Archives For WhatsApp

A raft of progressive scholars in recent years have argued that antitrust law remains blind to the emergence of so-called “attention markets,” in which firms compete by converting user attention into advertising revenue. This blindness, the scholars argue, has caused antitrust enforcers to clear harmful mergers in these industries.

It certainly appears the argument is gaining increased attention, for lack of a better word, with sympathetic policymakers. In a recent call for comments regarding their joint merger guidelines, the U.S. Justice Department (DOJ) and Federal Trade Commission (FTC) ask:

How should the guidelines analyze mergers involving competition for attention? How should relevant markets be defined? What types of harms should the guidelines consider?

Unfortunately, the recent scholarly inquiries into attention markets remain inadequate for policymaking purposes. For example, while many progressives focus specifically on antitrust authorities’ decisions to clear Facebook’s 2012 acquisition of Instagram and 2014 purchase of WhatsApp, they largely tend to ignore the competitive constraints Facebook now faces from TikTok (here and here).

When firms that compete for attention seek to merge, authorities need to infer whether the deal will lead to an “attention monopoly” (if the merging firms are the only, or primary, market competitors for some consumers’ attention) or whether other “attention goods” sufficiently constrain the merged entity. Put another way, the challenge is not just in determining which firms compete for attention, but in evaluating how strongly each constrains the others.

As this piece explains, recent attention-market scholarship fails to offer objective, let alone quantifiable, criteria that might enable authorities to identify firms that are unique competitors for user attention. These limitations should counsel policymakers to proceed with increased rigor when they analyze anticompetitive effects.

The Shaky Foundations of Attention Markets Theory

Advocates for more vigorous antitrust intervention have raised (at least) three normative arguments that pertain attention markets and merger enforcement.

  • First, because they compete for attention, firms may be more competitively related than they seem at first sight. It is sometimes said that these firms are nascent competitors.
  • Second, the scholars argue that all firms competing for attention should not automatically be included in the same relevant market.
  • Finally, scholars argue that enforcers should adopt policy tools to measure market power in these attention markets—e.g., by applying a SSNIC test (“small but significant non-transitory increase in cost”), rather than a SSNIP test (“small but significant non-transitory increase in price”).

There are some contradictions among these three claims. On the one hand, proponents advocate adopting a broad notion of competition for attention, which would ensure that firms are seen as competitively related and thus boost the prospects that antitrust interventions targeting them will be successful. When the shoe is on the other foot, however, proponents fail to follow the logic they have sketched out to its natural conclusion; that is to say, they underplay the competitive constraints that are necessarily imposed by wider-ranging targets for consumer attention. In other words, progressive scholars are keen to ensure the concept is not mobilized to draw broader market definitions than is currently the case:

This “massive market” narrative rests on an obvious fallacy. Proponents argue that the relevant market includes all substitutable sources of attention depletion,” so the market is “enormous.”

Faced with this apparent contradiction, scholars retort that the circle can be squared by deploying new analytical tools that measure attention for competition, such as the so-called SSNIC test. But do these tools actually resolve the contradiction? It would appear, instead, that they merely enable enforcers to selectively mobilize the attention-market concept in ways that fit their preferences. Consider the following description of the SSNIC test, by John Newman:

But if the focus is on the zero-price barter exchange, the SSNIP test requires modification. In such cases, the “SSNIC” (Small but Significant and Non-transitory Increase in Cost) test can replace the SSNIP. Instead of asking whether a hypothetical monopolist would increase prices, the analyst should ask whether the monopolist would likely increase attention costs. The relevant cost increases can take the form of more time or space being devoted to advertisements, or the imposition of more distracting advertisements. Alternatively, one might ask whether the hypothetical monopolist would likely impose an “SSNDQ” (Small but Significant and Non-Transitory Decrease in Quality). The latter framing should generally be avoided, however, for reasons discussed below in the context of anticompetitive effects. Regardless of framing, however, the core question is what would happen if the ratio between desired content to advertising load were to shift.

Tim Wu makes roughly the same argument:

The A-SSNIP would posit a hypothetical monopolist who adds a 5-second advertisement before the mobile map, and leaves it there for a year. If consumers accepted the delay, instead of switching to streaming video or other attentional options, then the market is correctly defined and calculation of market shares would be in order.

