Chapter 7 of our Digital Platforms and Concentration conference ebook looks at some critical junctures in the history of competition between Big Tech companies and explores how, even when regulators have detected anticompetitive behavior, the remedies they’ve imposed have done little to solve the problem.
Editors’ note: See also our 2017 conference volume Is There a Concentration Problem in America? Other chapters from this year’s ebook can be read at the following links:
It is rare for antitrust issues to reach public consciousness, but the question of what to do, if anything, about the dominant Internet firms of the day has clearly done so. The GAFA, as they are known in Europe—Google, Apple, Facebook, and Amazon, and let’s throw in Microsoft for good measure—have each achieved a remarkable market position in the technologies that seem to define so much of the modern economy, especially the consumer-facing economy. While each of these companies benefits from dynamics of platform economics, these firms are interestingly different and it is important not to lose sight of that as we consider possible regulatory responses to their market positions.
The rules of competition and antitrust are perhaps most important at the point where we have competition in a market adjacent to a market held by a dominant firm, especially where there is the promise that the adjacent market could turn into a springboard for competition back into the original market. Adjacency matters as it is quite difficult to attack a dominant firm in its home market, though even that might be possible if we have a leading firm in one market entering the market of another dominant firm.
Here, I look at the two most developed examples we have of the role of antitrust in adjacent market competition in platform industries. We have an extensive history for Microsoft and a now growing one for Google. Both situations show the difficulty of achieving meaningful remedies even when antitrust violations are found.
Two Decades of Microsoft Antitrust
It is interesting that it is the GAFA and not the GAFAM. Inspired by a magazine cover—Popular Electronics, January, 1975— Bill Gates and Paul Allen started Microsoft as a computer languages company to write the BASIC computer programming language for the new Altair 8800 personal computer. Today’s average computer user wouldn’t give the Altair 8800 even a glance and would probably be stunned to learn that this was the beginning, but it was. That world evolved quickly and reached a turning point on August 12, 1981, when IBM launched its first personal computer. It wasn’t obvious then that in doing so IBM would create two monopolies and yet would not end up with either. Intel and Microsoft both were defined by the success of the IBM PC and the clones that would follow from it.
Microsoft’s success is even more remarkable as Microsoft originally sent IBM elsewhere when IBM asked Microsoft to provide an operating system for its new computer. And IBM would eventually release its new computer with three different operating systems (bonus points if you can name the other two). But within a decade, Microsoft was an antitrust target. The US Federal Trade Commission started investigating Microsoft in 1991 and was believed to be looking at whether Microsoft was using its market position in operating systems to thwart competition in adjacent markets such as those for word processors.((Lawrence M. Fisher, Microsoft In Inquiry By F.T.C., New York Times, Mar. 13, 1991, pD1.)) Two years later, the FTC was stalled with an even 2-2 split on whether to bring an action against Microsoft.((John Markoff, F.T.C. Stays Deadlocked On Microsoft, New York Times, July 22, 1993, pD4.))
But the US has two federal antitrust agencies and with the FTC at a standstill, the US Department of Justice jumped in.((John Markoff, Justice Department Considers Inquiry on Microsoft, New York Times, Aug. 1, 1993, p33.)) By July, 1994, the government announced a settlement with Microsoft that would require it to alter its licensing practices for MS-DOS.((US v. Microsoft, Civil Action No. 94-1564 (SS), Final Judgment, July 15, 1994.)) The government believed that the settlement would end Microsoft’s monopoly in the operating system market, while Microsoft believed that its business would continue forward with minimal changes.((Elizabeth Corcoran, Microsoft Deal Came Down to a Phone Call, Washington Post, July 18, 1994, pA1.))
Adjacency matters as it is quite difficult to attack a dominant firm in its home market, though even that might be possible if we have a leading firm in one market entering the market of another dominant firm.
I think the history on that one is fairly clear. The government seemingly envisioned that new operating systems would take root if Microsoft altered its licensing practices. This wasn’t a bizarre idea—again IBM released its 1981 PC with three operating systems—but successful direct attacks on a dominant firm in its core market are rare. Successful attacks might be made at points of transition in a market—think the competition in phone OSs with the rise of new touchscreen devices like the iPhone—or in adjacent markets in which the position in the core market is less important—more on that below—but nothing suggests that the government had that idea in mind in settling in 1994.
