The Apple. vs. Epic trial finished up on May 24th and we await the judgement with anticipation. Epic’s complaint focuses on the restrictions that Apple applies on developers that sell via the iOS App store. It therefore focuses on Apple’s role as a gatekeeper of an eco-system that features numerous potential after-markets. The anticipation is only strengthened by the appointment of three prominent tech critics to leading roles at the FTC, DoJ and White House, and the publication of 5 bipartisan bills to strengthen antitrust legislation specifically to address anticompetitive behavior by leading tech firms.
However, while the tectonic plates are clearly shifting and the longer-term prospects for reining in anticompetitive behavior by big tech in the US and beyond are looking increasingly healthy, in the short-term the focus remains on whether more public and private enforcement of the existing rules can be effective. In that respect the Apple. vs. Epic trial marks an important first round in a series of key cases, with others involving Google, Amazon, and Facebook all to follow as they work their way through the system.
Given this context, we therefore take a look at the lead expert reports that Epic and Apple filed and consider their strengths and weaknesses.
Competition in Fore-markets and After-markets
Epic’s expert, Prof. David Evans, correctly identified a “fore-market / after-market” framework for market definition, defining a fore-market for Smartphone OSs and an after-market for iOS app distribution. Applying the smallest market principle, Evans ran a hypothetical monopolist test to consider:
§ whether a SSNIP on Apple app store would prompt consumers to switch to other app stores, and (since consumers typically have just one smartphone),2
§ whether a SSNIP in the after-market would prompt consumers to switch in the fore-market (from iOS to Android).3
Some commentators focus on the existence of fore-market competition, or competition between eco-systems such as iOS and Android, as a sign that competition in digital markets is healthy. However, as Epic have demonstrated, the fact is that purchasers in after-markets such as digital stores (and others such as tap-and-go payments) do not and will not switch their entire eco-system in response to small changes in the offer that is made in the after-market (e.g. a SSNIP on the commission on app sales). This is in part because the cost of app purchases in the after-market is such a small part of the lifecycle cost of a smartphone. It is also because consumer behavioural biases4 mean that many users do not react to after-market prices at the point of sale in the fore-market even if they know they will have to pay them in future.
This of course does not mean there is not competition between eco-systems, simply that such competition does not protect consumers and developers from a hypothetical monopolist applying a SSNIP in after-markets like the iOS digital store market.
Some commentary on the case has focused on the fact that Apple has never hidden the fact that it ties the app store to iOS, and so consumers are forewarned as to the implications of their decision to lock themselves into the iOS eco-system. However, the fact that consumers are forewarned does not imply that they react (or are likely to in future react) to the information that they are forewarned about. To make this jump is to assume a view on what would be rational behavior by consumers, rather than letting the facts of the case guide the assessment of how they would react to a SSNIP. Basing market definition assessment on assumptions that are in fact simply theoretical possibilities, rather than the evidence on how consumers are likely to in practice behave within the market, is both un-economic and can be expected to lead to errors.
While Evans identifies smartphone OSs as the fore-market, he also argues they are two-sided, with developers and consumers participating on opposite sides of a common platform. Since he also identifies these same two groups as participating on either side of the two-sided digital store market, Schmalensee, Apple’s expert, is able to use this to argue that the distinction between the two-sided digital store market and the two-sided smartphone OS market is artificial and misconceived.
In fact, both are wrong here. Contra to Evans, the two-sided smartphone OS market features consumers on one-side and OEMs (e.g. Samsung), not developers, on the other. OEMs decide which OS to put on their devices, buying a license for the one they choose and then selling the bundle to consumers. Of course, Apple decides to exclusively tie iOS to its iPhones, meaning consumers are unable to acquire other phones with iOS, or alternatively iPhones with Android (or other OS). The impact of that tie is however not the focus of this complaint.