The key problem is this: consumer switching among platforms is consistent both with competition and with monopoly power. In fact, consumers are more likely to switch to other goods when they are faced with a monopoly. Perhaps more importantly, consumers can and do switch to a whole range of idiosyncratic goods. Absent some quantifiable metric, it is simply impossible to tell which of these alternatives are significant competitors.

None of this is new, of course. Antitrust scholars have spent decades wrestling with similar issues in connection with the price-related SSNIP test. The upshot of those debates is that the SSNIP test does not measure whether price increases cause users to switch. Instead, it examines whether firms can profitably raise prices above the competitive baseline. Properly understood, this nuance renders proposed SSNIC and SSNDQ tests (“small but significant non-transitory decrease in quality”) unworkable.

First and foremost, proponents wrongly presume to know how firms would choose to exercise their market power, rendering the resulting tests unfit for policymaking purposes. This mistake largely stems from the conflation of price levels and price structures in two-sided markets. In a two-sided market, the price level refers to the cumulative price charged to both sides of a platform. Conversely, the price structure refers to the allocation of prices among users on both sides of a platform (i.e., how much users on each side contribute to the costs of the platform). This is important because, as Jean Charles Rochet and Jean Tirole show in their Nobel-winning work, changes to either the price level or the price structure both affect economic output in two-sided markets.

This has powerful ramifications for antitrust policy in attention markets. To be analytically useful, SSNIC and SSNDQ tests would have to alter the price level while holding the price structure equal. This is the opposite of what attention-market theory advocates are calling for. Indeed, increasing ad loads or decreasing the quality of services provided by a platform, while holding ad prices constant, evidently alters platforms’ chosen price structure.

This matters. Even if the proposed tests were properly implemented (which would be difficult: it is unclear what a 5% quality degradation would look like), the tests would likely lead to false negatives, as they force firms to depart from their chosen (and, thus, presumably profit-maximizing) price structure/price level combinations.

Consider the following illustration: to a first approximation, increasing the quantity of ads served on YouTube would presumably decrease Google’s revenues, as doing so would simultaneously increase output in the ad market (note that the test becomes even more absurd if ad revenues are held constant). In short, scholars fail to recognize that the consumer side of these markets is intrinsically related to the ad side. Each side affects the other in ways that prevent policymakers from using single-sided ad-load increases or quality decreases as an independent variable.

This leads to a second, more fundamental, flaw. To be analytically useful, these increased ad loads and quality deteriorations would have to be applied from the competitive baseline. Unfortunately, it is not obvious what this baseline looks like in two-sided markets.

Economic theory tells us that, in regular markets, goods are sold at marginal cost under perfect competition. However, there is no such shortcut in two-sided markets. As David Evans and Richard Schmalensee aptly summarize:

An increase in marginal cost on one side does not necessarily result in an increase in price on that side relative to price on the other. More generally, the relationship between price and cost is complex, and the simple formulas that have been derived by single-handed markets do not apply.

In other words, while economic theory suggests perfect competition among multi-sided platforms should result in zero economic profits, it does not say what the allocation of prices will look like in this scenario. There is thus no clearly defined competitive baseline upon which to apply increased ad loads or quality degradations. And this makes the SSNIC and SSNDQ tests unsuitable.

In short, the theoretical foundations necessary to apply the equivalent of a SSNIP test on the “free” side of two-sided platforms are largely absent (or exceedingly hard to apply in practice). Calls to implement SSNIC and SSNDQ tests thus greatly overestimate the current state of the art, as well as decision-makers’ ability to solve intractable economic conundrums. The upshot is that, while proposals to apply the SSNIP test to attention markets may have the trappings of economic rigor, the resemblance is superficial. As things stand, these tests fail to ascertain whether given firms are in competition, and in what market.

The Bait and Switch: Qualitative Indicia

These problems with the new quantitative metrics likely explain why proponents of tougher enforcement in attention markets often fall back upon qualitative indicia to resolve market-definition issues. As John Newman writes:

Courts, including the U.S. Supreme Court, have long employed practical indicia as a flexible, workable means of defining relevant markets. This approach considers real-world factors: products’ functional characteristics, the presence or absence of substantial price differences between products, whether companies strategically consider and respond to each other’s competitive conduct, and evidence that industry participants or analysts themselves identify a grouping of activity as a discrete sphere of competition. …The SSNIC test may sometimes be massaged enough to work in attention markets, but practical indicia will often—perhaps usually—be the preferable method

Unfortunately, far from resolving the problems associated with measuring market power in digital markets (and of defining relevant markets in antitrust proceedings), this proposed solution would merely focus investigations on subjective and discretionary factors.

