University of Calgary PRISM: University of Calgary's Digital Repository

Graduate Studies The Vault: Electronic Theses and Dissertations

2019-08-28 ’s Antitrust Woes and Google Shopping

White, Adam Joseph

White, A. J. (2019). Google’s Antitrust Woes and Google Shopping (Unpublished master's thesis). University of Calgary, Calgary, AB. http://hdl.handle.net/1880/110865 master thesis

University of Calgary graduate students retain copyright ownership and moral rights for their thesis. You may use this material in any way that is permitted by the Copyright Act or through licensing that has been assigned to the document. For uses that are not allowable under copyright legislation or licensing, you are required to seek permission. Downloaded from PRISM: https://prism.ucalgary.ca UNIVERSITY OF CALGARY

Google’s Antitrust Woes and Google Shopping

by

Adam Joseph White

A THESIS

SUBMITTED TO THE FACULTY OF GRADUATE STUDIES

IN PARTIAL FULFILMENT OF THE REQUIREMENTS FOR THE

DEGREE OF MASTER OF ARTS

GRADUATE PROGRAM IN ECONOMICS

CALGARY, ALBERTA

AUGUST, 2019

© Adam Joseph White 2019

2 ABSTRACT In 2017, Google was fined 2.7 billion USD by the European Commission for its abuse of dominance when it promoted its Google Shopping service above rival comparison shopping services on Google’s results page. Comparison shopping services suffered from the conduct as they received less traffic as users selected Google’s prominently placed Google Shopping service. This thesis will question whether Google’s conduct was anticompetitive, or were they incorrectly fined for pro-competitive conduct? Additionally, Canada’s Competition Bureau and the United States’ Federal Trade Commission exonerated Google for the same activity in past years, so why would these agencies with similar goals and legislation come to different opinions regarding Google’s conduct? This thesis concludes that Google’s conduct is pro-competitive and that the United States and Canada were correct in their decision to cease the investigation. This thesis also identifies three reasons why the European Commission could have legally justified fining Google, even if it was not economically justified. The analysis conducted in this thesis could give guidance to other similar cases which Google is being investigated for, such as the investigation into and .

i

ACKNOWLEDGEMENTS

I owe much to many people who have helped me get to where I am today. I’d first like to thank the Economics Department at the University of Calgary, and specifically, Dr. Church, who has been nothing but supportive and guiding during the past three years. Without Dr. Church, I would not be an inch of where I am today, and I owe him a great deal. Next, I would like to thanks my fellow graduate students and faculty of the Department of Economics as they have been nothing but supportive and helpful during my time there. I’d also like to thank my family for their constant support and encouragement, which I am very grateful for. Finally, I’d like to thank Jill for her sacrifices and willingness to give feedback about any arcane idea that pops out my mouth. I would not be where I am today without her.

ii DEDICATION

To Jill

iii Table of Contents ACKNOWLEDGEMENTS...... ii DEDICATION ...... iii Chapter 1: Introduction ...... 1 Chapter 2: Facts of the Case ...... 7 2.1 Chapter Introduction ...... 7 2.2 How Google Works ...... 8 2.3 Unique Economics of Google ...... 13 2.4 Vertical Search ...... 18 2.5 Overview of the Allegations Against Google ...... 20 Chapter 3: Case Theories ...... 25 3.1 Chapter Introduction ...... 25 3.2 Market Definition ...... 26 3.3 Standard for Harm ...... 27 3.4 Barriers to Entry and Economies of Scale ...... 30 3.5 Network Effects and Externalities ...... 33 3.6 Theories of Self-Preferencing ...... 36 3.6.1 The Economics of Raising Rivals Costs or Reducing Rivals Revenue...... 36 3.6.2 Discrimination Theory...... 41 3.6.3 Refusal to Deal Theory...... 43 3.6.4 Tying Theory...... 47 Chapter 4: The Facts Found by the European Commission ...... 53 4.1 The EC’s Document Regarding Their Decision in Google Shopping ...... 53 4.2 EC’s Market Definition ...... 53 4.3 EC’s Statements to do with Barriers to Entry ...... 57 4.4 The European Commission and Network Effects ...... 60 4.5 The EC and Google’s Self Preferencing ...... 61 Chapter 5: Comparison Between the Facts Found in this Thesis and The EC ...... 70 5.1 Chapter Introduction ...... 70 5.1.1 Barriers to Entry in Search Engines...... 73 5.1.2 Network Effects in Search Engines...... 77 5.2 Market Definition ...... 81

iv 5.3 Theories of Self-Preferencing in Search ...... 86 5.3.1 Raising Rivals Cost and Search...... 86 5.3.2 Discrimination Doctrine in Search...... 93 5.3.3 Refusal to Deal in Search...... 96 5.3.4 Tying in Search...... 100 5.4 Comparison Between Findings of EC and This Thesis ...... 107 5.5 General Comments about the EC Decision ...... 108 5.5 What’s Next for Google and the EC ...... 117 Chapter 6: Conclusion ...... 121 References ...... 125

v Chapter 1: Introduction

On June 27, 2017, the European Commission (EC), the European Union’s competition authority fined Google USD 2.7 billion for their abuse of dominance in promoting its Google

