Certain firms in the Internet economy may exclude competitors by refusing to deal data. Such conduct may impede innovation. But antitrust law lacks a coherent response to monopoly of data. This Comment proposes a policy inspired by duties to share. Over a century ago, courts devised an “essential facilities” doctrine that required monopolists to share inputs essential to competition with rivals. These inputs included phone lines and bridges. I contend that the essential facilities doctrine sometimes should require open access to data.
This Comment proceeds in two Parts. Part I describes the problems with data monopolies and provides an example of an essential data dispute. The Part goes on to explain the essential facilities doctrine and identify criticisms that led to the doctrine’s rejection. It closes by describing an essential data claim. Part II contends that criticisms of the essential facilities doctrine attenuate when a dataset becomes the facility to which a rival seeks access.
Part I has three sections. Part I.A explains the role of data in the online economy and provides an example of an essential data claim. Part I.B introduces the essential facilities doctrine, as well as the doctrine’s demise. Part I.C sets forth the elements of an essential data claim and situates the concept in commentary and precedent.
Sometimes data cause disputes. A company called PeopleBrowsr faced one late in 2012. According to PeopleBrowsr, its service helped clients monitor and analyze conversations online and relied on data from a social network called Twitter.1 PeopleBrowsr also claimed that it had used Twitter data for years.2 But Twitter told PeopleBrowsr that the social network would revoke access to its data at the end of November 2012.3 Twitter alleged that its business model had evolved.4 According to PeopleBrowsr, Twitter thought the monitoring company no longer “fit.”5
PeopleBrowsr alleged that a Twitter shutoff “would [have] devastate[d] PeopleBrowsr’s business.”6 So the company stated that it negotiated with Twitter for access.7 PeopleBrowsr said that negotiations failed and then it sued.8 Shutoff, PeopleBrowsr said, would violate California competition law.9 “[C]ompetition in the market for analysis of Twitter data” would founder and innovation in the data’s use would slow.10 Not so, Twitter said: shutoff preserved the incentives of entrepreneurs to innovate and violated neither California nor federal antitrust law.11 Twitter, the company said, “has the right to control its data.”12
This Comment challenges that and similar claims. Refusals to deal data can help firms free ride on rivals’ investments and maintain monopolies by excluding competitors.13 But courts supply no consistent response to the antitrust questions that data pose: PeopleBrowsr and Twitter settled and PeopleBrowsr got access for about eight months.14 Two software developers a decade apart sued online marketplaces for withholding data and got no answer on their antitrust claims.15 The Federal Trade Commission in 2011 reportedly opened an inquiry into claims that Twitter hobbled a potential rival by revoking access to data.16 The Commission never filed a complaint.17
Antitrust law generally preserves the “right[s] of trader[s] or manufacturer[s]” to choose the “parties with whom [they] deal.”18 But in “limited circumstances” a refusal to deal violates Section 2 of the Sherman Act, which prohibits monopolization.19 Under the essential facilities doctrine, a duty to deal arises when a monopolist refuses to share inputs essential to competition despite the feasibility of doing so.20
The essential facilities doctrine dates at least to 1912, when “a group of railroad operators obtained . . . the only railroad bridges across the Mississippi River at St. Louis.”21 Because the “most extraordinary” topography of the region rendered it “impossible for any railroad company to pass through . . . without using [the group’s] facilities,” the Supreme Court required that the group deal with outsiders on “just and reasonable terms.”22
The Supreme Court never adopted the essential facilities doctrine by name.23 But lower courts and commentators drew on the doctrine.24 The Court in 1972 made a power company share transmission wires with the company’s rivals.25 A decision of the Court a decade later required two ski mountains to continue offering a joint ticket after one sought to withdraw.26
Today, little remains of the essential facilities doctrine. Commentators weakened the doctrine with three criticisms. First, monopolists could not extract additional profits from consumers by refusing to deal.27 So efficiency and not exclusion likely motivated behavior scrutinized by the essential facilities doctrine. Second, the doctrine weakened incentives to compete: dominant firms would not erect infrastructure lest a court appropriate the investment for a rival’s use.28 Finally, the doctrine placed courts into the role of regulators, though they lacked the capacity to administer the sharing that the doctrine required.29 These concerns held sway: the Supreme Court in 2004 denied “ever recogniz[ing]” the doctrine of essential facilities.30
This Comment argues that a claim to essential data—data essential to competition—should require the same elements as a claim to an essential facility. First, the monopolist must control and deny access to the data that the plaintiff seeks.31 Second, competition must fail without the data.32 Third, the plaintiff must lack means to duplicate the data.33 Fourth, the monopolist must have means to share the data.34 Fifth and finally, an essential facility plaintiff must demonstrate the defendant’s monopoly power in an antitrust market.35
Several recent claims fit this description. One is Twitter’s attempt to disconnect PeopleBrowsr, discussed at the beginning of this Part. A second relates to a 2000 dispute between eBay and Bidder’s Edge, an aggregator of auction prices36: eBay, which reportedly controlled eighty-seven percent of auction traffic,37 refused to deal with an ecosystem firm that made tools for users to access auction prices.38 A third concerns a dispute that reached federal court in 201239: Craigslist, a dominant provider of online classifieds, sued 3Taps, a start-up that obtained and shared data based on Craigslist’s classifieds.40
Each dispute started with data created by users of a monopolist’s platform. Competitors could not duplicate the data because of network effects: each user who used the monopolist’s platform made that platform more valuable to every other user. So no competing dataset emerged. For example, the set of messages that Twitter controlled faced no competition from a rival network: users who wanted to listen went where people were talking. The set of prices that eBay controlled faced no competition from a rival auction: sellers went where people were buying.41 The disputes in each case involved refusals to deal monopolized inputs protected by barriers to entry. Those circumstances invite the application of the essential facilities doctrine.42
This Part contends that criticisms of the essential facilities doctrine attenuate when rivals invoke the doctrine against a defendant that has withheld data. Refusals to deal data may raise monopoly profits and lower consumer welfare. Essential data remedies benefit consumers without depriving innovators of incentives to invest. Finally, courts may administer access to data more easily than access to physical facilities.