This can be easily understood by looking at the FTC’s Facebook complaint regarding its purchases of WhatsApp and Instagram. The complaint argues that Facebook—a “social networking service,” in the eyes of the FTC—was not interchangeable with either mobile-messaging services or online-video services. To support this conclusion, it cites a series of superficial differences. For instance, the FTC argues that online-video services “are not used primarily to communicate with friends, family, and other personal connections,” while mobile-messaging services “do not feature a shared social space in which users can interact, and do not rely upon a social graph that supports users in making connections and sharing experiences with friends and family.”

This is a poor way to delineate relevant markets. It wrongly portrays competitive constraints as a binary question, rather than a matter of degree. Pointing to the functional differences that exist among rival services mostly fails to resolve this question of degree. It also likely explains why advocates of tougher enforcement have often decried the use of qualitative indicia when the shoe is on the other foot—e.g., when authorities concluded that Facebook did not, in fact, compete with Instagram because their services were functionally different.

A second, and related, problem with the use of qualitative indicia is that they are, almost by definition, arbitrary. Take two services that may or may not be competitors, such as Instagram and TikTok. The two share some similarities, as well as many differences. For instance, while both services enable users to share and engage with video content, they differ significantly in the way this content is displayed. Unfortunately, absent quantitative evidence, it is simply impossible to tell whether, and to what extent, the similarities outweigh the differences. 

There is significant risk that qualitative indicia may lead to arbitrary enforcement, where markets are artificially narrowed by pointing to superficial differences among firms, and where competitive constraints are overemphasized by pointing to consumer switching. 

The Way Forward

The difficulties discussed above should serve as a good reminder that market definition is but a means to an end.

As William Landes, Richard Posner, and Louis Kaplow have all observed (here and here), market definition is merely a proxy for market power, which in turn enables policymakers to infer whether consumer harm (the underlying question to be answered) is likely in a given case.

Given the difficulties inherent in properly defining markets, policymakers should redouble their efforts to precisely measure both potential barriers to entry (the obstacles that may lead to market power) or anticompetitive effects (the potentially undesirable effect of market power), under a case-by-case analysis that looks at both sides of a platform.

Unfortunately, this is not how the FTC has proceeded in recent cases. The FTC’s Facebook complaint, to cite but one example, merely assumes the existence of network effects (a potential barrier to entry) with no effort to quantify their magnitude. Likewise, the agency’s assessment of consumer harm is just two pages long and includes superficial conclusions that appear plucked from thin air:

The benefits to users of additional competition include some or all of the following: additional innovation … ; quality improvements … ; and/or consumer choice … . In addition, by monopolizing the U.S. market for personal social networking, Facebook also harmed, and continues to harm, competition for the sale of advertising in the United States.

Not one of these assertions is based on anything that could remotely be construed as empirical or even anecdotal evidence. Instead, the FTC’s claims are presented as self-evident. Given the difficulties surrounding market definition in digital markets, this superficial analysis of anticompetitive harm is simply untenable.

In short, discussions around attention markets emphasize the important role of case-by-case analysis underpinned by the consumer welfare standard. Indeed, the fact that some of antitrust enforcement’s usual benchmarks are unreliable in digital markets reinforces the conclusion that an empirically grounded analysis of barriers to entry and actual anticompetitive effects must remain the cornerstones of sound antitrust policy. Or, put differently, uncertainty surrounding certain aspects of a case is no excuse for arbitrary speculation. Instead, authorities must meet such uncertainty with an even more vigilant commitment to thoroughness.

On both sides of the Atlantic, 2021 has seen legislative and regulatory proposals to mandate that various digital services be made interoperable with others. Several bills to do so have been proposed in Congress; the EU’s proposed Digital Markets Act would mandate interoperability in certain contexts for “gatekeeper” platforms; and the UK’s competition regulator will be given powers to require interoperability as part of a suite of “pro-competitive interventions” that are hoped to increase competition in digital markets.