Instead, here the focus is on the fact that developers have two markets to operate in. Firstly, they need to decide which OS they want to develop an app for, and hence they need to purchase the license and app-building tools to do so from the relevant OS (a one-sided market).5 Second, they need to decide which two-sided app store to sell their iOS app in (that choice is restricted by Apple’s tying of the two). Therefore, Schmalensee is also wrong to suggest that this one-sided input purchasing market is the same as the app distribution market. Instead, as Evans points out, in MacOS the developer buys the rights and tools to build an iOS app, and then decides where to sell that app.6 The fact that Apple ties the distribution decision to the input purchasing decision does not mean they are one decision.
Schmalensee goes on to argue that Evans ignores indirect network effects when applying the two-sided SSNIP test. This has some merit, since Evans does not appear to explore whether a SSNIP on the Apple app store’s commission would lead, not only to developers switching away from iOS, but also whether any such loss of developers (or higher prices passed on as a result) would lead to a loss of consumers.
However, correcting this oversight is relatively straightforward, and is unlikely to result in a different relevant market. Firstly, Evans has identified that developers have not and are unlikely to divert in response to a SSNIP, therefore there is no reduction in app choice for consumers to respond to. Secondly, as to whether consumers would divert if the SSNIP were passed on, Evans has already provided powerful evidence that consumers are not going to switch smartphone OS in response to a SSNIP on the price of iOS app distribution.
Is iOS In-App Payment a Separate Market from the App Store?
Evans also defines a relevant market for iOS in-app payment processing solutions. However, Schmalensee argues that In-App Payment (IAP) is integral to the Apple App store and that like other digital stores, it enables efficient collection of Apple’s commissions. However, the absence of IAP from iOS apps available on the App store that offer physical rather than digital products suggests otherwise. Indeed this, taken alongside Apple’s acknowledgment that IAP revenue cross-subsidizes the provision of free apps on the store (Steve Jobs, 2008), suggests that the App store and IAP are in fact distinct products – products which it is sometimes in the interest of the store to combine, and sometimes not.
Evans therefore considers whether a hypothetical monopolist could profitably apply a SSNIP to In-App Payments. He concludes that it could. However, this allows Schmalensee to argue that firms such as Square and Braintree would compete in this market. In fact, the question that Evans should have asked himself is whether a hypothetical monopolist could profitably apply a SSNIP to In-App Payments for digital services. After all, Apple as the not-so-hypothetical monopolist in question clearly could, and in fact has, profitably introduced a SSNIP on only those firms that it identified as providing digital as opposed to physical services. Therefore, this is a clear instance of a price discrimination market, in which applying the smallest market principle means taking not only the smallest set of products, but also the smallest set of buyers over which a hypothetical monopolist could profitably impose a SSNIP.7
Theories of Harm
The Theory of Harm on Apple’s Restrictions
Having established the relevant market, Evans describes the allegedly anticompetitive restrictions that Apple uses to protect its monopoly of the iOS distribution market. He describes them as follows:
“Apple required that, as a condition of getting access to the tools and permissions for writing iOS apps, developers use the App Store for “exclusive distribution” and “agree not to distribute [their app] to third parties or to enable or permit others to do so.” It also barred developers from offering any app that would provide a distribution channel for users and developers and thereby prevented any other app store from being made available to iOS users.” Para 100.
Evans therefore does not set out his argument as a refusal to deal, or as anti-competitive tying.8 Indeed, he declined to put a label on his argument. Perhaps this is unsurprising given that Evans has previously suggested that “Tying is Antitrust’s Greatest Intellectual Embarrassment”.9 However, it does create a risk that Epic’s case will be interpreted as a refusal to deal. Indeed, some commentators have done so and cited the precedent on refusal to deal10 as reason to believe that Epic’s case will fail.11 If the court agrees and characterizes the restrictions as a refusal to deal, this might constitute a missed opportunity. In any case, this illustrates the risks of plaintiffs hemming themselves in through their choice of conservative economists.