Shopping service. The strangest thing about this fine was not the fact that it happened, it was that the United States’ Federal Trade Commission (FTC), along with the Department of Justice

(DOJ) and the Competition Bureau in Canada, decided that Google’s conduct was not anticompetitive a few years earlier. Was Google’s conduct anticompetitive? Also, why would these antitrust agencies with similar legislation and goals both come to different conclusions with regards to this case? The why behind the difference of opinions between these agencies is the question this thesis will hope to explore and analyze. This thesis concludes that Google should not be found guilty for promoting Google Shopping and that the United States and Canada were correct in their decision to exonerate Google.

Larry Page and launched their search engine in 1998 facing down incumbents like Yahoo! and AltaVista. Nowadays, Google handles over five billion searches a day (Sullivan, 2017), with profits reaching over 26 billion USD (Molla, 2017). While Google did start as an small entrant, their innovative search engine and superior business acumen lead them to overtake both Yahoo! and AltaVista in users by 2006. As Google started to take over general search, claims of anticompetitive behaviour started to roll in. These claims often come from the fact that with recent updates to Google’s search engine, they had started to promote their services above other companies providing the same services. For example, Google promotes its maps service above MapQuest, as well as ensuring that users do not have to leave their search engine results page, by way of an in-page map. Another example and the primary focus of this thesis is

1 Google’s promotion of Google Shopping over other comparison price services, which then led to reduced traffic and the shutdown of several rival comparison price services.

This case, like all high-tech cases, should balance the innovation incentives of firms, while also protecting the competitive process which breeds innovation. If the dominant firm today is not allowed to operate as they see fit due to the adverse effects on other players in the market, then consumers may not have the best version of the products, as well as having perverse investment incentives for the dominant firm. Conversely, if dominant firms cannot impede entrants, then these entrants may perhaps be able to offer a product which delivers large amounts of surplus to consumers. These cases are challenging for competition authorities to parse, as to make the correct decision, they need to measure the observable effects of the conduct against a hypothetical world where the conduct never happened. For the competition authorities to then make the right decision, they must lay out accurate facts and theories to create a consistent and coherent story explaining whether or not the conduct is anticompetitive. This thesis’ aim is then to set out the correct facts from the industry dynamics present, and then compare with what the competition authorities have found, and then diagnose why they made the decision they did.

After this introduction, Chapter Two will go into facts about Google and the industry it inhabits. There will be an explanation of how Google’s search engine works and how they make money by offering what seems to be a free service. The next section discusses the economics of the internet, which partially explain how Google has gotten so large and powerful. This includes discussions about network effects and economies of scale present in digital markets. The next section goes through the allegations of the firms harmed by Google’s conduct.

Chapter Three elucidates the theories necessary for an understanding of the market

Google operates in. First, this thesis looks at the standard of harm, exploring how to maintain the

2 balance of investment incentives for firms, while protecting competition. There is a discussion about whether the Total Surplus Standard is appropriate in reaching our desired results, and the prevalence of Type I and Type II errors1, and their impact on decisions made by competition agencies. Next, there is a discussion about the economics of both horizontal and vertical search engines and their respective roles in how users search for information online. Following this, this thesis explores the literature on barriers to entry and economies of scale, which are important in

Google’s market. After this, there is a discussion of the relatively new phenomenon of network effects and its effect on the industry which Google resides in.

While there are many characterizations of Google’s conduct, the first section of Chapter

Three will discuss whether the characterization of Google’s conduct matters and presents the idea that all three of these conducts fall under the Raising Rivals Cost’s literature. The Raising

Rivals Cost’s literature focuses primarily on the effects of the conduct, as opposed to focusing on which abuse the conduct falls into. Legally, one can think of the conduct in several ways. One characterization of the conduct is tying. Tying is where a firm has two goods in two separate product markets, but they condition the sale of one item on the sale of the second item. These ties can be done by selling the goods together, or by making the products only work with one each other. Another characterization of Google’s conduct is discrimination, where Google discriminates against comparison shopping services by promoting Google Shopping. The final characterization of Google’s conduct is refusal to deal, where Google is refusing to supply traffic to comparison shopping services, which then reduces the ability of them to compete against

Google Shopping.

