quality two-sided-markets final

Research Paper #268 May 2015

When Competition Fails to Optimise Quality: A Look at Search Engines

Maurice E. Stucke &

Ariel Ezrachi

Do not cite without authors’ permission

Copyright © 2015

This paper may be downloaded without charge

from the Social Science Research Network Electronic library at http://ssrn.com/abstract=2598128

Learn more about the University of Tennessee College of Law: law.utk.edu

http://ssrn.com/abstract=2591874

http://law.utk.edu/

Electronic copy available at: http://ssrn.com/abstract=2598128

- Working Paper -

WHEN COMPETITION FAILS TO OPTIMISE QUALITY: A LOOK AT SEARCH ENGINES

Maurice E. Stucke* & Ariel Ezrachi**

The European Commission’s Statement of Objections forms the latest addition to the ongoing debate on the possible misuse of Google’s position in the search engine market. The scholarly debate, however, has largely been over the exclusionary effects of search degradation. Less attention has been attributed to the dimension of quality – whether and how a search engine, faced with rivals, could degrade quality on the free side. We set out to address this fundamental issue: With the proliferation of numerous web search engines and their free usage and availability, could any search engine degrade quality? We begin our analysis with a review of the network effects that may impact the relative power of a search engine. We next identify three necessary, but not sufficient, variables for quality degradation to occur in search results. With these three variables in mind, we consider instances when a search engine could degrade quality despite competition from rival engines.

INTRODUCTION

The exponential growth of the Internet has seen a proliferation of email

platforms, social networks,1 texting, mapping, video sharing, gaming and

online communications,2 many of which are provided free of charge. The

common business model in these two-sided markets is based on the

potential income firms may generate from utilising their customer base.

Most noticeably, firms offer consumers a free product or service and earn

* Professor, University of Tennessee College of Law; Co-founder, Data Competition

Institute. ** Slaughter and May Professor of Competition Law, The University of Oxford.

Director, Oxford University Centre for Competition Law and Policy. 1 For example: Linkedin, Facebook, Academia. 2 See Case COMP/M.7217—Facebook/WhatsApp, Comm’n Decision, 2014 O.J. (C

7239) 24–25, ¶ 47 (noting that the “vast majority of social networking services are provided free of monetary charges”).

Electronic copy available at: http://ssrn.com/abstract=2598128

2 QUALITY, COMPETITION & TWO-SIDED MARKETS 23 April 2015

their income from in turn selling to advertisers the ability to access these

consumers.

By and large, when a product or service is offered for free, the primary

dimension of competition is typically quality.3 A competitive market

environment is therefore likely to stimulate investment in quality. Yet,

when a firm mainly earns its profits from one side of the market, such as

advertising, its incentive to invest in quality on the other side of the market

may be distorted. In such instances, it may have an incentive to degrade

quality for consumers below desirable levels on the free side of the market,

if doing so increases its profitability (or market power).

Examples of quality degradation in paid and free two-sided markets are

easy to find. These include, among others, newspapers, which may

compete vigorously but skew their news coverage through self-censorship

to avoid offending an important category of advertisers.4 Commercial radio

stations can skew playlists to music companies that pay them for airing their

songs.5 Entertainment producers may develop the show’s story line to

highlight a sponsor’s product at the expense of artistic quality.6 In those

instances, and others, the customer is not just the recipient of a product or

service, but also constitutes a commodity traded on the other side of the

3 European Commission Case No. Comp/M. 6281—Microsoft/Skype, Regulation (EC)

No. 139/2004 Merger Procedure (Oct. 7, 2011); European Commission Case No. Comp/M. 5727—Microsoft/Yahoo! Search Business Regulation (EC) No. 139/2004 Merger Procedure (Feb. 18, 2010); Office of Fair Trading, Completed Acquisition by Motorola Mobility (Google, Inc.) of Waze Mobile Ltd., ME/6167/13 at ¶ 28 (Dec. 17, 2013) [hereinafter Google/Waze] (OFT considering whether merger “may dampen Google’s incentive to innovate and improve quality as a result of the loss of an innovative rival”).

4 Allen P. Grunes & Maurice E. Stucke, Plurality of Political Opinion and the Concentration of the Media, in GENERAL REPORTS OF THE XVIIITH CONGRESS OF THE INTERNATIONAL ACADEMY OF COMPARATIVE LAW 575-76 (Karen B. Brown & David V. Snyder eds., Springer 2012).

5 After a series of scandals where music companies paid radio stations to play certain songs, the U.S. Federal Communications Commission promulgated “payola” rules where the broadcaster must disclose such payments. http://www.fcc.gov/guides/payola-rules.

6 See for example: http://www.ukessays.com/dissertations/media/the-effect-of-product-placement-in-movies-media-essay.php

23-Apr-15 QUALITY, COMPETITION & TWO-SIDED MARKETS 3

market. When the product or service is provided free of charge, the value

gained on the other side of the market becomes the focal point.

Consequently, in two-sided markets, producers may degrade the quality

of the free product along certain dimensions where doing so maximises

profit on the other side of the market. The extent to which quality may be

degraded depends on the level of competition and the ability of consumers

to identify and appraise changes in quality.7

In this paper we explore the interface between competition and quality

in the two-sided market of online search engines. This area is of

significance as search engines have evolved to become the gateway and

guide which link users and relevant websites.8 The competition between

these search engines focuses on quality – on their capacity to quickly

deliver the most relevant results in response to a search query. However,

the characteristics of two-sided markets may result in distortion of that

competition due to the impact of other revenue generating activities.

Search degradation and its impact on competition have been at the

centre of attention in recent years. In April 2015 the European Commission

issued a Statement of Objections concerning Google’s preferential

treatment in its general search results of its own comparison shopping

service. In its statement, the Commission outlined its preliminary view that

7 Ariel Ezrachi & Maurice E. Stucke, The Curious Case of Competition and Quality,

JOURNAL OF ANTITRUST ENFORCEMENT (forthcoming, October 2015), University of Tennessee Legal Studies Research Paper No. 256; Oxford Legal Studies Research Paper No. 64/2014, available at SSRN: http://ssrn.com/abstract=2494656 or http://dx.doi.org/10.2139/ssrn.2494656.

8 Ioannis Lianos & Evgenia Motchenkova, Market Dominance and Search Quality in the Search Engine Market, 9 J. COMPETITION LAW & ECONOMICS 419, 422 (Advance Access publication 17 April 2013) (discussing how search engines “act as ‘information gatekeepers’: they not only provide information on what can be found on the web (equivalent to yellow pages), but they also are ‘an essential first-point-of-call for anyone venturing onto the Internet’” and how they differ from other two-sided platforms, as “search engines detain an important amount of information about their customers and advertisers (the ‘map of commerce’)”).


this practice amounts to an abuse of a dominant position. In her statement

on this case, Commissioner Vestager noted:

when a consumer enters a shopping-related query in Google's search engine, Google's comparison shopping product is systematically displayed prominently at the top of the search results. This display is irrespective of whether it is the most relevant response to the query. Thus, Google's commercial product is not subject to the same algorithms as other comparison shopping service … with the result that consumers may not necessarily see the most relevant results in response to their queries, and Google's competitors may not get the commercial opportunities that their innovations deserve.9

The Commission’s Statement of Objections is predominantly concerned

with the leveraging of Google’s market power in the online general search

engine market to create an advantage in the related market of comparison

shopping services.

The issuing of the Statement of Objections formed the latest addition to

the ongoing debate on the possible misuse of Google’s position in the

search engine market. That debate has focused on possible manipulation of

search engine results, the impact this may have on actual or potential

competitors operating downstream, and the effects on the competitive

process.10 The scholarly debate has largely been over whether search

degradation violates the competition laws. Less attention has been

9 http://europa.eu/rapid/press-release_STATEMENT-15-4785_en.htm. 10 Andrea Amelio & Dimitrios Magos, Economic Background of the Microsoft/Yahoo!

Case, COMPETITION POLICY NEWSLETTER, Number 2 — 2010, at 51; Statement of the Federal Trade Commission Regarding Google’s Search Practices, In the Matter of Google Inc., FTC File Number 111-0163 (Jan. 3, 2013), http://www.ftc.gov/system/files/documents/public_statements/295971/130103googlesearchstmtofcomm.pdf. On the EU Commission’s proceedings which were initiated in November 2010, see http://ec.europa.eu/competition/elojade/isef/case_details.cfm?proc_code=1_39740.


attributed to the dimension of quality – whether and how a search engine,

faced with rivals, could degrade quality on the free side.11

In responding to the Commission’s ongoing concerns, Google stated

that while its search engine “may be the most used . . . people can now find

and access information in numerous different ways—and allegations of

harm, for consumers and competitors, have proved to be wide of the

mark.”12 Google also stated that “people have more choice than ever

before” citing “numerous other search engines such as Bing, Yahoo, Quora,

DuckDuckGo and a new wave of search assistants like Apple’s Siri and

Microsoft’s Cortana,” in addition to “a ton of specialized services like

Amazon, Idealo, Le Guide, Expedia or eBay” and “social sites like

Facebook, Pinterest and Twitter.”13

In what follows we set out to address this fundamental issue: to what

extent will competition from smaller search engines prevent larger search

engines from degrading the quality of search results (and vice versa)? With

the proliferation of numerous web search engines and their free usage and

availability, could any search engine degrade quality? After all, Internet

browsers enable users to run their queries on different search engines with

access to multitude of services, which are a click away.

