a uniform, centralized, and 24/7 available taxi service, thus granting the broadest possible access to taxi services. It also allowed an efficient management of peak taxi demand times and moments of low demand. Furthermore, the Council found that the price calculated by the algorithm rendered a tariff that was in the lower spectrum of prices offered in the market for similar services. Interestingly, the Council found that there was no less restrictive alternative to the fixed tariff that would have yielded the same benefits. Negotiating the prices individually for each ride would have likely increased the waiting times of the client, and it would be difficult for the platform to offer individualized transactions with the same speed and efficiency. The price-fixing algorithm was thus key to the efficiencies yielded by the business model of the platform. Lastly, Webtaxi only held 26 % of the relevant market, and there was thus no risk of a complete elimination of competition.

16One may wonder whether the classification of the Webtaxi platform’s pricing algorithm coordinating the taxi rides as a by-object restriction of competition was correct in the first place, given that the efficiencies yielded by it were immediately identified to justify the exemption. If the by-object category was really to be construed narrowly, the arrangement should probably have been analyzed as a by-effects restriction. Nevertheless, the case gives an indication that the use of a pricing algorithm by a platform is unlikely to be found contrary to EU competition law, at least when it comes to the horizontal dimensions of a hub-and-spoke arrangement implemented through a gig platform.³³

B. Vertical Restraints

17If we assume that the horizontal component of a gig platform as hub-and-spoke arrangement is unlikely to run afoul of Article 101 (1) TFEU, its vertical dimension might still attract scrutiny. In contrast to US antitrust law that has adopted a laxer attitude towards vertical restraints,³⁴ EU competition law still treats certain vertical restraints with suspicion. Apart from the general rule of Article 101 (1) TFEU, the Vertical Block Exemption Regulation (VBER)³⁵ further specifies which types of vertical restraints are likely to be found contrary to Article 101 (1) TFEU, and which types of vertical restraints fulfill the criteria of Article 101 (3) TFEU and thus do not violate competition rules. The VBER has a market share threshold,³⁶ and only applies to vertical agreements in which neither party holds more than 30 %.

18Resale price maintenance (RPM), i.e. the fixing of a (minimum) resale price by a buyer, is one form of vertical restraint that constitutes a by-object restriction of competition. Under the current version of the VBER, Article 4(a) lists RPM as a blacklisted practice that removes the benefits of being exempted from the application of Article 101 (1) TFEU. In principle, if gig platforms are considered a collection of vertical agreements between the platform and its users, the use of pricing algorithms by the platform could constitute a form of RPM.³⁷ Indeed, the EU Commission has recently made clear in the draft for its new Vertical Guidelines³⁸ that Article 4(a) VBER is also fully applicable to online intermediation services. The provider of an online intermediation service would be considered a supplier according to Article 1(1)(d) of the Proposal for a new VBER, and would thus be prohibited from imposing a fixed or minimum sales price for the transaction it facilitates.³⁹ The Commission still stressed that efficiency-enhancing RPM could benefit from an individual exemption under Article 101 (3) TFEU.⁴⁰

19Consequently, it seems that the current situation under EU Competition law would potentially lead to classifying the hub-and-spoke arrangements of gig platforms using pricing algorithms setting the prices for transactions as restrictions of competition that could benefit from an individual exemption under Article 101 (3) TFEU, but would otherwise run afoul Article 101 (1) TFEU. This situation would appear to create considerable legal uncertainty for any gig platform setting the prices for individual transactions through the use of algorithms.

III. Hub-and-Spoke under the Sherman Act

20Section 1 of the Sherman Act renders illegal every “contract, combination…, or conspiracy, in restraint of trade…”⁴¹ Courts have clarified that the vague language prohibits (1) agreements, which (2) unreasonably restrain trade.⁴² The hub-and-spoke agreements at issue in gig economy enterprises raise the question whether the series of parallel agreements between service providers and the platform aggregate to an agreement among the service providers.⁴³

A. Vertical Price Fix/RPM

21One question is whether the individual agreements between service providers and a gig platform implicate competition law, and whether the incidence of algorithmic pricing influences this. These agreements involve the platform’s providing terms of service for the service providers to accept, including an agreement that the algorithmically determined price will govern in transactions with consumers. Under the current state of U.S. law, such a vertical price fix is illegal only if it leads to harm to consumers, reflected in price increases or quality reduction.⁴⁴ Those outcomes arise in two primary circumstances. First is the case of market power, which may enable unilateral price increases and can encourage use of RPM to facilitate cartelization in an oligopoly market.⁴⁵ Second is the case of a dealer cartel, organized and implemented through an RPM scheme.⁴⁶ In fact, rather than harm in affected markets, existing evidence suggests quality increases, and price reductions, across gig-economy marketplaces.⁴⁷ Those outcomes would not support liability under an effects-based analysis. Intuition is consistent with that evidence. Market power leading to anticompetitive outcomes should be rare in a market with gig platforms because the best-known gig platforms have entered existing industries, rather than establishing new ones.⁴⁸