In 1997, the US government brought an action against Microsoft claiming that Microsoft had breached the terms of the 1994 consent decree by requiring computer makers who wanted Windows 95 to preinstall the current version of Internet Explorer, then IE 3.0.((US v. Microsoft, Petition by the United States for an Order to Show Cause Why Respondent Microsoft Corp. Should Not Be Found in Civil Contempt, Oct. 20, 1997.)) By this point, the government had a clear theory of what Microsoft was doing—Microsoft was acting “to thwart this incipient competition and thereby protect its operating system monopoly”—but the legal issue just turned on what the 1994 consent decree said or didn’t say and Microsoft would eventually win 2-1 in a federal appeals court in late June 1998.((United States v. Microsoft Corp., 147 F.3d 935 (D.C. Cir. 1998).))
But perhaps recognizing the weakness of its position on the language of the consent decree, on May 18, 1998, the US government filed an entirely new antitrust lawsuit against Microsoft. The government could easily have quit at this point. It seems unlikely that the 1994 licensing case was seen within the government as successful. The government might not have known that it would lose the contempt case in June 1998, but after seven years of chasing Microsoft, the government didn’t have much to show for its efforts. Of course, the Antitrust Division was used to the long haul: the IBM mainframe case started in January 1969, only to be dismissed by the government in 1982.
The new case focused on Microsoft’s response to the emergence of the Internet and in particular the competitive threat posed by Netscape Navigator. Again, the government saw Microsoft as trying to protect its position in operating systems and also attempting to extend its monopoly into the new browser market. Bill Gates had outlined the threat that Netscape posed in his famous Internet Tidal Wave memo of May 26, 1995, and Microsoft had responded aggressively to the upstart—or at least so the government’s complaint suggested.
This was clearly a critical juncture in the platform defined by the desktop operating system. Of course, in some ways, Microsoft had stumbled into its monopoly in operating systems. IBM had gone to Microsoft in the first place because of the hard, smart work that Gates and Allen had done in building BASIC for the Altair 8800, but faced with destiny, Microsoft had sent IBM elsewhere. It was only when IBM came back and basically insisted that Microsoft figured out how to cobble together MS-DOS.
But Microsoft had succeeded in the face of the original OS competition and had even successfully navigated from MS-DOS to Windows while IBM had tried to wrest back control of the PC platform with OS2 as OSs moved from characters to graphical user interfaces (GUIs). But the Internet was clearly the future and Netscape Navigator combined with Sun Java—the middleware threat—looked like it might be the next step in computer operating systems.
This was in many ways the best case for competition in platform markets. Netscape’s market share in the browser market had roared ahead as the browser offered an entirely new function. Microsoft’s OS monopoly just wasn’t in the way of Netscape’s organic growth. At the same time, with a strong position established in the new market, Netscape might have been able to fold in new functions into Navigator and diminish the importance of OSs generally. That was the threat that Gates had identified in his Internet Tidal Wave memo.
The government won its case in the federal district court and the core theory of its case was upheld on appeal even as the appellate court cut back on some of the government’s theories.((United States v. Microsoft, 253 F.3d 34 (D.C. Cir. 2001) (en banc).)) After a decade of pursuing Microsoft, the government was finally vindicated. The theory of the case had changed a little moving from the FTC’s 1991 theory on how Microsoft was using its OS monopoly to distort competition in adjacent markets to a theory instead of how Microsoft was using its OS monopoly to protect that monopoly, but it was still a substantial accomplishment.
What was the appropriate remedy? If the concern was that Microsoft had squelched a young possible OS competitor, how to restore that competition? That would seem like the natural goal of an antitrust remedy. The district court had called for Microsoft to be separated into two companies, one focused on the operating system and the other focused on applications like Microsoft Office. Each would have received a copy of Internet Explorer to distribute, though only one of the new firms would have been allowed to develop it further.
This would have been a bold remedy—Microsoft argued, probably correctly, that in the US at least no unitary company had ever been cleaved in two as an antitrust remedy—but there was a more basic conceptual problem. Had the remedy been put in place before the case it seems unlikely that it would have prevented the illegal behavior. MicrosoftOS Co. would have seen the same threat from Netscape Navigator and would have replicated the behavior of the real Microsoft. The divestiture remedy was rejected on appeal and a series of behavioral limits were put in place to limit Microsoft’s ability to engage in similar behavior going forward.