We begin our analysis with a review of positive feedback loops and

network effects that may impact the relative power of a search engine. In

Part II we consider how these network effects can impact a search engine’s

11 Lianos & Motchenkova, supra note, at 424 n. 18 & 425 (noting that most of the

existing literature focuses on the advertising side of search engines, finding “a threat of reduction in the quality of search results, if the search engine market is monopolized or dominated by a single firm” and suggesting that their model be extended to “an oligopoly setting” as a “natural step forward” since it “is a better description of the current practice with several engines competing in the search market (such as Google, Bing, and Yahoo, among others)”).

12 The Search for Harm, Google Europe Blog, April 15, 2015, http://googlepolicyeurope.blogspot.co.uk.

13 Id.


incentive and ability to degrade quality. Part II identifies three necessary,

but not sufficient, variables for quality degradation to occur in search

results.14 The first relates to the search engine’s incentive and ability to

degrade quality on the free side of the market (namely the search results).

This, in turn, depends on the degree of several network effects, and the

extent to which these benefit from scale and scope. The second variable is

consumers’ ability and incentive to accurately assess quality differences.

The third variable concerns imperfect information flows that make it

difficult or costly to convey to consumers the search engines’ inherent

quality differences. With these three variables in mind, we consider

instances when a search engine could degrade quality despite competition

from rival engines.

I. NETWORK EFFECTS AND THE INCENTIVE TO DEGRADE QUALITY

Search engines provide the gateway to Internet content, a crucial

mapping function without which the vast digitalised landscape would

become largely inaccessible.15 The major search engines do not charge

14 See Ezrachi & Stucke, supra note (discussing how competition authorities have

struggled in identifying the dimensions of quality competition important to many consumers).

15 This appears to remain true in the mobile world. As one recent report found: • The total U.S. multi-platform web search market grew 5 percent in

query volume in Q4 2014 vs. the previous year. Mobile search, which includes queries conducted via app and mobile browser, now accounts for 29 percent of all search activity, with smartphones driving a greater share (20 percent) than tablets (9 percent).

• As consumers shift their digital activity to mobile, growth in the search market is being driven by both smartphones (up 17 percent from the prior year) and tablets (up 28 percent). Desktop search, meanwhile, has declined marginally during the same period.

• Google remains the leader in the U.S. explicit core search market with 66 percent market share of search queries conducted in Q4 2014, followed by Bing at 20 percent and Yahoo at 11 percent. Bing increased its market share in 2014, while Yahoo’s recent search partnership with Firefox has also bolstered its share. In terms of


users for searching the web. Search engines therefore compete for users on

quality – providing search results that are quicker and more relevant.16

Their commercial model relies on their intermediary role, connecting

“between content providers (who want users), users (who want content),

and advertisers (who want users).”17 As the FTC Staff Report found,

“Although a user does not pay for the web search service, the user's focused

interest - or intent - is very valuable to advertisers, because users arc

effectively identifying themselves as potential customers through the

content of their queries.”18 The intermediary role of search engines enables

them to derive their revenue from advertising.19 For example, in 2013, 91

multi-platform search activity measured by comScore, Google’s strong leadership on both smartphones and tablets boosts its share of the multi-platform search market by several percentage points vs. desktop alone.

COMSCORE, 2015 U.S. DIGITAL FUTURE IN FOCUS 16 (2015). 16 Microsoft/Yahoo, supra note, at ¶ 101 (competition takes place of the quality of the

search results (i.e., their relevance and speed) and the user interface); Teresa Vecchi, Jerome Vidal & Viveca Fallenius, The Microsoft/Yahoo! Search Business Case, 2 EUR. COMM’N COMPETITION POL’Y NEWSL. 41, 46 (2010) (EC finding that the “quality and relevance of the algorithmic search engine” as “the most important factor in attracting users to a particular search engine”).

17 Lianos & Motchenkova, supra note, at 421. 18 Report from the FTC Bureau of Competition Staff to the Commission re Google

Inc., at 8. (Aug. 8, 2012), available at http://graphics.wsj.com/google-ftc-report/ [hereinafter FTC Staff Report]. A few caveats about portions of this report, which the FTC released (mistakenly) under the Freedom of Information Act to the Wall Street Journal. First, only the even pages were released, so the missing odd pages may have contained important qualifications. Second, other reports, including any prepared by the FTC economists and Google, were not released. Third, although the staff recommended the FTC to file a complaint, the Commissioners elected not to. Google said in response to the report’s disclosure, “We understand that what was sent to the Wall Street Journal represents 50% of one document written by 50% of the FTC case teams. Ultimately both case teams (100%) concluded that no action was needed on search display and ranking. Speculation about consumer or competitor harm turned out to be entirely wrong. On the other issues raised, we quickly made changes as agreed with the FTC.” http://graphics.wsj.com/google-ftc-report/

19 Microsoft/Yahoo, supra note, at ¶ 33.


per cent of Google’s revenues came from advertisers,20 and 79 per cent of

Yahoo!’s total revenue came from display and search advertising.21

The search engines provide users with “sponsored” results and

“organic” results, which are produced by the search engine’s algorithms.22

For the paid “sponsored” search results, most advertisers pay the search

engine on a cost-per-click basis, whereby the advertiser pays the search

engine only when a user clicks on its sponsored ad.23 Since they are paid

only when a user clicks on an ad, search engines generate “revenues

primarily by delivering relevant, cost-effective online advertising.”24

Indeed, the goal of Google’s AdWords, its main auction-based advertising

program, “is to deliver ads that are so useful and relevant to search queries

or web content that they are a form of information in their own right.”25

In this section we consider positive feedback loops and network

effects26 on a search engine’s incentive and ability to degrade quality.

20 Google 2013 10-K, supra note, at 9. 21 Yahoo 2013 10-K, supra note, at 13. 22 Microsoft/Yahoo, supra note, at ¶ 100. 23 Microsoft/Yahoo, supra note, at ¶¶ 35, 45; Amelio & Magos, supra note, at 50

(internal footnotes omitted): The expected revenue from an ad is contingent on the probability that an ad is clicked (measured by the likelihood that users click on ads, also known as the Click Through Rate (‘CTR’)) since advertisers pay platforms only when users click on the displayed ads. Search platforms use a ‘quality’ score, that reflects the expected CTR, to adjust the ranking accordingly. Google was the first to introduce the idea of ranking the ads in 2002 by weighting the advertisers’ bids with the ‘quality score’. As explained on its web site, Google currently uses a variety of indicators that try to measure quality and determine the quality score of an advertiser.

24 Google 2013 10-K, supra note, at 3; Microsoft/Yahoo, supra note, at ¶ 101. 25 Google 2013 10-K, supra note, at 4. 26 Network effects were at play in the Microsoft case:

In markets characterized by network effects, one product or standard tends towards dominance, because “the utility that a user derives from consumption of the good increases with the number of other agents consuming the good.” Michael L. Katz & Carl Shapiro, Network Externalities, Competition, and Compatibility, 75 AM. ECON. REV. 424, 424 (1985). For example, “[a]n individual consumer’s demand to use


A. Trial-and-error

The first potential network effect is linked to the scale of search

inquiries processed by the search engine. Trial-and-error, or learning by

doing, means that more searches increase the search engine’s likelihood of

identifying relevant results. A search engine cannot read the consumer’s

mind. It does not know when the consumer types “apple” and “orange,”

whether she is searching for fruit or technology companies. When a

consumer types “orange,” and “apple,” the search engine quickly generates

an opinion as to what information users will find most useful.27 As

Google’s Executive Chairman testified:

When a consumer enters search terms, those terms are processed by the search engine’s mathematical algorithms, which determine the probability that any given webpage will be responsive to the search. The user then receives results that are rank-ordered based on the search engine’s judgment of the likelihood that each result matches what the user was seeking in entering the search terms. This process necessarily depends on multiple variables and constant refinement.28

(and hence her benefit from) the telephone network TTT increases with the number of other users on the network whom she can call or from whom she can receive calls.” Howard A. Shelanski & J. Gregory Sidak, Antitrust Divestiture in Network Industries, 68 U. CHI. L. REV. 1, 8 (2001). Once a product or standard achieves wide acceptance, it becomes more or less entrenched. Competition in such industries is “for the field” rather than “within the field.” See Harold Demsetz, Why Regulate Utilities?, 11 J.L. & ECON. 55, 57 & n.7 (1968) (emphasis omitted).

United States v. Microsoft Corp., 253 F.3d 34, 49 (D.C. Cir. 2001). 27 Dissenting Statement of Commissioner Pamela Jones Harbour, In the Matter of

Google/DoubleClick, F.T.C. File No. 071-0170, at 7 (“Type the search term ‘apple’ into the Google search engine, and Google will ‘know’ whether the user is focusing on food (apple recipes) or technology products (Apple computers), depending on which websites the user recently visited (Cooking Light versus MacWorld) as well as what searches she recently conducted (Golden Delicious versus iPod).”).