1 A Type I error is when a competition authority condemns a conduct which is pro-competitive, while a Type II error is when a competition authority fails to find an anticompetitive conduct guilty

3 After the theory of the case, Chapter Four is a summary of the document provided by the

EC on its decision to fine Google. The EC goes through each of its assumptions and justifications for its explanation. In this section, a distinct contrast will emerge between the facts and theories put forward by the EC and this thesis.

Chapter Five starts by stating the facts found in this thesis with regards to barriers to entry and network effects as they exist in search engines. Next, Chapter Five explains why the preferred relevant market definition consists of general search engines and a market called retail searches, which will be expanded on later. This thesis agrees with the dominant nature of

Google’s general search, and that general search consists of its own distinct market. However, this thesis disagrees with other authors when it comes to who Google Shopping and comparison shopping services compete with. Rather than the market only including Google Shopping and comparison shopping services, this thesis argues that merchants and retailers such as Amazon and eBay should be included in the market.

Next, there is an economic discussion about Raising Rivals Cost’s and whether Google’s conduct was able to sufficiently raise its rivals' costs of business, by which Google could then better exercise its market power. This thesis concludes that Google’s conduct could not have created market power for three reasons. First, Google’s rivals have ample alternative ways to attract traffic, such as advertising and the use of mobile applications. Google should not be said to control market power in traffic to auxiliary websites. Since there are other ways which comparison shopping services can draw traffic to themselves, outside of Google’s general search results, it cannot be said that Google’s traffic is necessary to the operation of comparison shopping services.

4 Second, even if Google was able to force comparison shopping services out of business because the conduct made them unable to operate, there exist other services which can restrain

Google’s exercise of market power. With the market definition defined in this thesis, one can see that if there were no comparison shopping services, Google would not be able to raise prices to advertisers, because they could switch to Amazon or another online merchant or retailer. There is no locking in of merchants on Google’s ad network since there exist other legitimate places to advertise retail goods other than Google Shopping. Merchants such as Amazon were not harmed by Google’s conduct because they attract users directly to them, therefore even if Google’s conduct were able to knock comparison shopping services out of the market, there exist other constraints on Google’s market power. If advertisers should not see an increase in the advertising fees because of the conduct, then consumers should not be harmed by the conduct.

Finally, Google’s decision to promote Google Shopping should be seen as an exercise of market power, not conduct which creates market power. Google’s market power comes from its superior general search engine, and Google chooses to exercise its market power by extracting the maximum value downstream, whether that means fulfilling downstream retail searches themselves or allowing others to do it. Google’s conduct cannot be said to create market power in retail searches because its market power comes from its superior “upstream input” general search. Therefore, this thesis concludes that regardless of which classification the conduct falls under, the conduct is unable to harm consumers, by way of denying entry upstream, or weakening existing competition upstream, and that Google’s conduct is an exercise of market power as opposed to conduct which creates market power.

Chapter Five then goes into a discussion of the three potential characterizations of

Google’s conduct, and whether the conduct meets the legal thresholds outlined in the law and

5 past cases. This thesis will argue that from a legal perspective, Google’s conduct should not be considered any of the three conducts, as the conditions present in the relevant market do not match the provisions set out by the law or the jurisprudence.

Chapter Five continues with a discussion of the differences found between the EC and this thesis and follows with general comments on the EC decision. Next, Chapter Five questions whether the decision found by the EC could be deemed legally justified, even if it cannot be economically justified. There are peculiarities in European competition law which could explain the choice to fine Google, even if the conduct was unable to harm consumers. The final section then discusses what is next for Google and the EC, as the facts found in the decision align themselves closely with the realities of other similar industries, such as travel search and local search. Since Google’s conduct was deemed anticompetitive in this case, the same facts can permeate into these other cases. Thus, if the decision is wrong in Google Shopping because of the facts found by the EC, then the chance of getting it wrong in similar Google cases is very high.

Lastly, Chapter Six follows with the conclusion.

6 Chapter 2: Facts of the Case

2.1 Chapter Introduction

This Chapter first discusses how Google attempts to catalogue the internet, and how its

PageRank algorithm chooses how pages are displayed on Google’s search engine results page.

This Chapter then discusses Google’s advertising business, and how it can use personalized search results to pinpoint specific users for targeted advertising. Finally, there is some discussion about how the actual bidding process works for Google’s advertisements.

In the next section, there is a brief discussion of the unique economics of Google, and other New Economy firms. There is brief discussions of economies of scale and network effects, and how these phenomena tend to push industries into monopolies. There is also a discussion of the effect data has on competition in these firms, and how data could interact with both economies of scale and network effects.