It seems clear that the remedy did not restore Netscape to the market position it would have been in had Microsoft’s illegal behavior not occurred. The natural question is what more aggressive remedies might have accomplished and we can gain some purchase on that question by switching to Europe. At the end of August 2001, the European Commission announced that it believed that Microsoft had impermissibly tied Windows Media Player (WMP) to Windows. The concern here was not really that this was an effort to protect Microsoft’s dominant position in operating systems but rather that Microsoft would gain a decisive advantage in the adjacent media markets.((European Commission, Commission initiates additional proceedings against Microsoft, IP/01/1232, 30 Aug 2001.))
Given the theory of the Windows Media Player case in Europe, the fact that the remedy didn’t change the distribution of WMP should have suggested that Microsoft should have been able to extend its OS monopoly into the media player space. The fact that that didn’t happen … suggests the important ways in which even struggling firms … can compete outside the scope of dominant incumbents.
After a three-year investigation, the European Commission concluded that Microsoft had indeed violated European competition law. Microsoft was fined €497 million and ordered to offer to computer makers two versions of Windows, one with WMP and one without it. Microsoft didn’t have to charge a different price for the two OSs but giving PC makers a choice would ensure that other media player firms could bid to have their media players distributed instead of WMP. Think of this as a subtraction remedy, as Microsoft was required to create a version of Windows with reduced functionality.
What happened? In April 2006, Microsoft reported on how the market had embraced the new option. Over the relevant period, roughly 35.5 million copies of the full-blown version of Windows XP were sold in Europe. And the version without WMP? The total was 1,787 copies—or roughly 0.005 percent of all sales.((Microsoft News Center, Fact Sheet: Windows XP N Sales, April 2006.)) That doesn’t tell us whether there were financial payments made to OEMs, as Microsoft may have been forced to buy distribution of WMP from PC makers and those transfers would be important, but the actual distribution of WMP wasn’t altered by the subtraction remedy.
Perhaps we should have forced Microsoft to distribute Netscape Navigator as a remedy in the US browser case. Actually, we tried a version of that in Europe. Think of this as an addition remedy or a must-carry remedy. In January 2009, the European Commission set out its preliminary conclusion that Microsoft was impermissibly tying Internet Explorer to Windows. Rather than fight that case and possibly disrupt the release of Windows 7 in Europe, Microsoft settled.
In the settlement, Microsoft agreed to distribute something called the browser-choice screen in which a user turning on Windows for the first time in Europe would be presented with a screen of five different browser choices rather than just Microsoft’s Internet Explorer. Actually, a careful user would have noted that the screen offered the chance to scroll horizontally and that 14 different browsers were presented. It was subsequently discovered that Microsoft had broken the browser-choice window when it issued the first service pack update for Windows 7 and yet somehow that fact went undiscovered for 17 months.
This is not a pretty picture and it is important to see the full implications. The US government started chasing Microsoft a decade after the IBM PC’s 1981 release. Much of that was wheel spinning but the government moved successfully against Microsoft at a key competitive juncture when Netscape posed a possible threat to Microsoft’s OS position. But the remedy didn’t restore that threat and the subsequent parallel actions in Europe suggest that a broad set of available remedies might not have worked.
Two final points here. Given the theory of the Windows Media Player case in Europe, the fact that the remedy didn’t change the distribution of WMP should have suggested that Microsoft should have been able to extend its OS monopoly into the media player space. The fact that that didn’t happen—that everyone had Apple iPods and not Microsoft Zunes—suggests the important ways in which even struggling firms—and Apple Computer was that before it morphed into the Apple we know today—can compete outside the scope of dominant incumbents. Apple’s new MP3 player gave it a strong market position outside the dominance of Windows.
The second point of course is that Microsoft’s relative position has eroded mainly because what was once central, the desktop computer world defined by the IBM PC standard, has now been subsumed into a multi-device world defined by the Internet. The actual remedy implemented in the US case is seen as having slowed down Microsoft, making it less aggressive, less nimble, and more lawyer-hobbled. And that may have mattered when the next threat emerged.