28 The Power of Google: Serving Consumers or Threatening Competition?: Hearing before the Subcomm. on Antitrust, Competition Policy and Consumer Rights U.S. Senate


The search engine first has the benefit of observing which links, if any,

its users actually choose. If many choose a link that was originally offered

down the list (say on the third or fourth page of results), the search engine’s

algorithms can harvest that information to move that link up the list; the

search engine demotes less frequently tapped suggested links down the

list.29 Thus the more consumers using the search engine and the more

searches they run, the more trials the search engine has in predicting

consumer preferences, the more feedback the search engine receives of any

errors, and the quicker the search engine can respond with recalibrating its

offerings.30 Increased traffic volumes make more experiments possible,

thereby improving search results.31

Naturally, the product improvement attracts additional consumers to that

search engine compared to competitor sites. In effect, the more users, the

larger (and more heterogeneous) the sample size, and the better the search

Judiciary Committee, 112th Cong. (Sept. 21, 2011), at 2 (Testimony of Eric Schmidt, Executive Officer, Google Inc.).

29 FTC Staff Report, supra note, at 14, quoting Google's former chief of search quality Udi Manber:

The ranking itself is affected by the click data. If we discover that, for a particular query, hypothetically, 80 percent of people click on Result No. 2 and only l0 percent click on Result No. I, after a while we figure out, well, probably Result 2 is the one people want. So we'll switch it.

Other Google executives confirmed that “click data is important for many purposes, including, most importantly, providing ‘feedback’ on whether Google’s search algorithms are offering its users high quality results.” Id.

30 Steve Lohr, Can These Guys Make You ‘Bing’?, N.Y. TIMES, July 31, 2011, at 3 (“Consumer testing is key to the algorithm refining process, and Google uses both human reviewers and samples of real search traffic in order to measure whether a proposed algorithm change improves the user experience or not.”); Press Release, U.S. Dep’t of Justice, Statement of the Department of Justice Antitrust Division on Its Decision to Close Its Investigation of the Internet Search and Paid Search Advertising Agreement Between Microsoft Corporation and Yahoo! Inc. (Feb. 18, 2010), available at http://www.justice.gov/atr/public/press_releases/2010/255377.pdf.

31 FTC Staff Report, supra note, at 14 (stating “[t]he more search users there are at any given time, the more experiments can be run, the faster they can be completed, and the more improvements that can be made to the search algorithms” and how “search providers run experiments on large volumes of users”); Vecchi et al., supra note, at 44 & 46.


engine can identify relevant responses for both popular and less frequent

queries (“tail” queries).32

Entry barriers into the search engine market are high.33 Microsoft

reportedly invested in 2010 “more than $4.5 billion into developing its

algorithms and building the physical capacity necessary to operate Bing.”34

A new entrant can hire tech talent, but it would still lack the scale of this

trial-and-error experimentation.35 Microsoft argued, and the European

Commission found, that “scale is an important element to be an effective

competitor.”36 With fewer trials, entrants have fewer opportunities to

predict search terms (or what the consumer wants to know). Entrants have

fewer opportunities to observe subsequent errors and to perceive trends

(consumers’ search terms relating to a hot topic). Their ability to identify

sites that consumers prefer is likely to remain inferior, so the entrant

remains at a competitive disadvantage in attracting consumers and

advertisers.37 Recognising this, a smaller search engine can specialise in

32 FTC Staff Report, supra note, at 14; Microsoft/Yahoo, supra note, at ¶ 162 (noting

Microsoft’s claim). 33 FTC Staff Report, supra note, at 76 (“Along with specialized algorithms, search and

search advertising platforms require enormous investments in the technology and infrastructure required to crawl and categorize the entire Internet.”).

34 FTC Staff Report, supra note, at 76. 35 FTC Staff Report, supra note, at 76. 36 Microsoft/Yahoo, supra note, at ¶ 153; see also FTC Staff Report, supra note, at 76

(noting that Internet search, search advertising, and search syndication are “markets characterized by substantial scale effects”).

37 Pamela Jones Harbour & Tara Isa Koslov, Section 2 in a Web 2.0 World: An Expanded Vision of Relevant Product Markets, 76 ANTITRUST L.J. 769, 784 (2010) (“Google search engine has become further entrenched as the dominant search site, and the firm has accumulated even more search data. Given the role of network effects, one might wonder whether any other firm will be able to chip away at Google’s search supremacy without access to a comparable trove of data.”) (internal footnotes omitted); EUR. DATA PROT. SUPERVISOR, PRIVACY AND COMPETITIVENESS IN THE AGE OF BIG DATA: THE INTERPLAY BETWEEN DATA PROTECTION, COMPETITION LAW AND CONSUMER PROTECTION IN THE DIGITAL ECONOMY 35 (preliminary opinion Mar. 2014) [hereinafter EDPS Preliminary Opinion] (“Successful online providers persuade increasing numbers of customers to provide more personal information which increases the value of the service to


specific functions, such as travel and flight options on travel-specific

Internet sites, like Kayak and Expedia. However, one downside in

becoming a niche player is that the search engine can lose an important

segment of the population.38

A narrow user group would lead to the search results being skewed to

their preferences, which may differ from those of other groups.

Accordingly, an older user group may through usage make the search

engine less attractive to younger user groups. A collaboration between

medium size engines may be used to widen the search group. Indeed,

Microsoft justified its Yahoo partnership as necessary to achieve greater

scale of behavioural trial-and-error learning.39 Before the partnership,

fewer people used Microsoft’s and Yahoo’s search engines compared to

Google. As Microsoft’s then CEO said, “it turns out there’s a feedback

loop in the search business, where the most searches you serve, or paid ad

searches you serve, the more you learn about what people click on, what’s

relevant, and it turns out that scale drives knowledge which then can turn

around and redrive innovation and relevance.”40

Interestingly, increased market usage and share correlates with

increased quality. Each user’s utility from using the search engine increases

advertisers, thus generating ‘network effects’ whereby yet more customers are attracted to the service.”).

38 Yusuf Mehdi, Senior Vice President, Online Audience Business, Remarks at the Credit Suisse Annual Technology Conference (Dec. 1, 2009).

39 In December 2009, Microsoft partnered with Yahoo! to provide the exclusive algorithmic and paid search platform for the Yahoo! web sites. Microsoft believed this agreement would allow it over time to improve the effectiveness and increase the value of its “search offering through greater scale in search queries and an expanded and more competitive search and advertising marketplace.” Microsoft Corp., Annual Report for the Fiscal Year Ended June 30, 2011 (SEC Form 10-K), at 6 (filed July 28, 2011).

40 Transcript from Remarks from the Conference Call Held by Steve Ballmer, Chief Executive Officer, Microsoft, and Carol Bartz, Chief Executive Officer, Yahoo!, to Announce the Search Engine Agreement Between Yahoo! and Microsoft, Microsoft News Center (July 29, 2009).


as others use it as well. As more people use the search engine, the more

trial-and-error experiments, the more likely the search engine can learn of

consumer preferences, the more relevant the search results will likely be,

which in turn will likely attract others to use the search engine, and the

positive feedback continues.

B. Scope of Data

A second potential network effect involves the scope of data on the user.

Search results can improve from the variety of personal data on users. If

people use, besides the search engine, other services offered by the

company (such as email, browser, texting, mapping service etc.), the

company, in collecting a variety of personal data, can develop user profiles

to better predict that particular user’s tastes and interests, and better target

the user with more relevant organic and sponsored search results.41 Under

the first network effect, with more search inquiries, the search engine,

through learning-by-doing, can identify the likely universe of relevant

responses. Under the second network effect, the more that users rely on the

platform’s services, the greater the variety of personal data on particular

users, the better the search engine can segment search results by user

profiles, the better the search engines can offer personalised search results.

Suppose the search engine provides an array of services, including

maps. The search engine, in collecting users’ geo-location data, knows

where its users typically walk or drive during the week. When the user

searches for restaurants, the search engine can use the geo-location data to

recommend restaurants in the immediate vicinity, as well as other personal

data to recommend restaurants within the user’s price range (based on 41 Microsoft/Yahoo, supra note, at ¶ 40 (noting that a “growing number of both search

and non-search ads are also behaviourally targeted”).


inferences of where the user lives and shops), and that the user likely would

enjoy (based on what the user’s friends recommend in the user’s social

networks). So the more people that use the company’s platform of services,

the more personal data the company collects in developing user profiles, the

more likely the company can target users with sponsored and organic search

results, and the more opportunities across the platform to target users with

behavioural ads. As the OECD noted, “data linkage enables ‘super-

additive’ insights, leading to increasing ‘returns to scope’. Linked data is a

means to contextualise data and thus a source for insights and value that are

greater than the sum of its isolated parts (data silos)… diverse data set

allows the company to create even more detailed profiles about its users that

were not possible with each single service.”42

So the feedback loop adds another dimension: it is now no longer the

trial-and-error, learning-by-doing from earlier searches, but also learning of

user’s tastes and preferences from the variety of personal data it collects

across its platform (such as the user’s e-mail, geo-location data, social

network, browser history) to personalize search results and target users with

specific sponsored ads that they will likely click (as well as organic search

results).