A distinct theory of harm that featured heavily in the trial, but on which Evans spends little time, is related to the anti-steering restrictions in Apple’s developer agreements. For example, these restrictions steer transactions towards the Apple app store (and the use of IAP) by preventing developers from communicating with customers about options for making purchases outside the app.
Tim Cook says this “would be akin to Apple down at Best Buy saying ‘Best Buy, put in a sign there where we are advertising that you can go across the street and get an iPhone’” However as some commentators have pointed out, Apple’s ban is more comprehensive and extends to preventing general communications to users from discouraging the use of Apple’s IAP, therefore this is actually like Best Buy not allowing products in the store to have a website listed in the instruction manual that happens to sell the same products.12
Schmalensee argues this is comparable to the anti-steering rules that Visa and Mastercard have committed not to engage in, but which American Express is allowed to apply after the widely criticized US Supreme Court judgement (“AmEx”).13 Evans, a rare defender of the AmEx judgement, rejects this comparison. However, the same free-riding arguments which failed to hold water, but were nevertheless accepted in AmEx can certainly be posed in this case. For example, would allowing developers to steer consumers to cheaper purchases via their websites, not risk undermine the viability of the platform and its investment in innovating? In AmEx, this logic collapsed because rival payment solutions were not in fact free-riding on the fact that a merchant accepted AmEx. In this case the argument is marginally stronger, since it is true that the purchases made via the Epic store might be used when playing the game on iOS and so the free-riding could be genuine.
Nevertheless, preventing the use of a developer’s own website to discourage the use of Apple’s IAP, rather than simply preventing such discouragement occurring within the app itself, goes significantly beyond the conduct examined in the American Express decision. Furthermore, the rationale for this invasive restriction appears weak. For example, it is far from clear that an iOS app store would not be viable but-for the restriction on competition. Indeed, the only barrier to the viability of numerous additional stores are the very restrictions that Apple has imposed to prevent them opening. Similarly, whether the incentive to invest in the store that is created by the absence of free-riding is larger than the incentive to invest in the store that is lost through the lack of competitive constraints on the store is doubtful.
Apple suggested that these steering restrictions themselves demonstrate that the market extends beyond iOS, however, as discussed above, the hypothetical monopolist test does not simply list every conceivable constraint, it asks which ones are important enough to prevent a profitable SSNIP. The fact that Apple’s exclusionary behavior has denied the court the opportunity to observe with certainty the strength of these constraints should not be held in its favor. As Areeda & Hovenkamp note in discussing the Microsoft ruling: “To some degree, the defendant is made to suffer the uncertain consequences of its own undesirable conduct.”14.
Retail MFN Clauses
A further theory of harm that does not feature in Evans report is that of across platform parity agreements (APPAs). These often take the form of retail MFN or price parity agreements such as those that Amazon and Booking.com have been challenged for using in their seller contracts (but which it is alleged that Amazon have retained under the title of a Fair Pricing Policy).
In this case the relevant restriction is one which prevents developers from offering content on non-iOS platforms that they do not make available for in-app purchase through the Apple App store.15 This appears to fall into a category of conduct known as “contracts that reference rivals” or CRRs. These CRRs have raised increasing concern since coming to prominence within the retail MFN literature.16 Indeed the clause itself directly restricts output by preventing content being sold on rival non-iOS platforms and so also provides a helpful direct indication of Apple’s market power.
But-for the Restrictions
Evans argues that in the counterfactual (but-for) world, there would be healthy competition between app stores. He argues that entry would have occurred absent the restrictions, and points to a) the 60+ Android apps stores in China where there are no restrictions, b) the existence of competing app stores on the early smartphone OS (Symbian, Blackberry, Windows Mobile, and Palm), c) that in recent years 6 major rivals have sought to set up rival app stores, and d) that there are competing app stores on personal computing and macOS, and e) that consumers say they would use rival app stores if they existed.