22Pricing by algorithm does not change these outcomes. One unique feature of US law, the Colgate doctrine, may operate to make a vertical price term established by the algorithm immune from competition oversight. In United States v. Colgate, an imposed – as opposed to negotiated – price term was excluded entirely from the statutory requirement of an agreement.⁴⁹ In the case of a centrally imposed price term, established in real time by the pricing algorithm, service providers are presented with take-it-or-leave-it offers to contract each time they turn on their app. By providing services under the imposed terms, service providers do not agree to those terms – they merely acquiesce in them, immunizing the vertical relationship from illegality under Section 1.

B. Horizontal Agreements around the “Rim”

23The question under a hub-and-spoke analysis is whether there is a horizontal agreement among the service providers on a gig economy platform, perhaps facilitated by the intervention of algorithmic pricing in the enterprise structure. The US Supreme Court in Interstate Circuit held that an agreement could be reached through a series of parallel communications between one firm – the hub – and multiple competitors at a different level of the distribution chain – the spokes. In the specific case, the hub was a film exhibitor with market power and the spokes, eight otherwise competing movie distributors, each agreed to abide by a price minimum for movies that they distributed.⁵⁰ Keys to the modern understanding of the case are (1) mutual dependence among the distributors in their competitors’ acquiescence in the terms, and (2) the distributors’ being persuaded of their competitors’ acquiescence.⁵¹ Those two elements serve the function of establishing agreement among the spokes, as opposed to mere parallel agreements with the hub.

24The FTC’s prosecution of Toys ‘R’ Us, a mid-market brick-and-mortar toy retailer with substantial market share, turned on that firm’s vertical agreements with major toy manufacturers limiting sales to discount retail outlets.⁵² The court of appeals highlighted the reality of mutual dependence among the manufacturers that each would comply with the terms of the vertical sales limitations. In United States v. Apple Inc., a court of appeals affirmed a finding of a horizontal conspiracy among book publishers, orchestrated by Apple, fixing the price of e-books distributed through e-readers.⁵³ Apple reached individual vertical agreements with five leading publishing houses to establish minimum resale prices that were higher than that being charged by Amazon for Kindle editions. Apple’s vertical agreements came at a time of general awareness among the publishing house competitors that e-book discounting was a threat to their business model, but in the absence of collective action it could be expected to continue. The court of appeals rejected Apple’s arguments that these were parallel but independent vertical agreements [...] because, in context, they provide strong evidence that Apple consciously orchestrated a conspiracy [...]”⁵⁴.

25Gig economy enterprises share similarities with these foundational authorities, but are also fundamentally different in important ways. The first requirement from Interstate Circuit – mutual dependence – is present, as suppliers on gig platforms that control pricing will rely on their competitors’ prices also being controlled. The second element – persuasion of mutual acquiescence – is less clear. In the one US case to consider claims of hub-and-spoke conspiracy among gig service providers, Meyer v. Kalanick, the allegations of conspiracy were sufficient to survive a motion to dismiss, though the case was ultimately dismissed on other grounds and the facts of a hub-and-spoke conspiracy were never developed.⁵⁵ Had those facts proceeded to be developed, it is possible that the presence of a pricing algorithm, which may suggest to service providers in-built parity of treatment akin to a personal assurance from the hub, would have met the second Interstate Circuit requirement that service providers be persuaded of competitor acquiescence. As we conclude below, however, this is more likely with a simpler if-then algorithm than it is with a complex multivariate algorithm.

26Thus, and despite the theoretical appeal of the hub-and-spoke argument for gig platforms, evidence does not appear to support a finding that platforms, whether Uber, Airbnb, or others, have engaged in vertical negotiations with service providers in a way that reflects persuasion of competitors’ acquiescence, as distinct from merely permitting those service providers to assume mutual acquiescence.

Conclusion and Recommendations

27The competitive concern presented by algorithmic pricing on a gig economy platform turns to a great extent on (1) the nature of the platform’s pricing role and (2) the nature of the algorithm. At one extreme, where the platform does not intervene in suppliers’ pricing decisions (with Airbnb as the typical example), concerns for coordination among suppliers and consequent harm to competition or consumers are vanishingly small.⁵⁶ Under this scenario use of algorithms in pricing, if it occurs, is a unilateral choice by an individual supplier which can be expected to produce a price effect only under highly stylized facts. In contrast, where the platform offers to assist in pricing (e.g., the “Smart Pricing“ service on Airbnb), or the platform itself controls="true" pricing in the normal circumstance (e.g., Uber’s dynamic pricing), the concerns for anticompetitive coordination raised by hub-and-spoke price agreements become important.⁵⁷ The greatest risk arises if pricing is centrally controlled through the hub, with moderate risk of anticompetitive coordination if pricing is assisted, with that assistance taken or not at the choice of the supplier.