In the context of our discussion, one can therefore observe a correlation

between loyalty and quality. The more users on the platform and the more

loyal the users are to the platform, the more the search engine may learn

about user preferences and thus improve the search results’ relevance and

quality.

42 OECD, Data-Driven Innovation for Growth And Well-Being: Interim Synthesis

Report 29 (2014) [hereinafter OECD Interim Synthesis Report] (internal footnote omitted).


C. Spill-overs and Snowball Effect

A third network effect concerns the paid side of the market. As the

OECD observed, the network effects on the free side can spill over to the

paid side, and each can reinforce the other: “The reuse of data generates

huge returns to scale and scope which lead to positive feedback loops in

favour of the business on one side of the market, which in turn reinforces

success in the other side(s) of the market.”43 The inflow of many users with

heterogeneous search inquiries will attract a greater variety of advertisers to

the platform.44

This phenomenon is not unique to online search engines and may be

observed in other two-sided markets. For instance, as more people watch

the Superbowl or World Cup, the more advertisers want to reach these

viewers. But search engines, unlike television shows, can engage in more

advanced behavioural advertising. As the company attracts more users on

the free side of its services platform (such as search, email, mapping, etc.),

it can collect a greater volume and variety of personal data to develop user

profiles.45 The search platform can use the inflow of personal data to better

target consumers with specific targeted advertising across its platform of

free services (such as sponsored search results, ads in e-mail, display ads in

43 OECD Interim Synthesis Report, supra note, at 29; see also Vecchi et al., supra

note, at 46 (noting that the search business “is subject to network effects in that scale can improve the quality of the search results and the quality of the matching of the ads with the queries”).

44 Microsoft/Yahoo, supra note, at ¶¶ 157 (all of the advertisers responding to the market investigation highlighted that Google’s query volume was one of the main reasons why Google was a “must have” for search advertising campaigns), 163 (“Higher query volume in turn generate ad inventory.”); Vecchi et al., supra note, at 44.

45 See, e.g., In re Google Inc., No. 13-MD-02430-LHK, 2013 WL 5423918, at *2 (N.D. Cal. Sept. 26, 2013) (noting that although some Google Apps users, whether through the educational program or the partner program, “did not receive content-based ads,” their emails were “nevertheless intercepted to create user profiles”).


videos, etc.) in the moments that matter for a purchasing decision.46 In

targeting users with more relevant ads (or ads that the user will more likely

click), the search engine is more likely to attract advertisers (as consumers

are more likely to click on their ads) and thereby increase its advertising

revenue and profits. Moreover, the search engine can target users with

these personalised ads across media (such as on their personal computers,

smartphones, tablets and, soon, household appliances) and across services

(such as texts, maps, videos, etc.). This too increases the likelihood of

consumers clicking on a relevant sponsored ad (which generates revenue on

a cost-per-click basis) or seeing a display ad (which generates revenue on a

cost-per-impression basis). As more users are drawn to the platform, and as

the company amasses a greater variety of data to effectively target

consumers with relevant online ads, the broad platform can reduce the

advertisers’ fixed costs of managing multiple ad campaigns. In summary,

the network effects on the free side will spill over in attracting more

advertisers on the paid side. As the OFT observed, “[g]enerally, where

there is a two-sided platform, there is more value to both sides of having

more users.”47 As more people use the search engine, the more advertisers

will also use the platform, the more relevant and targeted the

advertisements, the more likely that users will click the ads, the more profits

the search engine has to expand its range of free services and to ensure that

its search remains the default search engine on various portals to the

Internet (for example, developing one’s own browser and paying other

browsers to have one’s search engine be the default). The search engine’s

accessing and quickly analysing this volume and variety of data can yield

46 Google 2013 10-K, supra note, at 25 (“The main focus of our advertising programs

is to help businesses reach people in the moments that matter across all devices with smarter ads that are relevant to their intent and context, reflecting our commitment to constantly improve their overall web experience.”).

47 Google/Waze, supra note, at ¶ 19.


significant value for advertisers.48

Then FTC Commissioner Harbour noted how her fellow commissioners

ignored these network effects when analysing the Google/DoubleClick

merger:

On the search side, Google is able to charge a premium for search advertising because Google has the highest volume of searches. More searches translate to more incoming data, which enables Google to enhance the quality of the underlying algorithms used to process searches and match them to relevant advertisements. Google’s search methodology and advertisement targeting become even better as consumers use Google’s search engine more. Improved searches drive still more traffic to the site, which further increases the value of search advertising on Google.49

Consequently, the above network effects can have a snowball effect,

whereby more users generate more search queries which generate more

48 As Harbour and Koslov state:

In online advertising, an advertiser gets the most “bang for its buck” when its ads generate more “click-throughs”—meaning that a user’s interest is piqued, and the user completes an action, such as clicking on the ad to pursue more information. The point of traditional, non-behavioural display advertising is a numbers game: it seeks to place a given ad in front of as many eyeballs as possible, figuring that if a certain percent-age of viewers will respond to the ad, more eyeballs will equal more click-throughs. Behavioural advertising takes a more targeted approach. It attempts to place highly relevant ads in front of the right sets of eye-balls, to maximise the likelihood of a click-through from each viewer. If the ads are likely to be more effective at attracting customers, an advertiser will pay more to place the ads, which will generate a larger revenue stream for the Web site.

Harbour & Koslov, supra note, at 781; see also Public Citizen, Mission Creep-y: Google Is Quietly Becoming One of the Nation’s Most Powerful Political Forces While Expanding Its Information-Collection Empire 10 (Nov. 2014) (“The more narrowly and accurately Google can target an ad to a user to match her interests, the more it can charge advertisers for each view or click”).

49 Dissenting Statement of Commissioner Pamela Jones Harbour, In the Matter of Google/DoubleClick, F.T.C. File No. 071-0170, at 6 (internal footnotes omitted).


trial-and-error, which yields better search results, which attracts more users

and advertisers to the search platform, which enables better profiling of

users and greater likelihood of users clicking the ads, which generates more

advertising revenue to enable the search engine to offer even more free

services,50 which enables consumers to spend more time on the company’s

platform, which allows it “to gather even more valuable data about

consumer behaviour, and to further improve services, for (new) consumers

as well as advertisers (on both sides of the market).”51 The OECD found

increasing returns to scale from data:

The accumulation of data can lead to significant improvements of data-driven services which in turns can attract more users, leading to even more data that can be collected. This “positive feedback makes the strong get stronger and the weak get weaker, leading to extreme outcomes” (Shapiro and Varian, 1999). For example, the more people use services such as Google Search, or recommendation engines such as that provided by Amazon, or navigation systems such as that provided by TomTom, the better the services as they become more accurate in delivering requested sites and products, and providing traffic information, and the more users it will attract.52

While the spillover and snow-ball effects may enhance the position of the

already successful search engine, to the possible detriment of its

competitors, we note that in themselves, these effects are not likely to

degrade the quality of the search engine’s services. One should also

consider whether, and the extent to which, these network effects, after a

50 OECD Interim Synthesis Report, supra note, at 29 (noting how these “self-

reinforcing effects may increase with the number of applications provided on a platform, e.g. bundling email, messaging, video, music and telephony as increasing returns to scope kicks in and even more information becomes available thanks to data linkage”).

51 OECD Interim Synthesis Report, supra note, at 29; see also FTC Staff Report, supra note, at 76 (discussing this “virtuous cycle” and how it represents a “significant barrier for any potential entrant”).

52 OECD Interim Synthesis Report, supra note, at 29.


certain point, taper off. We address these issues in the next Part.

II. REDUCTION OF QUALITY IN SEARCH RESULTS

Having explored the three network effects, we will now turn to consider

the possible reduction in quality of search results. We consider three

necessary conditions that could lead to quality degradation: First, the

search engine has an economic incentive and the ability to degrade quality;

second, consumers lack the ability to accurately assess quality; and third,

imperfect information flows make it difficult or costly for others to convey

to consumers the products’ or services’ inherent quality differences.

A. Search Engine’s Incentive and Ability to Degrade Quality

To benefit from the network effects, and the subsequent advertising

revenue, search engines, as the European Commission found, “will try to

attract as many participants on both sides of the platform as possible.”53

This makes sense. If the search engine attracts many users, but few

advertisers, then there will be few relevant sponsored ads for users to click

and less revenue for the search engine. If the situation were reversed, with

advertisers far outnumbering users, the quality of the organic search results

will suffer (with less trial-and-error), and the few users may be turned off

by many sponsored ads. In maximising both the number of users/search

queries and advertisers/sponsored search results, the search engine can

benefit from network effects.

Because the search engine is interested in increasing usage on both sides

of the market (advertising and search inquiries), the initial issue is whether

search engines have an incentive to degrade quality on the free side (search



results) to reap greater profit on the advertising side. Given that search

engines earn their revenues and profits from one side of the market (namely

paid advertising), they have an economic incentive to maximise the

likelihood of consumers clicking on sponsored advertisements. “The trade-

off may arise because when a platform tries to attract more users through

greater relevance on the organic search results it runs the risk of losing

revenues on the advertising side (i.e. users clicking less ads) due to users

clicking predominantly on the organic side.”54 Clicking on a sponsored

search result is not necessarily a bad outcome where the consumer finds the

sponsored search result helpful and the sponsored search result enables the

advertiser to effectively target the consumer. In such cases, no one would

begrudge the fact that the search engine is earning revenue from helping to

match the advertiser with the search engine user.

