Potential Competition in Platform Markets
It would seem to be impossible to speak publicly about antitrust today without at least touching upon issues relating to platform markets. We have all heard the calls for change. I put them in two distinct buckets. In the first bucket, we have calls to reinvigorate traditional conduct-based antitrust analysis, by enforcing existing laws more aggressively or by reevaluating our standards of proof. In the second bucket are calls to break up platforms that are deemed too big, to impose line-of-business restrictions to prevent such platforms from expanding into adjacent markets, and to tax and thereby discourage the use of certain disfavored business models. Underlying the second set of proposals is the belief that we must discard, or at least supplement, traditional antitrust enforcement to address the unique challenges presented by platform markets. Often missing in these calls for reform is a clear link between a firm’s conduct and its potential liability. History has taught us that without guidance on what conduct triggers liability, it is not obvious what a firm can do to reduce its risk of antitrust liability short of competing less aggressively and, as a consequence, losing market share to its rivals.
At the outset, it is helpful to step back and ask why we are hearing calls for increased antitrust enforcement in the first place. One reason is surely the ubiquity of today’s large platform businesses. We have seen large firms before, but perhaps not firms that touch so many aspects of our lives. A second reason, no doubt related to the first, is that many of these platforms are subject to pronounced network effects and scale and scope economies that may make them prone to tipping. In such environments, platforms can quickly gain market share and may enjoy market power, depending on the strength of network effects and scale and scope economies, switching costs, whether agents multi-home, and other barriers to entry.
In my view, our current antitrust enforcement regime works and is adaptable to the current environment. That is not to say that antitrust enforcement is perfect. For example, consider the case of the serial acquirer, a firm that slowly grows its share through a series of small acquisitions until it accounts for a sizable share of the market. It is exceptionally hard to establish that any individual acquisition leads to a substantial lessening of competition under the Clayton Act. Yet when we step back and look at the totality of the evidence, it is clear that a focus on individual transactions makes for a very blurry snapshot of what is happening in the market. Relying solely on the Clayton Act to evaluate and challenge this pattern of behavior might lead to under enforcement. I’ll return to this example later.
Today, I would like to discuss potential competition, a term that may mean different things to different people. I am focusing this evening on the case of a large incumbent platform acquiring one or more start-ups operating in adjacent markets that pose a potential competitive threat to the incumbent’s platform. I am not talking about the acquisition of a start-up that is already competing against the incumbent platform, even though such a start-up might have the potential to increase its market share if competition were allowed to run its course. In my view, such an acquisition is distinguishable because it involves actual as opposed to potential competition.
My focus is deliberately narrow. I will not address the broader debate about digital markets. I will not be talking about competition in multi-sided markets or big data, and I will not address calls for reform stemming from privacy or other concerns. Still, at a time when some question the efficacy of traditional antitrust enforcement, we benefit from slowing down and asking exactly what is or is not working in a specific circumstance.
Let me begin with an observation we can all agree with: potential competition cases in platform markets are hard. There are a few reasons for this. First, a case would usually involve an allegation that the start-up, were it not being acquired, would reposition itself to compete against the incumbent platform. To make this case we need evidence that this would happen but for the acquisition. Usually we would look to strategic documents and emails, in addition to deposition testimony, about the firm’s future plans and capabilities. But start-ups are by nature small firms. They may not have the robust processes in place that reliably generate these strategic documents, and other means of communications are less formal than we see in larger firms, with fewer memos and emails and more chatroom activity, which may or may not be archived.
A second problem is that our current economic tools are much better suited to a world in which we have evidence on current diversions and margins. Think merger simulation, UPP, and critical loss. By comparison, we don’t have the tools to accurately estimate the likelihood of, and consumer benefits from, entry and repositioning. To be clear, this is not a comment on the state of the economic research in this area, but rather the inherent difficulty of the exercise.
Finally, the incumbent platform often may have a credible merger-specific efficiencies story. The start-up may offer some functionality that complements, or has the potential to complement, the incumbent platform. Through acquisition, the incumbent platform may be able to integrate the two products or services more tightly or draw upon its resources and expertise to bring the start-up’s product to a much wider audience. The incumbent may also be able to utilize engineers at the start-up more efficiently by deploying them on a range of technical issues faced by the larger company.
So potential competition cases in platform markets are hard. Why not just make them easier to bring by, say, shifting the burden of proof to defendants or lowering our standards of proof?