The European Commission plans to require Apple to use USB-C charging ports on iPhones to allow interoperability among different chargers (to save, the Commission estimates, two grams of waste per-European per-year). Interoperability demands for forms of interoperability have been at the center of at least two major lawsuits: Epic’s case against Apple and a separate lawsuit against Apple by the app called Coronavirus Reporter. In July, a group of pro-intervention academics published a white paper calling interoperability “the ‘Super Tool’ of Digital Platform Governance.”

What is meant by the term “interoperability” varies widely. It can refer to relatively narrow interventions in which user data from one service is made directly portable to other services, rather than the user having to download and later re-upload it. At the other end of the spectrum, it could mean regulations to require virtually any vertical integration be unwound. (Should a Tesla’s engine be “interoperable” with the chassis of a Land Rover?) And in between are various proposals for specific applications of interoperability—some product working with another made by another company.

Why Isn’t Everything Interoperable?

The world is filled with examples of interoperability that arose through the (often voluntary) adoption of standards. Credit card companies oversee massive interoperable payments networks; screwdrivers are interoperable with screws made by other manufacturers, although different standards exist; many U.S. colleges accept credits earned at other accredited institutions. The containerization revolution in shipping is an example of interoperability leading to enormous efficiency gains, with a government subsidy to encourage the adoption of a single standard.

And interoperability can emerge over time. Microsoft Word used to be maddeningly non-interoperable with other word processors. Once OpenOffice entered the market, Microsoft patched its product to support OpenOffice files; Word documents now work slightly better with products like Google Docs, as well.

But there are also lots of things that could be interoperable but aren’t, like the Tesla motors that can’t easily be removed and added to other vehicles. The charging cases for Apple’s AirPods and Sony’s wireless earbuds could, in principle, be shaped to be interoperable. Medical records could, in principle, be standardized and made interoperable among healthcare providers, and it’s easy to imagine some of the benefits that could come from being able to plug your medical history into apps like MyFitnessPal and Apple Health. Keurig pods could, in principle, be interoperable with Nespresso machines. Your front door keys could, in principle, be made interoperable with my front door lock.

The reason not everything is interoperable like this is because interoperability comes with costs as well as benefits. It may be worth letting different earbuds have different designs because, while it means we sacrifice easy interoperability, we gain the ability for better designs to be brought to market and for consumers to have choice among different kinds. We may find that, while digital health records are wonderful in theory, the compliance costs of a standardized format might outweigh those benefits.

Manufacturers may choose to sell an expensive device with a relatively cheap upfront price tag, relying on consumer “lock in” for a stream of supplies and updates to finance the “full” price over time, provided the consumer likes it enough to keep using it.

Interoperability can remove a layer of security. I don’t want my bank account to be interoperable with any payments app, because it increases the risk of getting scammed. What I like about my front door lock is precisely that it isn’t interoperable with anyone else’s key. Lots of people complain about popular Twitter accounts being obnoxious, rabble-rousing, and stupid; it’s not difficult to imagine the benefits of a new, similar service that wanted everyone to start from the same level and so did not allow users to carry their old Twitter following with them.

There thus may be particular costs that prevent interoperability from being worth the tradeoff, such as that:

  1. It might be too costly to implement and/or maintain.
  2. It might prescribe a certain product design and prevent experimentation and innovation.
  3. It might add too much complexity and/or confusion for users, who may prefer not to have certain choices.
  4. It might increase the risk of something not working, or of security breaches.
  5. It might prevent certain pricing models that increase output.
  6. It might compromise some element of the product or service that benefits specifically from not being interoperable.

In a market that is functioning reasonably well, we should be able to assume that competition and consumer choice will discover the desirable degree of interoperability among different products. If there are benefits to making your product interoperable with others that outweigh the costs of doing so, that should give you an advantage over competitors and allow you to compete them away. If the costs outweigh the benefits, the opposite will happen—consumers will choose products that are not interoperable with each other.

In short, we cannot infer from the absence of interoperability that something is wrong, since we frequently observe that the costs of interoperability outweigh the benefits.