Data essential to competition—essential data—can exist when firms act alone or with others.43 Courts have forced access in the latter case but not the former.44 Firms that act alone may originate data or build platforms for others to originate data. Microsoft and Intel, for example, originated technical data45 essential to competition in downstream markets. Marina Lao has argued for access to data in such a case.46 This Comment extends the case for access to platforms: that is, to firms whose data monopolies derive from users who must originate data to consume the functionality that the firms’ technology enables.
Critics of the essential facilities doctrine begin by asking why monopolists would refuse to deal. Single monopoly profit theory holds that monopolists may extract monopoly rents from a market without selling to consumers.47 Assuming there is a competitive market for the end product whose input the monopolist controls, monopolists may charge downstream firms one fee or royalty per product and thereby induce downstream firms to produce only the monopoly quantity. The monopolist could do no better if it sold to consumers itself.
Data monopolists in emerging industries lie beyond this model. First, one dataset can supply zero or infinite final goods and services. When 3Taps gets data from Craigslist ads, 3Taps can serve that data to anyone who wants to see goods and services listed for sale on Craigslist’s exchange. Second, a data monopolist may lack the ability to monitor the quantity of final goods and services produced using the monopolist’s data. When Bidder’s Edge scrapes prices from eBay, eBay may never learn that a user has viewed those prices on Bidder’s Edge. As a result, uncertainty may blur the final demand curve that a data monopolist faces.
If the monopolist cannot predict final demand and must make sunk investments in order to enter the final market, the monopolist may prefer temporarily to deal with a downstream rival. The rival’s success or failure provides a proxy for otherwise unobservable final demand. If the data monopolist retains the ability to terminate the rival’s access, then the monopolist has obtained a costless option on a downstream market. For example, PeopleBrowsr said that it started analyzing Twitter data on the basis of the social network’s promise to make that data available.48 Twitter, PeopleBrowsr said, refused to share data only after PeopleBrowsr demonstrated the existence of a lucrative market for analytics.49
So a data monopolist might pursue a strategy of free riding that ends with a refusal to deal. That means that uncertainty about market opportunities makes real the monopolist that Judge Richard A. Posner could imagine only “with difficulty”: the monopolist who “entice[s] new firms into its market only to destroy them.”50 If a plaintiff cannot bring an essential data claim to mitigate the threat of exclusion, then the risk of entry—and, therefore, the cost of innovation—will rise.
Data monopolists might also refuse to deal in order to protect their monopolies. A monopolist may fear that a downstream rival’s tools will eventually supplant the monopolist’s product altogether.51 If the monopoly product provides to users data produced by a network, then refusal to share those data may impede a rival that seeks to develop a competing product. For example, according to Craigslist’s rivals, they cannot promise market prices to buyers without Craigslist’s data.52 Or a social network’s rival might choose to challenge the network by first attracting users with messages passed on the incumbent network.53 Refusing to deal data forecloses such a tactic.
Data monopolists may have multiple motives for refusing to deal data to potential competitors. Courts should not automatically ascribe to those refusals the procompetitive explanations put forward by critics of the essential facilities doctrine.54
Critics next charged that the essential facilities doctrine distorted firms’ incentives to invest.55 The prospect of future antitrust liability “could significantly reduce the incentive of entrepreneurs to innovate in areas . . . involv[ing] essential facilities.”56 The Supreme Court’s Trinko decision adopted this argument,57 which has since met with approval in courts of appeal.58
However, antitrust law does not offer clear guidance about when a defendant’s argument about reduced incentives will suffice to rebut an essential facilities claim. The Supreme Court last ruled for an essential facilities plaintiff in Aspen Skiing Co. v. Aspen Highlands Skiing Co.59 Aspen’s unanimous Court held that the prospect of profits from “exclusionary” conduct does not justify a refusal to deal.60 But as Einer Elhauge asserts, “[m]onopolization doctrine currently uses vacuous standards and conclusory labels that provide no meaningful guidance about which conduct will be condemned as exclusionary.”61 So data monopolists’ claims that refusals to deal protect “incentives of companies to innovate and compete,”62 or reduce “free-riding on [the monopolist’s] substantial investment of time, effort, and expense,”63 presuppose the sufficiency of business justifications that antitrust law has yet to accept.