This is therefore a powerful case, on which Schmalensee does not comment. Apple argued that the presence of rival app stores introduced safety risks. Craig Federighi said the level of malware on the Mac was “unacceptable” and that iOS would get “run over” by malware attacks if it adopted the same model. In effect this accepts that the counterfactual would be the entry of competing app stores. In a competitive app store market, we would expect that those new app stores (Epic, Amazon, Google, Microsoft, Facebook and Nvidia) would compete not only on price, but on the strength of their safety records. This would allow consumers themselves to choose the best combination of price and safety for them, and Apple would be free to inform them of the risks posed by using app stores with poor safety records.
While there are some peculiarities in Epic’s expert report relating to market definition and the theory of harm, the case nevertheless appears a strong one. Moreover, by setting out a clear framework on after-market cases it would open the door to numerous other actions. In contrast if it falls there will no doubt be renewed demands, from those signatories of the Utah Statement (Khan, Steinbaum & Wu, 2019) that are newly empowered in their new roles at the FTC and the White House, for legislative reform to facilitate action against the application of these restrictions by Apple and others.
1 Epic Games, Inc. v. Apple Inc. (4:20-cv-05640), District Court, N.D. California
2 A SSNIP on that store price might be unprofitable if it results in a significant diversion of sales to other retailers that are distributing apps or games on the same platform (e.g. other sellers of steam or xbox codes). In that case the relevant market needs to expand to take in those other retailers. If there is also significant switching to purchasing the app or game via a different platform (e.g. from sales of Epic games on an xbox to sales of Epic games on a smartphone iOS), then distribution via stores on other platforms will also need to be added to the relevant market. However, switching from purchasing apps on the apple app store to the Google Playstore would require the purchasing of a second smartphone.
3 In contrast, Apple’s expert, Prof. Richard Schmalensee argues that the relevant market is for digital game transactions.
4 For example, hyperbolic discounting, this refers to changes in preferences over time such that immediate rewards are seen as disproportionately more attractive. See CCP, 2013.
5 In 2007, Apple announced that it would enable developers to write apps and would provide them with a software development kit (“SDK”). In 2008, Apple released an iOS SDK and APIs for core services to developers, making the relevant tools and permissions available to developers for nominal cost.
6 Similarly, an ecommerce seller decides where to buy fulfilment/delivery services and then decides whether to sell on Amazon or not. They might decide to buy fulfilment from amazon and sell on amazon, or just one of the two.
7 See the US Horizontal Merger Guidelines for a description of a price discrimination market.
8 For example, consumers can buy an iOS smartphone but only on condition that they use the Apple app store, effectively tying the Apple app store to their purchase of an iOS smartphone. Nor does Evans point to the fact that Apple agree to sell inputs to developers (app building tools and licenses), but only on condition that the developer does not use those inputs to compete with Apple’s app store to distribute iOS apps. As Eric Hovenkamp (2021), points out, such tie-like conditionalities on sales improve the prospects of cases that might also involve a refusal to deal. In particular, he notes that Kodak refused to supply repair parts to rival repairers, but it also agreed to sell parts on the condition that the buyer did not use those tools with any repair service from a rival of Kodak. (The Antitrust Duty to Deal in the Age of Big Tech).
10 Aspen Skiing Co. v. Aspen Highlands Skiing Corp., 472 U.S. 585 (1985)
11 Thompson, 2021, Stratechery, Epic v. Apple Trial Begins, EU Files Statement of Objections, Who Owns the App?
12 Thompson, 2021, Stratechery. Or similarly Airbnb might prevent those renting homes from leaving materials informing renters that they can rent the same property through other channels.
13 Ohio v. American Express Co., 585 U.S. ___ (2018). See criticism Hovenkamp (2019)
14 Areeda & Hovenkamp, Antitrust Law p 651c, at 78
15 p59 Evans Report
16 Scott Morton, Antitrust 2013, Contracts that Reference Rivals