Of course, markets do not always lead to optimal outcomes. In cases where a market is “failing”—e.g., because competition is obstructed, or because there are important externalities that are not accounted for by the market’s prices—certain goods may be under-provided. In the case of interoperability, this can happen if firms struggle to coordinate upon a single standard, or because firms’ incentives to establish a standard are not aligned with the social optimum (i.e., interoperability might be optimal and fail to emerge, or vice versa).

But the analysis cannot stop here: just because a market might not be functioning well and does not currently provide some form of interoperability, we cannot assume that if it was functioning well that it would provide interoperability.

Interoperability for Digital Platforms

Since we know that many clearly functional markets and products do not provide all forms of interoperability that we could imagine them providing, it is perfectly possible that many badly functioning markets and products would still not provide interoperability, even if they did not suffer from whatever has obstructed competition or effective coordination in that market. In these cases, imposing interoperability would destroy value.

It would therefore be a mistake to assume that more interoperability in digital markets would be better, even if you believe that those digital markets suffer from too little competition. Let’s say, for the sake of argument, that Facebook/Meta has market power that allows it to keep its subsidiary WhatsApp from being interoperable with other competing services. Even then, we still would not know if WhatsApp users would want that interoperability, given the trade-offs.

A look at smaller competitors like Telegram and Signal, which we have no reason to believe have market power, demonstrates that they also are not interoperable with other messaging services. Signal is run by a nonprofit, and thus has little incentive to obstruct users for the sake of market power. Why does it not provide interoperability? I don’t know, but I would speculate that the security risks and technical costs of doing so outweigh the expected benefit to Signal’s users. If that is true, it seems strange to assume away the potential costs of making WhatsApp interoperable, especially if those costs may relate to things like security or product design.

Interoperability and Contact-Tracing Apps

A full consideration of the trade-offs is also necessary to evaluate the lawsuit that Coronavirus Reporter filed against Apple. Coronavirus Reporter was a COVID-19 contact-tracing app that Apple rejected from the App Store in March 2020. Its makers are now suing Apple for, they say, stifling competition in the contact-tracing market. Apple’s defense is that it only allowed COVID-19 apps from “recognised entities such as government organisations, health-focused NGOs, companies deeply credentialed in health issues, and medical or educational institutions.” In effect, by barring it from the App Store, and offering no other way to install the app, Apple denied Coronavirus Reporter interoperability with the iPhone. Coronavirus Reporter argues it should be punished for doing so.

No doubt, Apple’s decision did reduce competition among COVID-19 contact tracing apps. But increasing competition among COVID-19 contact-tracing apps via mandatory interoperability might have costs in other parts of the market. It might, for instance, confuse users who would like a very straightforward way to download their country’s official contact-tracing app. Or it might require access to certain data that users might not want to share, preferring to let an intermediary like Apple decide for them. Narrowing choice like this can be valuable, since it means individual users don’t have to research every single possible option every time they buy or use some product. If you don’t believe me, turn off your spam filter for a few days and see how you feel.

In this case, the potential costs of the access that Coronavirus Reporter wants are obvious: while it may have had the best contact-tracing service in the world, sorting it from other less reliable and/or scrupulous apps may have been difficult and the risk to users may have outweighed the benefits. As Apple and Facebook/Meta constantly point out, the security risks involved in making their services more interoperable are not trivial.

It isn’t competition among COVID-19 apps that is important, per se. As ever, competition is a means to an end, and maximizing it in one context—via, say, mandatory interoperability—cannot be judged without knowing the trade-offs that maximization requires. Even if we thought of Apple as a monopolist over iPhone users—ignoring the fact that Apple’s iPhones obviously are substitutable with Android devices to a significant degree—it wouldn’t follow that the more interoperability, the better.

A ‘Super Tool’ for Digital Market Intervention?

The Coronavirus Reporter example may feel like an “easy” case for opponents of mandatory interoperability. Of course we don’t want anything calling itself a COVID-19 app to have totally open access to people’s iPhones! But what’s vexing about mandatory interoperability is that it’s very hard to sort the sensible applications from the silly ones, and most proposals don’t even try. The leading U.S. House proposal for mandatory interoperability, the ACCESS Act, would require that platforms “maintain a set of transparent, third-party-accessible interfaces (including application programming interfaces) to facilitate and maintain interoperability with a competing business or a potential competing business,” based on APIs designed by the Federal Trade Commission.