Next, there is a section on vertical search engines, which starts by explaining the differences between general search engines and vertical search engines. There is then a dialogue about when users might prefer to use vertical search engines instead of general search engines, and how advertisers may prefer to advertise with vertical search engines as opposed to general search engines.

Finally, there is a overview of the Google’s accusations from its rivals. Through both algorithm updates and the promotion of Google Shopping, Google’s rivals accuse Google of attempting to monopolize the downstream industry of comparison shopping services. Both conducts limited the traffic that comparison shopping services received, which sapped them of a large portion of traffic which they received altogether, which then caused a large decline in

7 traffic to comparison shopping services. There is then a quick overview of the cases by Canada’s

Competition Bureau, and the United States Federal Trade Commission (FTC).

2.2 How Google Works

Google’s search engine starts with grunt work which happens before queries are input into Google, the crawling and indexing of most of the web pages on the internet. There is then the interpretation and analysis of the inputted query done by Google’s search algorithm which then gives the user their personalized search engine result page. Somewhat abstract from this process is where Google makes its money to keep afloat since organic (non-paid) search results do not pay Google’s bills. To get a well-rounded view of Google, one must also look at their advertising business. Google uses content, like a newspaper, to attract eyeballs, which it then sells said eyeballs to advertisers (Ratliff and Rubinfeld, 2014, p. 5). The content provider,

Google, in this case, can pay for providing the content for users by using the revenue from selling advertising (Ratliff and Rubinfeld, 2010).

How does Google crawls and indexes pages? Google sends out software which goes to web pages and scans the web page for keywords and links to other pages. After scanning the page, the crawler follows any links to other pages found on the internet (Google, 2010). These crawlers repeat this step for all the interconnected pages, creating an extensive web of interconnected links, hence the colloquial term spiders for these crawlers. These crawlers send this data back to Google, which it then indexes and records which web pages are connected and keywords found on each page.

Once this grunt work is out of the way, Google is ready for users to input queries. When a user inputs a query, Google looks through its index of the internet for those keywords and pulls

8 those out first (Google, 2010). For some very general searches, there can be millions of results which can be triggered by just the simple keyword, so how does Google know which one the user likely wants to see? They continue to ask questions of their index to measure the relevance of any one web page. They ask questions like, how many times is the keyword referenced on this web page? Is the keyword in the title of the page, or the URL? Google also uses its patented

PageRank system to determine the quality of websites. They determine this quality by looking at the number of web pages that link to this website, as well as how many links to external pages are on its website (Sullivan, 2016). For example, many websites will link to the New York Times website if they have a breaking story, but they will likely only have a couple of external links towards sources or additional information. These parameters would lead to a very high PageRank score for the New York Times. On the other hand, if many other pages do not have many external links pointing towards their website, they likely are not very popular, which means they will likely be irrelevant to most searches.

Additionally, Google has released several different updates to their algorithm to try and combat websites which were trying to game PageRank into giving them a higher score than they truly deserved. Two updates, which are central to the case were the Big Daddy update, and the

Panda update, which will be expanded on in the next section. Both of these updates were targeting websites which purposely had many inbound and outbound links to make themselves look popular to PageRank, and then had advertising space which will be seen by users who click on their link expecting something relevant to their search.

As demonstrated with these “link farm” pages, websites see the inherent value of being highly ranked by Google and continually try to end up at the top, sometimes by less than legitimate ways. Since much of internet traffic goes through search engines, websites can employ

9 Search Engine Optimizers (SEO) to try and boost their page scores for a quick boost in relevance. From Google’s perspective, they would rather see pages end up at the top of the search engine results page because they are genuinely relevant, as opposed to cheating the system in some way, as this damages Google’s reputation with bad results. Therefore, to combat this, Google is continually changing its algorithm to ensure that the most relevant pages float to the top. An apt analogy is one of a game of cat and mouse as SEO’s find a hole to exploit which prompts an algorithm change by Google, and so on (Levinson and Salinger, 2015, p. 36).

Another vital business strategy employed by Google in its search engine is the use of personalized search results. When users use the same computer and browser when making searches on Google, Google keeps tabs on what users pick for individual queries, and learns what the user wants to see from specific searches. For example, if a user always searches “email” and chooses AOL mail every time, Google can discern that this user wants to see AOL mail at the top of their searches for “email” (Teevan, Dumais, and Horvitz, 2010). Google will not display AOL mail at the top of all searches for “email” after learning this; only for the intended user. With more and more users coming online in the last couple years, more of which are using

Google Search as their primary search engine, Google has continued to amass vast swathes of data which it can use to better predict what users are looking for based on context and phrasing of their queries. Not only have they been better at getting data, but users are also making it easier for Google to identify them. For example, when using users can log in to an account, and their information can be shared across many of Google’s services so that users only have to log in once to have access to their , YouTube, and other accounts. Implicitly, what users do here is link what they do and look at across several separate websites in exchange for easier navigation and time-saving. This data is invaluable for search engines as with more data

10 about its users and their preferences; they can give users what they want more consistently, making their service more valuable and ensuring return users.

