Antitrust and Tech: One Network to Rule Them All?
The argument that Amazon, Apple, Facebook, Google, and Microsoft — dubbed the “Frightful Five” in the pages of the New York Times — pose imminent threats to fair and competitive markets are grounded on two claims: their advantages of network effects and big data. This post argues that proponents of these claims have overstated the anticompetitive implications of network effects and big data in the recent past, making these claims a poor guide for the future of U.S. competition policy.
First, critics point out that social media platforms, clouding computing services, online marketplaces and the like benefit from non-linear network effects. Facebook has attracted over one billion users because it had 500 million users in prior years. Firms that benefit from strong network effects have a natural tendency, like the physical property of inertia, to see increasing market share, requiring antitrust intervention to trim their size or closely watch their conduct towards competitors.
Second, critics note that the tech firms that dominate the S&P 500 have accumulated troves of data on their users, advertisers, and competitors. Data has increasing returns to scale; the firms with the most data will have more fine-tuned pricing, search result, and advertising algorithms. The forthcoming book Reinventing Capitalism in the Age of Big Data by Viktor Mayer-Schönberger and Thomas Ramge argues that as firms depend more on AI and machine learning to glean insights and make progress, the financial gains from innovation will reward firms with stronger market positions and user bases than their competitors.
The more conspiratorial-minded fear that tech giants, who have positioned themselves as intermediaries in digital advertising and cloud computing, can weaponize their algorithms and data against their smaller competitors. How can incubating startups, who depend on AWS and Azure to deploy and scale their applications, compete on fair terms when their activity is under the careful watch of Amazon and Microsoft?
Timothy Taylor, managing editor at the Journal of Economic Perspectives, argues in a recent blog post that these two marks against today’s tech giants aren’t persuasive justifications for antitrust intervention. Simply put, in the two decades of booming (then busting) information technology firms, the incumbent advantages of large user bases and troves of data collected from them have been poor predictors of continued success.
In October of last year, Chris Koopman and I wrote an op-ed for The Hill highlighting a notable example of the high levels of churn in information technology — AOL Instant Messenger, or AIM. In the early 2000s, AIM had huge market share in the instant messaging space, the subject of much hand-wringing at the FCC and FTC. In order for the AOL-Time Warner merger to receive the FCC’s approval in 2001, the sure-to-be Internet-media juggernaut accepted several behavioral restrictions and made several promises to the agency.
One sticking point with antitrust watchdogs was AIM. The agency and several economists argued AOL Time Warner was poised to monopolize the instant messaging services market — network effects predicted that large platforms would only get larger. To give the good people at the FCC a good night’s rest, AOL Time Warner agreed to focus significant engineering efforts to making its instant messaging service interoperable. That is, they would make the platform work such that consumers wouldn’t have to set up an AIM account to chat with AIM users. Until the platform was interoperable, AOL Time Warner was forbidden from leveraging its broadband networks to introduce videoconferencing and other new features.
The platform was never made interoperable, yet AIM’s hold on the market weakened in only a couple of years. Having made good on its promise to level the instant messaging playing field among AIM, Yahoo, and Microsoft, the FCC approved a petition by AOL Time Warner to free it from engineering purgatory.
What did the network effects hypothesis that predicted AIM’s imminent dominance fail to account for? One possible explanation is that the network effects were strong, and the FCC’s intervention simply worked to undermine them. After all, the FCC put restrictions on AOL Time Warner to check AIM’s growth, and voila, its market share feel within a few years! But if network effects are as strong determinants of AIM’s growth as the FCC believed them to be, then AIM should have only continued to outpace the growth rates of its competitors until it was made interoperable. AIM was the biggest game in town during the merger review, so without interoperability most new consumers should have been pulled into AIM.
Yet Microsoft and Yahoo appeared to be closing the gap with AIM as early as 2003. Economist Gerald R. Faulhaber and and computer scientist David J. Farber, who lent strong support to the interoperability requirement in the merger order and wrote against AOL Time Warner’s petition to be relieved of it, acknowledged in April of 2003 that “there are two stable competitors to AOL Time Warner, and that is at least suggestive (if not dispositive) that market tipping is less of an issue today than at the time of the case.”
The issue with the network effects hypothesis is that it assumes the costs of setting up and managing multiple IM accounts, social media accounts, what have you, are prohibitively high. In the 2001 Merger Order, the FCC writes,
Using several IM services … entails much inconvenience. A user must download several kinds of IM software; must register and maintain accounts, unique names, and passwords with several IM providers; must use each one enough to become comfortable with its ‘look and feel’; must keep several buddy lists and remember which buddies are on which IM service (and with what names); and must keep several IM sessions open simultaneously.
The claim is that consumers will pick only one IM service, social media network, or search engine. In which case, why not pick the most popular one? To the contrary, consumers in the IT services space have shown incredible flexibility to try new services and experiment with multiple similar services. This phenomenon, called “multihoming”, has contributed to considerable churn in the tech sector for two decades: AIM gave way to MSN and Yahoo Messenger gave way to Gchat and Slack; MySpace and Friendster gave way to Facebook, Snap, WhatsApp, and Twitter; eBay gave way to Amazon; and Netscape gave way to Internet Explorer gave way to Chrome.
A paper by David S. Evans and Richard Schmalensee published in the Cato Institute’s Regulation (quoted at length by Taylor) argues that the same market churn in tech casts doubt on the claim that the market will strongly favor incumbents that already have amassed troves of consumer and retail data. Google emerged as the dominant player in online search, but only after years of competition with other search firms like AltaVista that surely amassed their own cache of search and user data. Further, Google’s search share does not appear to be asymptotically converging to one. Indeed, its market share has seen downward adjustments when Bing and Yahoo see upticks, and Amazon has put a serious dent in Google’s share of e-commerce search. A 2017 survey estimates that about half of online retail searches started at Amazon.
To state the point generally, today’s top tech companies started with no users, no customers, no retailers, and none of their data, and were still able to get a toe-hold in markets with incumbents that had plenty of data to inform their business strategies. These incubating startups came to topple the giants of their time, rendering market capitalizations in 2000 a poor predictor of market capitalizations in 2008, and 2008 market capitalizations a poor predictor of market capitalizations in 2018.
Yet two decades into the IT revolution, antitrust hawks continue to presume that network effects and big data will allow large tech firms to squash new entrants, or simply consume emerging threats via acquisition. Some have called for a heavy hand — only breaking up the firms or regulating them like public utilities will do — while others seek a lighter touch. Favoring the latter, The Economist recommends that the tech firms be required to share anonymized data with competitors, analogous to AIM’s obligation to make its large network of users accessible to smaller IM players.
We have not seen the former approach, yet, but the case for breaking up tech firms is self-defeating. If network effects are as strong as critics claim, then breaking up Facebook, Amazon, or Google will simply kick-start a new race towards a concentrated tech market.
The light touch approaches have long been advocated and tried, and AIM’s case highlights a general pattern of “creative destruction” that shows that fretting about network effects and big data is much ado about nothing. Instead of seeing tech as a winner-take-all market with entrenched power players, we should hope the decades of dynamism in information technology spills over into other stagnant sectors in the American economy.¹
¹ See Tyler Cowen, The Complacent Class, St. Martin’s Press (2017), and Michael Mandel and Bret Swanson, “The Coming Productivity Boom: Transforming the Physical Economy with Information,” Progressive Policy Institute (2017).