However, consumers and companies are harmed when the search engine

intentionally degrades the quality of the organic search results below levels

that consumers prefer. Here the search engine can, but elects not to, provide

users with the most relevant results in response to their queries. One

concern is that search engines, to incentivise users to click on sponsored

advertisements or the results of its affiliated business, can give a slight

preference to these sponsored search results by promoting – and ranking

higher – its sponsored results and providing fewer – and ranking lower – the

organic results. The search engine could lower the ranking of potential

advertisers appearing in the organic search to pressure the businesses to

advertise with the search engine, namely to bid for keywords to get the

attention of viewers who do not scroll down the list of search results.55

“Alternatively, the platforms may alter the ranking of the organic search

54 Amelio & Magos, supra note, at 51. 55 MARTIN CAVE & HOWARD WILLIAMS, THE PERILS OF DOMINANCE: EXPLORING THE

ECONOMICS OF SEARCH IN THE INFORMATION SOCIETY 8 (March 2011).


results such that, from the user’s perspective, firms offering competing

products to the sponsored links are given a less-than-optimal ranking on the

organic side.”56 This can hurt consumers with higher prices (due to higher

advertising costs), less relevant results (when companies refuse to

advertise), and less innovation (when companies know that however good

their products or services, they will be unable to effectively reach

consumers online).

In its Microsoft/Yahoo investigation, the European Commission

considered a search engine’s incentive to degrade the quality of its search

results:

The trade off arises because when a platform tries to attract more users through greater relevance on the organic search it runs the risk of losing revenues on the advertising side (i.e. less clicks on ads) due to users clicking predominantly on the organic side (especially if both types of clicks would bring the user to the same kind of information). The degradation of organic search could possibly be achieved through various means. For example, platforms might have an incentive to dedicate a smaller part of the result page to organic results in favour of search advertising links (also called sponsored links) thereby providing proportionally more advertising links. Alternatively, the platforms may rank the sponsored and organic search results in a way that firms offering competing products to the sponsored links are ranked, from the user's perspective, on the organic side lower than optimally.57

As the search engine expands to other services (such as offering vertical

search, such as shopping, restaurant reviews, etc.), the search engine can

56 Amelio & Magos, supra note, at 51 (“For instance, instead of displaying links to

additional merchants in the organic search results, search engines could display links to ‘informational’ sites or placing the links winning the auctions also in prominent positions in the organic search results, in order to decrease substitution between organic and paid searches.”).



demote the rival services.58 The search engine can also blur the distinction

by mingling relevant organic results with less relevant sponsored results.59

It can provide relevant restaurant information, for example, but promote the

reviews from its affiliated online review service and bury the reviews from

a competitor’s more robust sample. The search engine can also degrade

other dimensions of quality, such as the privacy afforded to the search

queries and limitations on using the search data in developing user profiles

and for behavioural advertisements.

Thus, a search engine has the capacity to degrade the quality of search

results. The issue is whether it has the incentive to do so. An important

variant concerns returns to scale and scope. Although search engines can

degrade quality, some, given the network effects discussed in Part I, will

have less incentive or ability to degrade quality. The OECD found non-

linear, increasing returns to scale from searches and personal data. But one

issue is whether the returns to scale and scope taper off at a certain point. A

simple illustration is if the returns to scale level off at a specific level (say

10,000 search inquiries per day per relevant geographic market). Google,

Yahoo!, and Bing will have a relative advantage in terms of relevant search

results over a new entrant (for example, Apple), but that advantage will

disappear as soon as Apple averages 10,000 daily inquiries in those locales.

Moreover, if the two other conditions of our theory hold (customer

detection and competitors’ effectively alerting consumers of degradation),

then neither Google, Yahoo!, nor Bing would have the incentive to

unilaterally degrade quality. Competition would be an effective check.

58 See, e.g., FTC Staff Report, supra note, at 92 (stating that “Google’s threat (and

willingness) to degrade its own web search product—by banishing high-quality vertical websites from its web search results altogether—suggests that Google’s motive in scraping high-quality content from its vertical competitors was not pro-competitive”).

59 See, e.g., FTC Staff Report, supra note, at 24 (noting that Google Universal Search results “often were not labeled as being provided by Google affiliated services, but were integrated directly into the search results”).


Now suppose that the returns to scale taper off for different types of

searches. Suppose two search engines attract a random cross-section of the

population. Suppose one has five million daily searches and the other five

billion. Suppose one per cent of the daily searches involve a hot topic (such

as a scandal involving a celebrity). The smaller search engine has 50,000

searches from which to experiment, the larger 50 million. For the more

popular search, both search engines may have a sufficiently large sample to

identify the more popular, relevant sites, even though the larger website

might do a better job identifying and ranking the second or third tier of

results. However, the results are different for esoteric search queries (“tail

queries”). Suppose a search, such as “law and economics professors” and

NCAA, averages one per 10 million searches. The smaller search engine

will get one search every two days, the larger search engine gets 500 queries

per day.

Consequently, to identify relevant responses for frequent and popular

queries, the search engine might need fewer average daily searches (say

100,000) than for less frequent, more esoteric tail inquiries. So the smaller

search engine may perform as well in terms of quality for popular searches,

but struggle for less frequent searches, where its sample size is much

smaller than the larger search engines. If new searches are a significant

component of daily searches (say 30 per cent), then the larger search engine

will enjoy an inherent advantage. In averaging five billion searches per

day, a leading search engine is likely to provide more relevant results across

the board (long tail to popular queries) than a search engine averaging five

million searches per day. Under this scenario, the larger search engine’s

results will continually improve as more people use the search engine for all

of their searches, but the incremental benefits from scale (increase in

relevance) will taper off as the sample size approaches all users.


Thus, a search engine’s ability and incentive to degrade quality will

depend on whether the company is at a relative advantage or disadvantage

in scale compared with other search engines.

Microsoft and Yahoo advanced this argument to the European

Commission to justify their 10-year partnership, whereby Microsoft would

control the online web-wide algorithmic search and search advertising

business of Yahoo. Microsoft and Yahoo argued that they needed to attain

greater scale and thereby produce more relevant search results to better

compete against industry leader Google. As the parties argued, their

incentive was to increase the number of queries and users to attain similar

scale with Google, which lessened their incentive to shade quality.60

This argument, of course, assumes that the benefits from scale have not

already levelled off for Microsoft and Yahoo. For if Microsoft, Yahoo, and

Google were all operating at or above the minimum efficient scale for

search queries, if the other network effects were not involved, and if the

quality of the algorithms were similar, then the quality of their search

results would be roughly similar. Under this scenario, quality degradation

could still occur, but along the lines of tacit collusion. Each search engine

could monitor the other, and follow the quality degradation of the other. In

effect, they could tacitly collude on one quality parameter (such as slightly

favoring their sponsored ads over their organic search results). It would be

akin to three television stations, each following the other in terms of product

placements, advertising time, etc.

60 Microsoft/Yahoo, supra note, at ¶¶ 211, 213. Moreover, Microsoft argued that the

structure of its transaction with Yahoo provided “a large incentive to innovate on the search experience because, according to the definitive Agreements, the publishing businesses of Microsoft and Yahoo will remain separate, and the revenue attributable to either Yahoo or Microsoft depends on the source of the search query.” Microsoft/Yahoo, supra note, at ¶ 208. Yahoo and Microsoft would only earn revenue from ads that appear on their respective websites (and those of its publisher affiliates), so each would have a strong incentive to attract users to their own entry points. Microsoft/Yahoo, supra note, at ¶ 208.


The U.S. Department of Justice accepted that the benefits from scale

might enable a combined Microsoft/Yahoo to better compete against the

dominant search engine:

The search and paid search advertising industry is characterized by an unusual relationship between scale and competitive performance. The transaction will enhance Microsoft’s competitive performance because it will have access to a larger set of queries, which should accelerate the automated learning of Microsoft’s search and paid search algorithms and enhance Microsoft’s ability to serve more relevant search results and paid search listings, particularly with respect to rare or “tail” queries. The increased queries received by the combined operation will further provide Microsoft with a much larger pool of data than it currently has or is likely to obtain without this transaction. This larger data pool may enable more effective testing and thus more rapid innovation of potential new search-related products, changes in the presentation of search results and paid search listings, other changes in the user interface, and changes in the search or paid search algorithms. This enhanced performance, if realized, should exert correspondingly greater competitive pressure in the marketplace.61

The European Commission also found scale to be an important factor to

effectively compete,62 with a caveat: “While the Commission notes that

Google appears to perform better in terms of relevance especially for [...]

queries, this does not provide evidence that scale leads to higher relevance

for users, since the above studies do not take into account the technology of

the different search engine which are not related to scale.”63 The

Commission found that the search platform’s revenue per search increased

61 Press Release, U.S. Dep’t of Justice, Statement of the Department of Justice

Antitrust Division on Its Decision to Close Its Investigation of the Internet Search and Paid Search Advertising Agreement Between Microsoft Corporation and Yahoo! Inc. (Feb. 18, 2010), available at http://www.justice.gov/atr/public/press_releases/2010/255377.pdf.