The only nod to the costs of this requirement are provisions that further require platforms to set “reasonably necessary” security standards, and a provision to allow the removal of third-party apps that don’t “reasonably secure” user data. No other costs of mandatory interoperability are acknowledged at all.

The same goes for the even more substantive proposals for mandatory interoperability. Released in July 2021, “Equitable Interoperability: The ‘Super Tool’ of Digital Platform Governance” is co-authored by some of the most esteemed competition economists in the business. While it details obscure points about matters like how chat groups might work across interoperable chat services, it is virtually silent on any of the costs or trade-offs of its proposals. Indeed, the first “risk” the report identifies is that regulators might be too slow to impose interoperability in certain cases! It reads like interoperability has been asked what its biggest weaknesses are in a job interview.

Where the report does acknowledge trade-offs—for example, interoperability making it harder for a service to monetize its user base, who can just bypass ads on the service by using a third-party app that blocks them—it just says that the overseeing “technical committee or regulator may wish to create conduct rules” to decide.

Ditto with the objection that mandatory interoperability might limit differentiation among competitors – like, for example, how imposing the old micro-USB standard on Apple might have stopped us from getting the Lightning port. Again, they punt: “We recommend that the regulator or the technical committee consult regularly with market participants and allow the regulated interface to evolve in response to market needs.”

But if we could entrust this degree of product design to regulators, weighing the costs of a feature against its benefits, we wouldn’t need markets or competition at all. And the report just assumes away many other obvious costs: “​​the working hypothesis we use in this paper is that the governance issues are more of a challenge than the technical issues.” Despite its illustrious panel of co-authors, the report fails to grapple with the most basic counterargument possible: its proposals have costs as well as benefits, and it’s not straightforward to decide which is bigger than which.

Strangely, the report includes a section that “looks ahead” to “Google’s Dominance Over the Internet of Things.” This, the report says, stems from the company’s “market power in device OS’s [that] allows Google to set licensing conditions that position Google to maintain its monopoly and extract rents from these industries in future.” The report claims this inevitability can only be avoided by imposing interoperability requirements.

The authors completely ignore that a smart home interoperability standard has already been developed, backed by a group of 170 companies that include Amazon, Apple, and Google, as well as SmartThings, IKEA, and Samsung. It is open source and, in principle, should allow a Google Home speaker to work with, say, an Amazon Ring doorbell. In markets where consumers really do want interoperability, it can emerge without a regulator requiring it, even if some companies have apparent incentive not to offer it.

If You Build It, They Still Might Not Come

Much of the case for interoperability interventions rests on the presumption that the benefits will be substantial. It’s hard to know how powerful network effects really are in preventing new competitors from entering digital markets, and none of the more substantial reports cited by the “Super Tool” report really try.

In reality, the cost of switching among services or products is never zero. Simply pointing out that particular costs—such as network effect-created switching costs—happen to exist doesn’t tell us much. In practice, many users are happy to multi-home across different services. I use at least eight different messaging apps every day (Signal, WhatsApp, Twitter DMs, Slack, Discord, Instagram DMs, Google Chat, and iMessage/SMS). I don’t find it particularly costly to switch among them, and have been happy to adopt new services that seemed to offer something new. Discord has built a thriving 150-million-user business, despite these switching costs. What if people don’t actually care if their Instagram DMs are interoperable with Slack?

None of this is to argue that interoperability cannot be useful. But it is often overhyped, and it is difficult to do in practice (because of those annoying trade-offs). After nearly five years, Open Banking in the UK—cited by the “Super Tool” report as an example of what it wants for other markets—still isn’t really finished yet in terms of functionality. It has required an enormous amount of time and investment by all parties involved and has yet to deliver obvious benefits in terms of consumer outcomes, let alone greater competition among the current accounts that have been made interoperable with other services. (My analysis of the lessons of Open Banking for other services is here.) Phone number portability, which is also cited by the “Super Tool” report, is another example of how hard even simple interventions can be to get right.

The world is filled with cases where we could imagine some benefits from interoperability but choose not to have them, because the costs are greater still. None of this is to say that interoperability mandates can never work, but their benefits can be oversold, especially when their costs are ignored. Many of mandatory interoperability’s more enthusiastic advocates should remember that such trade-offs exist—even for policies they really, really like.