28Within these use models, concerns for anticompetitive coordination among the spokes differ depending on the nature of the pricing algorithm deployed. A surprising result is that the simpler, if-then algorithm is more likely to lead to a conclusion of anticompetitive coordination, while a more complex algorithm, reaching instant decisions based on dozens of variables and involving recursive learning, is less likely to do so, under the current state of the law in both the US and the EU. The simple algorithm gives suppliers on the platform substantial insight into the pricing decision that is either imposed by agreement or that they opt in to in an assisted pricing scenario. It also limits incentives for supplier competition on other facets of their service, such as the time or location of their choosing to work. Suppliers can also be expected to depend on consistency among competitors for a particular service in a particular geography. It could be plausible for a court or competition enforcer to infer an understanding among those suppliers that instead of competing on price, they provide services pursuant to an agreement not to compete on price.

29In contrast, a more complex algorithm allowing for dynamic pricing by its nature processes more variables than are reasonably considered by an individual supplier, including taking into account the successes or failures of prior decisions by all platform participants. Variables include facets of service competition including time and location, and it is not difficult to imagine other factors influencing demand including complementary services (such as entertainment venues or flight schedules), weather, and public health circumstances, among others. Such an algorithm both provides ample opportunity for competition among suppliers by responding to signals of higher demand, and masks prices charged for any one instance of service being provided, reducing the possibility of inferring a dependency on price consistency among suppliers. In this sense, the “black box” concern sometimes identified with algorithmic decision-making operates to mitigate concerns for competitive harm.⁵⁸ It would also make inferring an understanding among suppliers more tenuous.

30A complex algorithm is less likely to produce competitive harm for reasons of efficiency as well. As the Webtaxi decision demonstrates, possible competitive harms from hub-and-spoke agreements can be mitigated by the efficiencies of resource allocation a well-designed dynamic pricing algorithm permits. This determination is consistent with observations of increased availability and reduced price in ride-hailing markets in the presence of Uber.⁵⁹ Those efficiencies are unlikely to arise in the case of a simple price-setting algorithm.

31Both the US and the EU regimes would provide a flexible approach that takes into account the efficiencies of using pricing algorithms by platforms. Under US antitrust law, a rule of reason approach would be adopted that entails a balancing of pro-and anti-competitive effects, with ultimate liability premised on the actual or likely effects. Both because of this expected result, and because of the inherent challenges of prosecuting rule of reason cases, challenging a hub-and-spoke platform agreement involving a complex algorithm under a rule of reason framework is unlikely to produce a finding of liability.⁶⁰ That hands-off approach provides the benefit of allowing space for innovation. However, given the possibility of developing evidence of efficiency outweighing competitive harm, and the real possibility of harm where efficiencies do not arise and the Interstate Circuit elements are present, the US approach may benefit from increased space for scrutiny.

32Under EU competition law, we would likely observe an analysis of whether an individual exemption under Article 101 (3) TFEU applies in each case, which entails a calculation of efficiencies, an assessment of consumer benefit, a proportionality analysis, and that not all competition on the relevant market in question is eliminated. At the same time, however, the fact that gig platforms would need to invoke an individual exemption might expose them more easily to liability in the EU than in the US. First, platforms would have to carry the burden of proof to show that they meet the requirements of Article 101 (3) TFEU.⁶¹ Second, since the entry into force of Regulation 1/2003, the EU Commission has not once accepted that the requirements under Article 101 (3) TFEU were met by a defendant.⁶² Third, national competition authorities apply the criteria under Article 101 (3) TFEU differently.⁶³ Fourth, having to rely on an individual exemption offers much less legal certainty than being able to assess one’s agreements and cooperation under a Block Exemption Regulation. With increasing empirical evidence of the competitive effects of pricing algorithms, depending on their design, it might be possible to envisage a Block Exemption Regulation for certain algorithmic restrictions implemented by online intermediation platforms.

33At the present moment, however, it appears that albeit sharing similar approaches, EU and US antitrust law might be defaulting into two different positions on the use of algorithms fixing prices on platforms: permitted unless anti-competitive effects can be proven (US) versus prohibited unless pro-competitive effects can be shown (EU). This aligns with US and EU antitrust traditions more generally, where antitrust intervention in the US has been more cautious than in the EU for a variety of reasons.⁶⁴ Due to the lack of sufficient case law, the full picture of the treatment of pricing algorithms under EU and US antitrust law still needs to develop.