62 Microsoft/Yahoo, supra note, at ¶ 153; Vecchi et al., supra note, at 46 (noting that EC’s market investigation “confirmed that scale is an important aspect in the economics of the industry”) & 47.



with the volume of search queries.64 Competitors that responded to the

Commission’s market investigation indicated “almost unanimously that

scale is important in order to be an effective competitor in search

advertising”65 and that “Microsoft did not have enough traffic to compete

effectively with Google.”66 Google was the exception, arguing that “while

scale is an important and necessary ingredient of having a successful search

engine, its degree of importance has been largely overstated.”67 The

Commission did not elaborate, but noted how both Google and Microsoft

acknowledged that “the value of incremental data decreases as the amount

of data increases.”68 The issues of when that happens, for which types of

searches and the quality differential were left unresolved.69

Even when the scale effects taper off, another issue is whether the

search engines can still have a relative advantage in scope, by being better

able to offer users personalised search results. And to the extent to which

users are attracted to one search engine – even after scale and scope effects

taper off – the search engine benefits in being the “must use” option for

advertisers, and thus can deliver more personalised ads.

64 Microsoft/Yahoo, supra note, at ¶ 168. 65 Microsoft/Yahoo, supra note, at ¶ 173. 66 Vecchi et al., supra note, at 46. 67 Microsoft/Yahoo, supra note, at ¶ 174. 68 Microsoft/Yahoo, supra note, at ¶ 174. 69 The FTC Staff Report, in discussing the “scale curve,” also noted the dispute.

Google acknowledged “the importance of scale in the abstract.” FTC Staff Report, supra note, at 16. Its internal documents were “replete with references to the ‘virtuous cycle’ among users, advertisers, and publishers.” Id. Its executives, however, testified of diminishing returns from scale, and that Google has “enough users already that more users don’t make it much better.” Id. at 124 n. 77 (quoting Schmidt). Google also argued that “Bing’s query and advertiser volume have passed the point at which scale should - or would - matter significantly to Microsoft, and that any volume gains made by Bing would yield minimal improvements in either Bing’s search quality or its monetization ability.” Id. at 16. Microsoft agreed that there were generally diminishing returns to scale. Id. So the “main bone of contention between Google and Microsoft is where on this scale curve Microsoft currently operates. This is an important question, but one which evades easy answers. This is, in part, because neither party can identify a fixed number of queries or ads that constitutes the ‘minimum efficient’ point of operation.” Id.


Accordingly, some search engines will be less likely to degrade quality

than others. A dominant search engine may have the incentive and ability

to degrade the quality of its search results to a slight degree, but could still

produce better quality search results, due to the network effects, than its

smaller rivals. Under this scenario, Google could shade its search results to

a greater extent than the smallest search engine, DuckDuckGo. So unless a

smaller search engine seeks to differentiate its services to such a degree that

it no longer competes against the dominant search engine based on quality

of search results, the smallest search engine will have less incentive and

ability to degrade quality. The quality of its search results is already

relatively inferior (given the scale at which it operates, the scope of data it

collects, and the number and diversity of advertisers). It already faces an

uphill battle to sway consumers to use its search engine; it cannot afford to

further degrade its search results.

Conversely, even with multiple competitors, the largest search engine,

with a significant advantage from its scale and scope, could degrade quality

on the free side when it increases its market power or profits. But even a

dominant search engine will have a disincentive to skew search quality

when it reduces profits. For example, the dominant search engine, if it

provides many irrelevant sponsored ads, will not earn more income if users

do not click the irrelevant ads. If one searches for running shoes and gets

many sponsored search results for dress and children shoes, one will not

click the sponsored links. If the search engine continues to provide mainly

irrelevant sponsored results, then it risks consumers generally ignoring

sponsored ads or switching to other search engines. Moreover, the

dominant search engine would still need to invest in improving its search

algorithms to match users and advertisers.70 However, within these



parameters the dominant search engine can afford to degrade quality in the

face of competition (i.e., it enjoys a relative quality advantage from network

effects over its smaller search engine rivals) and when doing so increases its

profits or market power.71

B. Consumers’ Ability to Accurately Assess Quality Differences

Even when search engines have the ability and economic incentive to

degrade the quality of their search results, they will not do so where

consumers would likely perceive the quality degradation and punish them

by switching to rival search engines.

In Microsoft/Yahoo, Microsoft, for example, argued that it could not

degrade the quality of their search engine results, given Google’s presence:

Users, according to the notifying party, notice inferior (i.e. less relevant) search results and given the strong presence of Google in the market they would direct their queries back to Google lowering user engagement in Microsoft’s platform. Microsoft also states that the search industry is closely monitored by industry participants and opinion leaders. Their view as well as mainstream media influences users’ perceptions of search engines. Microsoft therefore claims that it cannot run the risk of even small degradations in quality, as users will become aware and switch to Google.72

This argument assumes that consumers can detect degradations in the

quality of search results. Search engine users undoubtedly can detect

significant degradations in quality. If one searches for tennis shoes, and

receives results for the search engine’s other, unrelated offerings (such as

voice-mail), the degradation in quality is evident. A more complex scenario

may concern the limited relevance of organic search results and the high

relevance of sponsored ads. Users may consider the overall quality to be

higher in such cases where they do not distinguish between the sponsored

71 Microsoft/Yahoo, supra note, at ¶ 204. 72 Microsoft/Yahoo, supra note, at ¶ 212.


and organic results.

Overall, beyond significant degradations in quality, the quality of

search results is relative. First it is relative across time. Search results can

improve in quality as, among other things, more people use the search

engine. Second, quality is relative to the amount of potentially relevant

information available online. When more information (such as books) is

brought online, and when more of the online material is searchable (such as

texts), the more likely the search engine is to identify relevant information.

Thus, the quality of the search results today on a smaller search engine may

be better than the search results from Yahoo or Google ten years ago.

Third, the quality of a search engine’s results is relative to the

contemporaneous results of other search engines. Smaller search engines

may identify more relevant results for popular, more straightforward search

queries (which requires a smaller number of average daily searches) than

the long tail queries (which might arise once every million searches).

Thus, consumers will likely base the quality of search results not on

some objective, ideal benchmark but relative to what other search engines

offer. If quality is relative, then to notice its degradation, users must run the

same search query around the same time on different search engines.

Technically, nothing prevents consumers from multi-homing searches. But

in reality consumers infrequently make the same inquiry on multiple search

engines. As the Commission found in its earlier investigation, “The very

limited share of user multi-homing between Microsoft and Yahoo shows

that users rarely run checks between these two platforms.”73 Yahoo users’

second choice was Google,74 and the Commission confirmed the parties’

73 Microsoft/Yahoo, supra note, at ¶ 221; see also Vecchi et al., supra note, at 44

(Commission finding that “users tend to ‘single-home,’ meaning that they perform over 90% of their search queries within a month on one single search engine”).



figures that the “vast majority of users ‘single-homed’ on Google.”75

Moreover, the market shares do not bear out significant multi-homing.

For if users were frequently running the same searches on Google, Bing,

Yahoo, and DuckDuckGo, then each competitor would control

approximately 25 per cent of the search market. If users were frequently

multi-homing among the search platforms, one would not see the disparity

in market share for search queries.76 In March 2014, Google continued to

lead the U.S. explicit core search market “with 67.5 percent market share,

followed by Microsoft Sites with 18.6 percent (up 0.2 percentage points)

and Yahoo Sites with 10.1 percent. Ask Network accounted for 2.5 percent

of explicit core searches (up 0.1 percentage points), followed by AOL, Inc.

with 1.3 percent.”77 In Europe, Google has an even larger share of the

search market. In 2013, the Commission observed how Google “has been

holding market shares in web search well above 90% in most European

countries for several years now, a level which is higher than in many other

parts of the world.”78

Another possibility is spot-checking. Even if consumers do not

routinely run the same search query across multiple search engines, they

may occasionally use another search engine to assess the relative quality of

search results. As the Commission stated, the “mere presence of an

alternative check may suffice to induce the search engines to enhance the

relevance of their organic search.”79

It is unclear how frequently consumers actually spot check (and whether 75 Vecchi et al., supra note, at 46; see also Microsoft/Yahoo, supra note, at ¶ 102. 76 Microsoft/Yahoo, supra note, at ¶ 157 (noting how the search engine market shares

“clearly demonstrate that Google has a much greater user reach and a much larger number of single homers that never use another search engine”).

77 Press Release, comScore Releases March 2014 U.S. Search Engine Rankings (Apr. 15, 2014), http://www.comscore.com/Insights/Press-Releases/2014/4/comScore-Releases-March-2014-U.S.-Search-Engine-Rankings.

78 EC April 2013 Google Memo, supra note, at 1. 79 Microsoft/Yahoo, supra note, at ¶ 221 n. 71.


they randomly spot check or are likelier to spot check for certain categories

of search inquiries or at certain times). For example, one survey asked

search users “what they would do if a Google search result did not contain

the expected information”: 34 per cent of respondents said they would

return to the search results page and try a different result; 25 per cent said

they would return to Google to enter a new search.80 None of the

respondents indicated trying another search engine.