Well, we need to be very cautious. At the risk of oversimplifying, consider an entrepreneur with two innovation paths. Path one: the entrepreneur can innovate to compete directly against the incumbent platform. Path two: the entrepreneur can innovate to complement the incumbent platform with the hope of being acquired. This type of innovation is very common with operating systems, where we see small start-ups developing functionalities that expand or refine the capability of the core operating system. These start-ups are often acquired and their technologies are then incorporated into the operating system itself.
We can and should debate which of these two innovation paths provides greater benefit to consumers. But we shouldn’t lose sight of the fact that often it is the entrepreneur who chooses which path to follow. When we make potential competition cases against start-ups in adjacent markets easier, we may well depress the profitability of innovating to complement the incumbent platform. Some entrepreneurs will instead choose path #1, innovating to compete directly against the incumbent platform. Other entrepreneurs, however, may stop innovating altogether. Innovating to complement, after all, can be a much safer business model, for entrepreneur and venture capitalist alike, than rolling the dice to take on the incumbent platform directly.
Of course, we might conclude that these costs are worth it, that we would rather have more entrepreneurs innovating to compete against incumbent platforms even if it means reduced innovation in adjacent markets. But on what basis do we come to that conclusion? I know of no empirical support for the proposition that the government should favor one form of innovation over another.
Some scholars have proposed taking a probabilistic approach to potential competition in such markets. The approach is best understood by example. Suppose a start-up has a 1 in 20 chance of unseating a dominant incumbent platform and thereby generating $1 billion in consumer benefits. If instead the start-up and the incumbent merged, they would more tightly integrate their services and create, say, $40 million in consumer benefits. The proposal would involve blocking the merger if the expected harm exceeds the expected benefit. Here, the expected merger harm is $50 million, or 5% of $1 billion. The expected benefit is the assumed efficiencies of $40 million. Because expected harm exceeds expected benefit, this approach suggests the transaction should be blocked. In this example, it would be suggested that the government seek to block a transaction even though there is only a 1 in 20 chance that it would result in any harm to competition. Obviously, this would mark a dramatic departure from how we approach such acquisitions today in the U.S.
A few remarks on the proposal. First, I agree with and applaud the emphasis on consumer surplus, as opposed to whether the acquisition disadvantages competitors or is outside the incumbent platform’s core line of business.
That said, there are reasons to be skeptical about the proposal. Simply put, it is not clear that we currently can estimate reliably the probabilities and other inputs we would need to weigh expected harms and benefits. Courts have a hard enough time determining whether a merger between two existing competitors is anticompetitive. Now we are asking the court to assess: 1) the probability that a start-up would reposition its product or service to compete against the incumbent platform but for the merger, 2) the probability the market tips in favor of the incumbent or the upstart, and 3) consumer surplus in each of these states of the world. I’d give a graduate student writing a Ph.D. dissertation a fighting chance of making such a calculation with the help of a few dozen assumptions and several years to conduct the research. A judge holding a two-week trial with a full docket might have a harder time.
Of course, we should not discard a methodology simply because it is difficult. Here, however, the decision to challenge an acquisition turns on probabilities that we cannot estimate with any degree of precision using existing economic tools. In the example I gave above we sued to block the transaction because there was a 5% chance that consumers would benefit. If instead the probability were 3%, we’d let the transaction through. With so much hanging on small, imprecisely estimated probabilities, judges may simply retreat to their prior beliefs as to whether a transaction harms competition. Ironically, by placing greater demands on economic analysis in such cases, we might render it less relevant.
Thus far I have argued that potential competition cases are difficult, but that the solution isn’t to make antitrust challenges easier. Dramatic changes to our merger enforcement regime threaten to alter innovation incentives in ways that may harm consumers despite the best intentions.
What then is the solution?
One solution may be Section 2 of the Sherman Act.
As I mentioned earlier, we’re concerned about acquisitions of nascent competitors in platform industries because these markets are prone to tipping, and with tipping comes the potential for durable market power and substantial barriers to entry. Anticompetitive conduct by firms seeking to maintain or acquire monopoly power is precisely what Section 2 is intended to address.
The immediate implication of using Section 2 to evaluate potential competition is that it allows us to step back and put greater emphasis on a pattern of conduct, including past acquisitions. I mentioned serial acquisitions earlier. In that context, with many acquisitions involving targets with market shares of 5-10%, we tend to miss what is happening in the market when we look at each transaction in isolation. We also struggle to identify which single transaction leads to a substantial reduction in competition. That same logic applies to the acquisition of start-ups operating outside a platform’s core market.