This data could be invaluable for advertisers. In the past, the main issue for advertisers was the need to advertise widely to reach the audience they wanted to reach (Pope, 1983). With this information, advertisers can spend less money to reach a more engaged and interested audience, while hopefully spending less advertising dollars (Evans, 2009, p. 381). Not only can advertisers find a more engaged and interested audience with their advertisements, but they can also target users who are using the search engine for commercial reasons. For example, if someone searches “Canon camera” and opens several links to retailers that sell the camera, the search engine can conclude that this user is looking to purchase a Canon camera. Alongside this immediate clue, browsers and search engines can collect cookies its users produce when interacting with a website to record what they are interested in as well. This behaviour advertising/targeting is done by collecting data on online purchasing behaviour and correlating it with user behaviour so that advertisers can focus their advertising at the time where users are most likely to make their purchasing decision (Long, 2007). Röhle (2007) went on to say that this type of advertising requires a great deal of data on its users and, therefore, behavioural advertising can only be done by websites that have a large dataset of its users. So while all websites would love to offer this type of advertising for their advertisers, not all websites attract enough users to amass the data required to provide this, putting them at a disadvantage compared to their potentially large incumbent rivals.

Finally, this section will talk about Google’s advertising business. Google’s primary source of revenue has been online advertising since its beginning, and it continues to grow as more users come online (Rosenberg, 2019). As discussed before, Google attracts eyeballs to its

11 search engine by offering search results free of charge to its users. In exchange for the search results, advertisements are displayed for users, as well as allowing Google to use users’ data for improvement of Google. Advertisements show up in several ways online, the primary way for

Google is search advertising which looks like an organic search link at the top of the page, with the word “Ad” or “Sponsored” beneath it. Google also has its AdSense system which buys advertising space from publishers and resells it to advertisers, which is called contextual advertising. Since their purchase of DoubleClick, they have also gotten into graphic advertising, which is software which lets publishers and advertisers manage their advertising inventory and produce ads which show up as banners on webpages (Hahn and Singer, 2008).

Search ads are the most critical subset of advertising for Google, and they may be generated by a particular keyword inputted into a query, or they may be shown based on past searches and past websites the user has visited (Google, n.d.). These advertisements come from the fact that advertisers will want to try and boost their sales by showing their product/service when users search for a certain word or phrase. For example, when someone searches for

“television” Sony or LG may have chosen to bid on that keyword, and they are shown in prominent positions with the word “Sponsored” or “Ad” underneath it. The placement for these advertisements depends on the result of an auction for specific keywords. A bid is how much a business is willing to pay Google when their ad is clicked on, hence the Cost-Per-Click name given to this form of advertising. All else equal bidding higher for a keyword will increase the chance of a prominent placement, but advertisements are also subject to the same quality scores that organic search results are. Quality scores for advertisements are essential because if a business bids high for a broad swath of keywords, some of which may be utterly unrelated to their business, this implicitly hurts Google (Manne and Wright, 2010, p. 23). For example, if

12 someone searches for “Coca-Cola,” they are looking for information on specifically Coca-Cola.

Without quality scores, Pepsi would be able to bid high on a rival cola’s keyword to be seen first, which deteriorates Google’s search engine results by having irrelevant results showing up.

Therefore, for any keyword users search for several metrics decide whether or not an advertisement will show up, and which one will receive the prominent position.

One last point to touch on with Google’s advertising business is the actual bidding process. The auction process for keywords is a Generalized Second Price (GSP) Auction. As described by Edelman, Ostrovsky, and Schwarz (2005), advertisers decide what keywords to show their product or service, and how much they are willing to pay to be placed with those keywords. All else equal, the advertiser that pledges the highest amount will show up as the first sponsored link, and so on. If the user clicks the sponsored link, the advertiser then pays Google an amount equal to the next highest bid (Edelman et al., 2005).

In this section, this thesis illuminates how Google’s general search engine works. While all general search engines crawl and index the internet, how Google set itself apart is how it displays results to users queries. Google has also decided to change how it shows results in response to bad actors who wished to game Google’s algorithm. Because users continue to come back to Google, they continue to collect information on its users, making its search engine better, along with the ability to provide personalized search results. The way they fund this is by way of advertising. By way of sponsored links at the top of some search results, as well as banner advertisements on websites through Google’s AdSense program, Google funds its operations through advertising.