Google and Antitrust Neutrality Remedies
Sergey Brin and Larry Page described their Google prototype in a 1998 paper that they prepared for an academic conference in Brisbane, Australia, on the World Wide Web.((S. Brin & L. Page, The Anatomy of a Large-Scale Hypertextual Web Search Engine (online here).)) Their new search engine would be launched into a crowded field of search engines—Alta Vista, Lycos, Excite, and more if you remember your Internet history—and yet it would come quickly to dominate the search market. The vision behind the Google prototype and what would become the Pagerank patent and algorithm was better use of hyperlinks as a signal of website value. Google’s search algorithm has evolved over time to augment that original key insight with the ability to evaluate value based upon the behavior of searchers in clicking and not clicking on organic search results—the search version of the wisdom of crowds.
But as the 1998 Brin and Page paper made clear, there was a fundamental contradiction at the core of building an advertising-supported search engine. Consumers who would click on high-quality organic search results would have little reason to engage with advertising. Indeed, if the search engine was going to try to get consumers to click on ads, it might have an incentive to degrade the quality of the organic search results. Plus bias for or against particular sites would be very difficult to detect.
It was exactly that concern about bias that led to antitrust investigations against Google in both the US and the EC. The FTC action stalled out, but the EC investigation eventually focused on exactly these issues of bias. On June 27, 2017, the EC announced a fine of €2.42 billion against Google relating to Google’s shopping product where the central finding was that Google had preferred its own shopping site to those of its competitors.((European Commission Press Release, Antitrust: Commission fines Google €2.42 billion for abusing dominance as search engine by giving illegal advantage to own comparison shopping service – Factsheet, 27 June 2017 (online here).)) The EC’s decision in the Google shopping case is complex—215 pages single-spaced—but the core of the remedy is to “subject Google’s own comparison shopping service to the same underlying processes and methods for the positioning and display in Google’s general search results pages as those used for competing comparison shopping services.”((European Commission, Google Search (Shopping) Case AT.39740, redacted public decision of Dec. 18, 2017, ¶ 700.)) Comparison shopping service neutrality as it were.
Even when officials have a good theory of liability, they have struggled to come up with successful remedies. Direct regulatory remedies have been ineffective in the examples seen in this paper, while remedies directed at firm culture or internal transaction costs have been perhaps more effective.
Nondiscrimination duties are quite traditional in regulated industries and common antitrust remedies. Google started implementing its interpretation of the remedy on September 27, 2017. Google is continuing to implement a version of its product shopping unit as an ad on the top of particular search result pages but now outside comparison shopping services can bid against Google for the product slots as part of a standard Google auction.((Google AdWords Blog, Changes to Google Shopping in Europe, Sept. 27, 2017.))
Google’s approach to advertising has evolved over time. Google started with traditional impression-based advertising before switching to pay-per-click style advertising. The move to the rich, product slot ads backed by detailed up-to-the-minute product inventories represents something meaningfully different from Google’s traditional search market, which is based on public information available on the Internet. The EC understandably concluded that this was a separate market—an adjacent market as it were—and acted to attempt to preserve competition in that market. The EC will monitor the remedy and Google’s competitors are already complaining about it.((FairSearch.org, Open letter to Vestager: Google remedies fail to comply with decision, Feb. 28, 2018.))
Platform markets are often punctuated equilibria markets: short periods of competition followed by a market characterized by dominance, with that cycle to repeat. Schumpeterian competition. The point at which a new cycle might start is critical to healthy competition and the competitive spark to that cycle will often arise in a market near or even adjacent to the market current subject to dominance. Antitrust officials therefore have good reason to act to try to protect that nascent competition in these adjacent markets, as they did repeatedly for Microsoft and have commenced doing for Google. But even when officials have a good theory of liability, they have struggled to come up with successful remedies. Direct regulatory remedies have been ineffective in the examples seen in this paper, while remedies directed at firm culture or internal transaction costs have been perhaps more effective.
Randal C. Picker is the James Parker Hall Distinguished Service Professor of Law at the University of Chicago Law School and Senior Fellow at the Computation Institute of the University of Chicago and Argonne National Laboratory.
Disclaimer: The ProMarket blog is dedicated to discussing how competition tends to be subverted by special interests. The posts represent the opinions of their writers, not necessarily those of the University of Chicago, the Booth School of Business, or its faculty. For more information, please visit ProMarket Blog Policy.