Moreover, the incentive claim rests on the ex ante expectations of entrepreneurs, but the experience of the data monopolists identified in this Comment suggests that facility ownership did not motivate entry into the markets that the monopolists came to dominate. For example, Twitter launched a tool to connect; early monetization discussions revolved around advertising.64 Craigslist began as its founder’s events circular; the site remained “wedded to the idea that [it] was a community service” years after its launch.65 eBay’s first revenues came from transaction fees, and its business plan predicted future revenue from software licensing.66
This could change. Data licensing revenues at Twitter rose almost fifty percent in 2013, to $70 million.67 In April 2014 the company bought Gnip, a data reseller.68 LinkedIn, a social network for professionals, received most of its 2013 revenue from hiring professionals who bought access to the network’s data.69
The application of any essential data doctrine to those who invest in pursuit of data monopolies will require finesse. But essential facilities precedents supply a framework for such a future: the Supreme Court’s refusal-to-deal precedents impose sharing only after reviewing a monopolist’s reasons for exclusion.70 Whether those reasons may include monopoly profits, if the profits require exclusion and motivate investment, remains an unresolved question.71 Partial answers exist; courts and scholars view with skepticism justifications advanced by monopolists who deal with some, but not with rivals.72 Moreover, courts may scrutinize proffered justifications for pretext, safeguarding ex ante incentives only where those incentives are endangered.73
Finally, critics questioned courts’ capacity to identify and remedy anticompetitive refusals to deal.74 This administrability critique asserted that generalist judges’ reviews of novel practice and complex economics for “exclusionary” conduct became risky affairs.75 And once judges condemned refusals to deal, they could not enforce remedies without taking on the burdens of a regulator.76
Essential data remedies escape some of these criticisms. First, the nonrivalrous character of data “facilities” relieves courts of the analytical effort otherwise required to prevent “congestion through competing uses” of physical facilities with finite capacity.77 Second, the data monopolist faces costs of sharing that likely approach zero;78 any nonzero costs likely arise from markets for commoditized infrastructure, such as servers, bandwidth, or processors. Therefore, no sustained judicial inquiry into industry idiosyncrasies or extant plant characteristics would be necessary to determine the sharing costs borne by data monopolists.79
Finally, courts can preserve incentives to invest by permitting data monopolists to recover their average total costs.80 Courts have long paired cost recovery with the essential facilities doctrine.81 The standard—which includes a reasonable return on capital—“reflects equilibrium in the market for investment.”82
The argument that innovators and their backers require higher than reasonable returns to capital presumes that one firm—but not others—can identify a superior investment opportunity. Theories of efficient capital markets, however, deny that such opportunities for arbitrage exist.83 Together, these observations suggest that courts may more easily administer access to data than to physical facilities.
The benefits of access to data should also enter the analysis of courts confronted with claims to essential data. In no markets do monopoly prices produce static deadweight loss more than in markets for information.84 Further, refusals to deal essential data stall innovation. Data power many applications: Twitter data have predicted social unrest and power outages and directed humanitarian aid.85 In that regard, data resemble technologies that support multiple rounds of innovation.86
Innovation scholars suggest that unqualified control of such technology “tends to hinder technical progress.”87 First movers focus on past experiences88 or lack expertise to develop all applications.89 Improvers who would build on data may fail to secure permission to do so because monopolists hold divergent beliefs about an improvement’s value.90
Broad exclusion rights favor innovation and consumer welfare when “the overall potential for modifications and improvements based on the original achievement is relatively clear and bounded.”91 Little suggests that essential data fit that description92: AOL failed to recognize that transaction data could power the recommendation engine that Amazon built.93 Yahoo considered creating a spell check tool from users’ search engine queries, but it was Google that actually pursued the project.94
The welfare case for an essential data doctrine has caveats. I have simplified questions of access quality over which parties have litigated in the last twenty years.95 I have assumed that conduct that resembles both exclusion and justified competition excludes with frequency sufficient to justify scrutiny.96 But the prima facie case remains: revitalizing essential facilities in the context of data may speed innovation and increase consumer welfare.97
This Comment has argued that criticisms of the essential facilities doctrine carry less weight when a dataset becomes the facility. In the data context, the essential facilities doctrine captures suspect conduct and better withstands criticisms linked to ex ante incentives. Remedies that enforce access to data would entail less judicial inquiry into costs of service and facility capacity. The case for any essential data doctrine will evolve with the objectives of aspiring data monopolists. But that case will always build on the allocative inefficiencies of information monopolies and their negative effects on innovation with data.
Essential data implicate networks through which consumers connect and transact—activities fundamental to Internet economies. Courts and agencies should consider whether a doctrine devised to safeguard competition in the last century has become more salient in this one.