2.3 Unique Economics of Google

13 In this section, this thesis will discuss the unique economics of Google, which is a New

Economy firm, and is characterized by high innovation rates and quick turnover. The introduction of the internet has led to the growth of industries which were unfathomable decades ago. One such industry which came from the internet was search engines. Like many New

Economy industries, Google operates as a free platform which tried to bring users in so they could sell advertising to them. Since most platforms could not charge for access to their content, without losing a majority of their users, they gave their content for free to attract users who would then look at advertisements.

Advertisers like to advertise in places where there will be lots of eyes watching, so the more popular the website, the more valuable the advertising was to advertisers. This creates a network effect in which more users makes a platform more valuable to advertisers. Network effects will be expanded on later, but the critical point to see is that these network effects only go in one direction for Google. Advertisers care about how many users a website has, but users are generally indifferent to the number of advertisers there are on a platform. In fact, in some cases, users would prefer fewer advertisers because sometimes the platform can be so saturated with advertisements it can be troublesome to use and navigate (Goldfarb and Tucker, 2011). While there are still economists who believe that Google operates as a two-sided platform,2 including economists at both the FTC and EC, this author remains unconvinced by their arguments. The main issue with describing Google as a two-sided platform is that the network externalities only work in one direction. Thus, it seems more likely that Google is a pure advertising firm, which uses consumers as an input to make advertising revenues, as described by Luchetta (2013).

2 See Evans and Schmalensee, (2016), where they conclude that there are direct network effects between Google’s users and advertisers. Users value content from websites, but do not want to see advertising. Websites want advertising revenue. Therefore, for users to be attracted to the website the content must outvalue the advertising. Although they call it a direct network effect, what they describe is more akin to usage externalities

14 Luchetta (2013) describes Google as a firm whose primary purpose is to sell advertising, and one of the inputs used to sell advertising is users. Therefore, since users would rather there not be advetisements, they act as an input to Google, which they use to sell advertising.

While there may not be network externalities from advertisers to users, there are likely network effects in Google’s collection and usage of user data. When a user searches with

Google, they presumably click a link which was offered by Google’s algorithm on the search results page. When the user clicks a link, they provide information to Google about what people generally are looking for when they input that given query, and what that specific person was looking for when searching said query. This information can be collected for every search done by users and goes to improve Google’s search engine. So, since Google at this time gets more searches than any other search engine, they have more data about their searcher's preferences, which they can then use to improve their search engine. Since Google is improved when they collect more data, this attracts more searches to their platform, which improves it further, and the cycle repeats ad infinitum. Hence, one can see why Google continues to grow and get more popular as the years go on. One caveat of this point is to ask what the relationship between search quality and data amassed is. If the relationship between data and search quality is diminishing marginally, each gigabyte of data Google collects makes Google’s search engine better than before, but less than the last gigabyte did, then there is less to worry about Google’s market power. Since a search engine with the relatively same amount of data can have relatively similar quality as Google with this relationship, then they should be able to constrain Google’s market power. However, if the relationship between data and search quality is increasing marginally, where each gigabyte improves Google’s search engine more than the last one, one would expect no search engine to rival Google with regards to search quality. Intuition points to

15 diminishing returns to data, but there are reasons as to why data may create economies of scope for firms which have vast swathes of data. For example, in Allen Grunes and Maurice Stucke’s

2016 book entitled, Big Data and Competition Policy, they argue that the diversification of services leads to better insights if data linkages are possible, which is to say, that with more services a firm provides the better each of the services are due to the “super-additive” insights of data linkages. Economists will recognize these as economies of scope, where the cost of producing two services within the same firm will be less than if produced by two different firms.

Another important and unique characteristic of internet platforms is the extent of their economies of scale. The most costly part of starting a platform on the internet is the backend work to try and get the website up and running, as well as some hardware servers. For example,

Google needed to create crawlers which crawl websites on the internet, buy servers to maintain their index of websites, and create their patented PageRank algorithm to rank websites based on queries. This initial investment, along with the minuscule cost of providing one more search leads to significant economies of scale (Manne and Wright, 2010, p. 40). If a search engine can amass vast swathes of users, then their average cost per search will be minuscule. This thesis will expand on scale economies in Chapter Three, but one thing to note is that markets which experience scale economies, as well as network effects, tend towards one dominant firm with several fringe firms perhaps nipping at the dominant firm's heels (Economides, 2006, p. 104).