These results may suggest that the “switching costs” between search

providers (or, alternatively, the “costs of spot checks”) are perceived by the

users as higher than the benefit they expect to obtain through these actions.

Even if many consumers did spot check, given the network effects and

benefits from scale and scope in the search market, the spot checks would

likelier keep the smaller search engines in check, but not necessarily the

larger ones. To illustrate, suppose one generally runs 100 searches per

week on Google, one weekly quality check on Bing, and one monthly

quality check on DuckDuckGo. At the end of the year, Google would have

5200 searches, compared to Bing’s 52 searches, and DuckDuckGo’s 12

searches. Google, with its larger user platform, would attract still more

advertisers and have more advertising revenue. With more search queries,

it could offer more relevant organic search results and more relevant

sponsored search results to target users. But, despite spot checks, Google

could degrade quality, while maintaining its quality advantage due to the

network effects.

Spot check comparisons become even harder with personalised

searches. If the consumer not only uses the platform’s search engine, but

also its email, mapping services, etc., then the search engine, using the

personal data, can offer personalised search results. This makes

80 http://searchengineland.com/?p=155708.


comparisons harder with other search engines that do not have the personal

data and user profiles.

Consumers would also have to factor in whether their query is long-tail

or popular. As we have discussed, the scale effects will likely vary

depending on the amount of trial-and-error. For example, DuckDuckGo

might suffice for a popular search (such as a recent celebrity scandal)

because with many of its users searching that topic, it has a sufficiently

large sample size to experiment. That will not be the case for more esoteric

searches (such as information on antitrust law professors). Rather than try

to ascertain which search engine does better for which type of inquiry, the

user may simply opt for the one that does a better job overall for all types of

searches (both popular and long tail queries).81

Spot checking will also be less feasible with greater differentiation

among the search engines. Smaller search engines, given these network

effects, will likely realise the uphill challenge in improving search quality

(in terms of relevance) relative to their larger competitors. To improve the

relevance of its organic and sponsored results, the smaller search engine has

to attract more users, which afford more opportunities to experiment with

many different kinds of search inquiries, which in turn attracts more

advertisers. However, the search engine cannot attract more users until its

search results are at par with the larger competitors. Thus, to increase in

size, a smaller search engine may differentiate its services on some other

parameter of quality. DuckDuckGo, for example, highlights its superior

privacy protections, using the tag line “The search engine that doesn't track

81 FTC Staff Report, supra note, at 66 (“In effect, users are habituated into using

Google for all their queries because of its comprehensive scope, and so they may be more likely to tum to Google when they have commercial queries, instead of starting at a vertical website. [Google’s Eric] Schmidt’s testimony is corroborated by the representations of several of the vertical search firms, who note that they are dependent on horizontal search providers for significant amounts of their traffic, because even many vertical search users tend to begin their search with a query on Google, Bing or Yahoo!.”).


you.”82

Consequently, even with consumers’ spot checking, larger search

engines can degrade quality, while smaller search engines, given their

quality disadvantage, cannot. The consumer would be unable to discern the

extent to which the dominant search engine has degraded quality, only that

the larger search engine generally produces somewhat better results for the

universe of searches.

The discussion thus far highlights the difficulties for rational consumers

to objectively measure quality differences and detect deteriorations in

search quality. However, as the recent developments in the economics

literature show, consumers are not necessarily rational, self-interested

profit-maximisers with perfect willpower. One issue is the extent to which

quality is part of branding. Consumers at times subjectively believe that a

higher priced good is indeed better.83 Because searching is free, these price

effects are absent. Nonetheless, consumers’ perceptions of quality may be

affected by brand, as one report describes:

In a recent study by SurveyMonkey examining SEO assumptions, respondents were given two search result pages, one with a page header labeled “Google” and the other with a page header labeled “Bing,” and asked which page of results they preferred. Even when the page header labels were swapped, more users preferred the Google search results.

Of 641 survey respondents, 379 participants received a

survey asking which of two search result pages they preferred. One page of results for the term “file taxes” included true Google results and the other page included true Bing results. The Google page was chosen by majority of the respondents.

82 https://duckduckgo.com. 83 DAN ARIELY, PREDICTABLY IRRATIONAL: THE HIDDEN FORCES THAT SHAPE OUR

DECISIONS 181-86 (HarperCollins 2008).


A second survey was given to 262 participants. Using the same search term, respondents had to choose between a Google search results page and a Bing search result page. In this survey the SERP headers were swapped with Google results listed as Bing results and Bing results listed as Google results. Of the respondents who received the swapped search result pages, a larger percentage of respondents still chose Google results, even though they were actually Bing search results.84

The European Commission found that search engine users were not

aware of the promotion of Google’s services within the search results.85

This is consistent with our findings that consumers may observe substantial

quality degradations, but not the slow degradation of quality in search

results by a leading search engine, which enjoys a quality advantage due to

scale, network and branding effects.

C. Imperfect Information Flows Under the third variable of our theory, quality degradation is feasible

when companies recognise that neither they nor their competitors can easily

or inexpensively convey to consumers the inherent quality differences in

their and their competitors’ product offerings.

Several factors favour information flows for search engines. First, rival

search engines have the incentive to alert consumers of quality degradation

given the importance of scale. If consumers do not multi-home search, then

the contest for users can become zero-sum: every search query on Bing or

Yahoo makes it less likely that consumers will redo a search on Google,

narrowing the gap between the competitors, and advantage from network 84 http://searchengineland.com/users-prefer-google-even-when-155682; see also FTC

Staff Report, supra note, at 66. 85 European Commission Press Release, Antitrust: Commission Obtains From Google

Comparable Display of Specialised Search Rivals (Feb. 5, 2014), http://europa.eu/rapid/press-release_IP-14-116_en.htm.


effects. Second, companies harmed by search degradation also have the

incentive to inform consumers and competition authorities, as some did

before the FTC and European Commission.

Third, rival search engines have alerted consumers to superior quality.

Microsoft, for example, advertised its quality advantage over Google with

its Bing-it-on challenge. Microsoft asked users to take a blind test, where

users select five search queries, and choose which results (Google’s or

Bing’s) they favour.86 In blind tests, according to Microsoft, more U.K.

residents preferred Bing’s search results over Google’s.87 Despite this

advertising campaign, however, U.K. residents overwhelmingly use

Google.88

Why does Google still dominate by such a margin in search results in

the U.K.? One possibility is that Microsoft’s sample does not reflect the

U.K. Internet user population. A second possibility is that Bing

outperforms Google on some searches (such as popular searches) but not

other inquiries (such as long-tail queries), where Google’s scale comes into

play.89 A third possibility involves objective or perceived quality. A fourth

possibility may be linked to the benefit of loyalty and customised search

results. This consideration may be further amplified due to status quo bias.

Status Quo Bias - The Power of Defaults

86 http://www.bingiton.com. 87 http://www.bingiton.com (noting that “[d]espite having used Google’s own top

queries, after carrying out 10 searches, 46% of people surveyed picked Bing search results more often, 37% of people picked Google results more often, and 17% of people chose Bing and Google results an equal number of times”).

88 http://theeword.co.uk/info/search_engine_market/. 89 For a criticism of the Bing results, see

http://islandia.law.yale.edu/ayres/BingItOn_Draft%209.pdf. For Microsoft’s response, see http://blogs.bing.com/search/2013/10/02/challenging-the-challenge-to-the-bing-it-on-challenge/.


Many consumers, as the European Commission discussed, will stick

with the default option even when a superior alternative exists.90 Status quo

bias arose in the Commission’s case involving Microsoft’s Windows Media

Player.91 As the Commission recently recognised, pre-installing software

on a large number of smartphones may make it less likely that consumers

will switch to alternatives.92 As the OFT discussed in its Google/Waze

decision, integrating the pre-installed application with other functions on

the platform may further reduce the likelihood of consumers switching.93

Thus, another confounding variable is the search engine’s ability to use

status quo bias to degrade quality, including less-visible aspects such as

privacy protections.

One recent case reflects this power of defaults. After the Internet

browser Firefox made Yahoo the default search engine, Yahoo’s share grew

from 8.6 per cent in November 2014 to 10.9 per cent of the U.S. search

market in January 2015, its highest share in the past five years.94 It appears

that this growth was attributable to Firefox. When Firefox users were

separated out from Yahoo’s overall share, Yahoo’s remaining share was

reportedly “flat or down slightly vs. last month.”95 Here, many Firefox

users were using Google, and after Firefox changed the default, Google

sought to persuade Firefox users to switch back to Google. But, according

90 For the importance of default options generally, see RICHARD H. THALER & CASS R.

SUNSTEIN, NUDGE: IMPROVING DECISIONS ABOUT HEALTH, WEALTH, AND HAPPINESS (Yale Univ. Press 2008).

91 Commission decision of 16 December 2009 in Case C-3/39.530—Microsoft (Tying); Maurice E. Stucke, Behavioral Antitrust and Monopolization, 8 J. COMPETITION LAW & ECONOMICS 545 (2012).