Putting greater emphasis on a pattern of conduct and acquisitions has a direct impact on the types of data and documents we would seek in an investigation. We may need more extensive data, both over time and across geographies, to evaluate the competitive impact of past transactions in adjacent markets. We might ask whether acquired start-ups’ products evolved differently than those of independent firms. To the extent those start-ups’ products complemented other platforms, were those functions disabled or deprioritized post acquisition? Senior management’s strategic plans for future acquisitions also would shed light on the rationale behind the incumbent’s acquisitions. Similarly, a pattern of unrealized efficiency claims would suggest a campaign of defensive acquisitions.
We might also seek documents regarding the incumbent’s business practices vis-à-vis start-ups. Did those practices vary predictably based on the threat posed by the start-up to the incumbent’s platform? Did those practices change as start-ups began to reposition their products to compete more directly against the incumbent?
Of course, challenging acquisitions as part of a broader pattern of conduct presents hurdles that we don’t face under the Clayton Act. For monopoly maintenance, we must prove the acquirer has monopoly power; for attempted monopolization, a specific intent to monopolize and a dangerous probability of achieving monopoly power.
There are ways, however, to meet these elements, and the presence of network effects could be useful in proving such a case.
As mentioned, pronounced network effects and economies of scale and scope can render platform markets susceptible to tipping. Once a market tips in favor of a given platform, we would expect that platform to enjoy significant competitive advantages over its rivals and to price accordingly, with prices well above competitive levels. If there are substantial barriers to entry, market power in these circumstances could be durable. In this light, one might ask whether traditional share-based thresholds for monopoly power may be too high.
We also may want to consider whether a firm’s acquisition or series of acquisitions had the purpose and effect of achieving monopoly power. Mindful of the competitive dynamic in markets prone to tipping, strategic business documents may identify the scale the platform needs to tip the market and lay out an acquisition strategy to effect that end. Such documents may well suffice to show a specific intent to monopolize and block future entry.
To be clear, it is not unlawful for a firm to obtain a monopoly in a tipped market through organic growth by offering a superior product or business acumen. Indeed, in markets with strong network effects, consumers may benefit from having only one or two large platforms with many participants. The problem arises when a firm obtains or maintains a monopoly through exclusionary conduct rather than on the merits. A pattern of acquisitions with the purpose and effect of squashing potential entrants potentially could qualify as such anticompetitive behavior under Section 2.
That is not to say that using Section 2 to evaluate and potentially challenge such acquisitions will be any easier than challenging under the Clayton Act. Rather, I am arguing that in many respects Section 2 may be a better fit than the Clayton Act given the types of concerns that arise from acquisitions of potential competitors. Consider the parallels to what I have discussed today and U.S. v. Microsoft in the following quote from the D.C. Circuit’s opinion:
[T]he question in this case is not whether Java or Navigator would actually have developed into viable platform substitutes, but (1) whether as a general matter the exclusion of nascent threats is the type of conduct that is reasonably capable of contributing significantly to a defendant's continued monopoly power and (2) whether Java and Navigator reasonably constituted nascent threats at the time Microsoft engaged in the anticompetitive conduct at issue.
This logic might extend to acquisitions of potential competitors by platforms with monopoly power. The acquisition of a potential competitor may be an “exclusion of a nascent threat” to borrow the words of the D.C. Circuit. So we would ask, under Microsoft, whether the acquisition of a potential competitor “is reasonably capable of contributing significantly to a defendant’s continued monopoly power.” Defendants would then have an opportunity to provide evidence on business justifications and efficiencies, as the framework of Microsoft allows.
Applying a different standard to firms with monopoly power makes good sense as a matter of economics. In markets with several small or mid-size firms, an individual firm’s incentives to acquire a potential competitor are relatively weak. While failing to acquire a potential entrant will result in lower prices and profits, those costs will be borne by all firms. In contrast, in markets that are prone to tipping, a dominant firm with monopoly power stands to lose all its rents to a successful entrant. It would not be surprising in the least to learn that such a firm dedicated considerable resources to identifying and acquiring potential rivals well before they emerge as legitimate challengers to its monopoly. In this light, additional scrutiny and unique treatment of such acquisitions may well be warranted.