These markets will often have one sizeable dominant firm which controls most of the market's profits. As stated by Economides (2006), there need not be any anticompetitive action for this conclusion to come about in this market, due to the presence of strong network effects and

16 economies of scale (p. 107). Furthermore, Economides (2006) goes on to say that free entry will likely not change market structure due to lock-in effects3 and status-quo bias (p. 108).

While this form of competition seems quite weak, it can be healthy. Instead of firms competing inside the market, they try to innovate and predict where the market goes next to capture the entire market. The market power of the dominant firm is often fleeting, they currently are the best at providing a service or product, but if they do not adapt to what is coming next, they may find themselves as the fringe when the market shifts next. So, while markets which experience network effects, substantial innovation, and economies of scale may appear unhealthy, they compete in different ways than firms in more traditional markets. A perfect example of this is Google and incumbent search engines like Yahoo! and AltaVista. As mentioned before, Google started as a small new entrant into this market and did not catch up to the incumbents until around 2006. While there are many reasons to point to for both Google’s rise and the incumbents fall, one sees that market positions can change if a current monopolist does not predict what is coming next, or ceases to innovate. Now that Google has a dominant position in general search, it may feel challenged by vertical search engines which specialize in travel, retail, and social media who could steal Google’s market power. Google must keep itself relevant to its users evolving tastes’ and desires to stay relevant and maintain its desired market position.

This section briefly commented on the peculiarities of the subsection of Google’s industry. The New Economy is characterized by the collection and use of large amounts of data to further increase the usefulness of their products. Firms also experience large economies of

3 Further Discussed in Section 3.5.

17 scale, whereby if they can hypothetically serve an infinite amount of customers, its average cost would be indistinguishable from zero. These firms also experience differing degrees of network effects, which can entrench an incumbent in the market. These firms often compete for the market instead of in the market, because the presence of network effects and economies of scale, as well as the rapid innovation, lend itself to shortlived dominant monopolies.

2.4 Vertical Search

This section discusses the subtle differences between a vertical search engine and a horizontal (general) search engine. A vertical search engine is a more specialized search engine than a horizontal search engine. They often specialize in a particular form of information or product; for example, there is Expedia for travel goods, Flickr for images, and Amazon for retail goods. Often vertical search engines rely on outside avenues to get users to them, like advertising or horizontal search engines. They require this because they will not be the first stop on the internet like horizontal search engines are, users do not start their search with them. If a user is starting to search for new shoes online, they will likely start by searching for shoes on a horizontal search engine, then open a couple of the links which appear on the search engine results page. They will likely next compare on prices, quality, and selection between all the vertical search engines which specialize in shoes, or more generally, retail goods. When searching on a general search engine, one would be bombarded with ancillary information, which may well be irrelevant to the user's query in their head, making the cost of using a general search engine higher than using a vertical search engine. One way to circumvent the horizontal search engine altogether is to advertise and try and become synonymous with a specific type of vertical search product and try to induce users to go right to the vertical search engine when they require specialized information.

18 Why do users use both horizontal and vertical search engines? As stated, horizontal search is helpful as a starting point for users of the internet who do not know who could provide the information they seek. Vertical search is for users who have a general idea of what they are looking for, but they seek a service which will parse and display more detailed information which the user seeks. Vertical search engines may also have access to propreitary databases which users may not be able to get anywhere else. One facet of vertical search which is helpful for advertisers is that users self-select into the specialized search engine, which gives advertisers a hint of the user's preferences. For example, if someone is searching on a golf club trading website, advertisers can feel reasonably safe assuming that they enjoy golfing, and they can use this information to advertise local golf courses. The fact that users give some indication of their preferences is precious to advertisers, as this information can make it easier to turn an advertisement into a sale. If someone is using a horizontal search engine and sees an advertisement for a golf course, there is no indication that they enjoy golfing and the advertisement is likely not very valuable to the user. This is analogous to traditional advertising on television, advertising on the Golf Channel is likely very different than advertising during

Primetime on NBC.

The point to get out of these examples is that in the absence of any previous proffered information on the user using the search engine, advertising on vertical search engines is likely to be considered more valuable for advertisers than horizontal search, because of limited use of horizontal search advertising (Blake, Nosko, and Tadelis, 2015, p. 3). Since a considerable amount of vertical search engines are related to retail, such as travel, computer parts, or cameras, there is a higher chance that if a user is using the service, they want to make a purchase. An anecdote from Chief Media Officer at Verizon, John Nitti aptly describes this phenomenon.