92 FaceBook/WhatsApp, supra note, at ¶¶ 111, 124 (discussing status quo bias). 93 Google/Waze, supra note, at ¶¶ 57-61. 94 http://searchengineland.com/firefox-deal-continues-boost-yahoo-us-search-share-

grows-january-213998 95 http://searchengineland.com/firefox-deal-continues-boost-yahoo-us-search-share-

grows-january-213998


to reports, many Firefox users stuck with the new default engine Yahoo.96

It will be interesting to see whether many Firefox users continue to stick

with the default search engine Yahoo.

Search engines estimate the value of being the default search engine on

a browser, and pay the browser for this right.97 Google, for example, is the

default engine on Apple’s Safari Internet browser. Google reportedly paid

Apple $82 million in 2009, and $1 billion in 2013 for this partnership.98

With Google’s deal with Apple set to expire in 2015, there are rumors that

Apple may launch its own search engine.99

So why does Google pay Apple $1 billion annually to be the default

search browser on all Apple devices running iOS, when it could simply tell

Apple users that Google is just one (or two clicks) away? If Google offered

superior search results, why did many Firefox users simply stay with the

new default, Yahoo?

This demonstrates that defaults matter. It will be interesting to see

whether another company becomes the default search engine on Safari, and

whether Apple launches its own search engine. If many browser users, like

the Firefox users, stick with the browser’s default search engine, then the

battle over search will be the battle over the default option as inertia may

trump over quality differences among the search engines.

96 http://searchengineland.com/firefox-deal-continues-boost-yahoo-us-search-share-

grows-january-213998 (StatCounter CEO Aodhan Cullen posted, “Some analysts expected Yahoo to fall in January as a result of Firefox users switching back to Google. In fact Yahoo has increased US search share by half a percentage point.”); Gregg Keizer, Yahoo Loses Some U.S. Share Gained from Firefox Deal, COMPUTERWORLD, Apr 17, 2015, http://www.computerworld.com/article/2911108/yahoo-loses-some-us-share-gained-from-firefox-deal.html (noting that after a January 2015 peak of 13 per cent, Yahoo's search share dropped in February to 12.8 per cent, and in March to 12.7 per cent).

97 http://searchengineland.com/apple-google-deal-expires-will-win-safari-default-search-business-214277

98 http://ibnlive.in.com/news/apple-working-on-its-own-search-engine-aims-to-take-on-google-report/527597-11.html

99 http://ibnlive.in.com/news/apple-working-on-its-own-search-engine-aims-to-take-on-google-report/527597-11.html


Status quo bias has important implications for search quality and

competition. Search engines, in order to grow (or maintain share), will

battle to be the default search engine at key entry points for search. This

includes, as the Firefox and Safari deals show, being the default search

engine on the Internet browser, and entering into syndication agreements

with third-party websites.100

However, given the significant revenue stakes in capturing search

advertising markets, the battle will also mean the search engines capturing

the key entry points with their own products. Search engines will have the

incentive to develop and promote their own Internet browser (such as

Google’s Chrome and Microsoft’s Internet Explorer) and their own

operating system (such as Google’s Android for mobile and Microsoft’s

Windows for mobile and personal computers).101 While the battle for

search may appear to be one click away, it actually means controlling

(either internally or through contractual agreements) other key entry points

for search. Thus, to the extent to which the search engine is the default

option and integrated with the operating system, the battle for search may 100 Microsoft/Yahoo, supra note, at ¶ 51. 101 According to one report, http://www.benedelman.org/news/021314-1.html,

Google’s Mobile Application Distribution Agreement requires the following from mobile phone manufacturers:

"Devices may only be distributed if all Google Applications [listed elsewhere in the agreement] ... are pre-installed on the Device." See MADA section 2.1.

The phone manufacturer must “preload all Google Applications approved in the applicable Territory … on each device.” See MADA section 3.4(1).

The phone manufacturer must place “Google's Search and the Android Market Client icon [Google Play] ... at least on the panel immediately adjacent to the Default Home Screen,” with "all other Google Applications ... no more than one level below the Phone Top." See MADA Section 3.4(2)-(3).

The phone manufacturer must set “Google Search ... as the default search provider for all Web search access points.” See MADA Section 3.4(4).

Google's Network Location Provider service must be preloaded and the default. See MADA Section 3.8(c).

These provisions are confidential and are not ordinarily available to the public. MADA provision 6.1 prohibits a phone manufacturer from sharing any Confidential Information (as defined), and Google labels the MADA documents as "Confidential" which makes the MADA subject to this restriction.


be the battle for the dominant browser and operating system, especially on

mobile phones, where search is increasingly occurring.102 This further

raises the entry barriers to the search market, and the costs of expanding in

this market.103

Although consumers can perceive differences in quality between search

engines when confronted with side-by-side comparisons in blind tests, it is

not altogether clear that consumers, even with direct “Bing-it-on” quality

challenges, act upon quality differences in real life. If many consumers

stick with the default option, then rival search engines using Bing-it-on and

other direct comparisons of quality will be unable to overcome users’ status

quo bias. If most users will use the default search engine, then the search

engine that becomes the default option on most “entry points” for search

(such as the Internet browser) gets the most users and attracts the most

advertisers. Depending how strong the status quo bias is, the dominant

search engine can afford to shade quality to an even greater degree.

Thus, the likelihood of anticompetitive quality degradation increases

when the search engine controls the essential portals to search (including

the mobile operating system), when the search engine controls and limits

the portability of data so that users cannot export their search and other

personal data to rival engines (thereby helping the search engine provide

more relevant results), and when the parties’ applications are pre-installed

on a large base of mobile phones, tablets or PCs, so that “status quo bias”

would affect consumers' choices.104

CONCLUSION

Quality forms a fundamental aspect of competition and a key non-price

102 http://uk.businessinsider.com/google-search-share-below-75-2015-2?r=US 103 Microsoft/Yahoo, supra note, at ¶ 111 (noting that entry barriers “appear to be

high”). 104 Facebook/WhatsApp, supra note, at ¶ 134.


consideration, which affects consumers’ choice. As such, one would expect

it to be the prime variable that determines use in free goods, and, in the

context of our discussion, in search engine competition. The European

Commission in the Microsoft/Yahoo venture considered the joint venture’s

degrading the quality of the Internet search results as a theory of harm.105

But the Commission’s underlying assumption was that competition would

prevent the degradation of quality.106

Yet, as we have shown, given several conditions, in a two-sided market

scenario, a search engine may have the incentive and ability to degrade its

search results’ quality to the detriment of consumers. Quality degradation

may be limited in nature, yet present and intentional. Here, consumers

prefer different, more relevant results, which the search engine opts not to

provide. Instead the search degrades the search quality to earn greater

profits or market power.

Our conclusion on the possibility for quality degradation leads us to the

wider debate over when such quality degradation violates the antitrust laws.

The answer to this question depends on one’s view on the role of

competition law and the ability of markets to correct themselves. Even if

one favors intervention, the triggering point for intervention and the

effectiveness of the remedies may be challenging.107 One issue is if the

search engine simply degrades quality. Suppose a monopolist intentionally

degrades the quality of its search results, but its quality, due to network

effects, is still superior to rivals. Should competition policy condemn such

105 Microsoft/Yahoo, supra note, at ¶¶ 202-04. The Commission in that case left open

whether Internet search constituted a separate market. Id. ¶¶ 85-86. 106 Amelio & Magos, supra note, at 52 (“In summary, the literature suggests that an

important role for competition is to induce search engines to provide more relevance.”). 107 Understanding why rival search engines do not prevent quality degradation may

also help competition authorities with designing remedies. Competition authorities might consider, for example, how search and user data portability might level the playing field, and the use of choice screens (as the Commission did in its Microsoft browser case).


relative degradation? And if so, how does one identify the optimal point of

intervention?

Less controversial is if the quality degradation is part of the company’s

plan to leverage its dominance into other markets and thereby harm

consumers and lessen the incentives to innovate. Mechanisms that may

foster status quo bias – such as long-term agreements to be the default

search engine and bundling search engines with operating systems – might

warrant scrutiny, when used by a dominant firm. A related and interesting

development is the European Commission’s announcement, on the same

day it issued its statement of objections, that it was investigating Google’s

Android. The Commission will look, among other things, “into Google

allegedly requiring or incentivising smartphone and tablet manufacturers to

exclusively pre-install Google’s own applications or services, in particular

Google’s search engine.”108

Ultimately, the European Commission’s Statement of Objections in

Google opens the door to wider considerations of consumer welfare and

quality. The case signals the growing importance of quality competition as

consumers are increasingly offered free goods and services in exchange for

their personal data and ability to be targeted with behavioural ads. No

doubt we will see other cases involving quality degradation along various

dimensions, including privacy protection, even when competition is a click

away.

108 European Commission, Statement by Commissioner Vestager on Antitrust

Decisions Concerning Google, April 15, 2015, available at http://europa.eu/rapid/press-release_STATEMENT-15-4785_en.htm. The Commission will also investigate “the alleged bundling together of certain Google products with other apps and services” and “if Google is hindering the ability of manufacturers of smartphones or tablets, who want to use the Android operating system, from being able to use and develop other open-source versions of Android (so-called ‘Android forks’).” Id.

quality two-sided-markets final

Documents