19 When asked why Verizon has expanded its advertising on Amazon he said, “They [Amazon] have people who are in a shopping mind-set, so that’s valuable for Verizon to be seen as a resource within that mind-set” (Cresswell, 2018). This anecdote is supported by the paper done by Blake, Nosko, and Tadelis (2015) which states, “While these [horizontal search] consumers may look like good targets for advertising campaigns, they are also the types of consumer that may already be informed about the advertiser’s product, making them less susceptible to informative advertising channels” (p. 2). That is to say that the people who click advertising links at the top of Google’s general search results page are likely users which in the absence of the advertising link, would have found their way to the advertised webpage regardless of the advertisement. Therefore, the efficacy of horizontal search engine advertising is likely overstated, especially for well-known brands, which represent the lions-share of online advertising revenues, which may partly explain the increase in advertising on vertical search engines such as Amazon.

The primary differences between vertical and horizontal search are then the scenarios in which users tend to use each, as well as the value of advertising on each platform due to the intentions of the user when using the service. Vertical search engines must also worry about directing users to their service, as they are less likely to be the user’s first choice of search engine due to the limited scope of their use.

2.5 Overview of the Allegations Against Google

This section will focus on Google’s conduct and its alleged harm. A UK comparison search engine called Foundem was one of the complainants of Google in the European Union.

Foundem was a vertical search engine, a search engine which focused primarily on a particular subset of information. For example, Expedia is a vertical search engine for travel goods, or

20 Amazon is a vertical search engine for retail goods. In the owner's own words, “[Foundem] was one of the worlds broadest vertical search services, covering product-price-comparison, travel search, jobs search, and property search” (Raff and Raff, 2018, p. 3).

In 2006 Google released an update to their search engine which was intended to combat spam websites. Spam websites had no other goal than attracting eyeballs so the owners could collect advertising revenue. They were filled with links to other websites, which tricked Google into thinking it was a relevant search result. This update targeted websites which primarily hosted other websites’ data, a characteristic of spam websites (Arriola, 2017). However, caught in the crossfire of this update were vertical search engines like Foundem. Since price comparison services need up-to-date information about prices, they link to other pages and do not host their own data. This update caused Foundem to be excluded entirely from results, which was very detrimental to their business, as they received most of their business from

Google Search results (Raff and Raff, 2018, p. 4). This initial complaint was the start of a many- year battle with Google, trying to get Foundem back into Google’s search results.

This initial Big Daddy update in 2006 was compounded by another algorithm update called Panda in 2011. Panda, like Big Daddy, was meant to target spam and websites which tried to cheat by having an artificially high PageRank score. While the Big Daddy update was to try and target very apparent spam websites which were packed full of keywords and links to other spam websites, Panda was created to try and target sites which copied other sites content and sites with low amounts of original content (Sullivan, 2017). As expected, when Google foreclosed one type of spam service with its Big Daddy update, spam websites moved onto a new loophole, which then Google has to patch, which was the primary goal of the Panda update.

These “scraper” sites were the principal target of this update, and it is unsurprising that Foundem

21 and other price comparison services got caught in this new update again. For the same reasons as

mentioned while discussing the Big Daddy update, price comparison services do not host their

own content, they scrape other websites prices, and act as an intermediary of this information.

So, for the same reasons as before Foundem’s PageRank was demoted, and they had to plead

with Google to whitelist (make a manual exception) for Foundem to show up in Google’s search

results.

Google launched Google Shopping in 2002, under the name Froogle. By 2007 Google

changed the name to Google Product Search, which would yet again be changed to Google

Shopping in 2012. When Foundem found their businesses PageRank score decimated by

Google’s 2006 Big Daddy update, they also started to see Google’s Google Shopping shown in a

prominent position on the search engine results page due to Google’s rollout of Universal Search

in 2007. Universal Search was designed by Google to make use of their specialized search

engines which were rarely used because they were somewhat hidden on Google’s home page. At

the time of the release of Universal Search, Google had over a dozen vertical specialized search

engines, from to . Before the update, these vertical search engines

were barely used because they were treated as separate silos separate from Google’s general

search engine and had to be sought out on purpose, said then Vice President of Search Products

and User Experience at Google, (Sullivan, 2016). By rolling these vertical search

engines together, Google could hopefully better answer user’s questions that before would have

been answered by going to some other webpage.4 When a user inputs a query to Google, the

4 See Universal search: The best answer is still the best answer. (2007, May 16). Retrieved from https://googleblog.blogspot.com/2007/05/universal-search-best-answer-is-still.html. (“[W]e're attempting to break down the walls that traditionally separated our various search properties and integrate the vast amounts of information available into one simple set of search results.”)

22 assumption is that users like to find the information as quickly as possible, and by ensuring that

Google can answer more questions from their services embedded in the search engine results

page, this makes users better off.

At this point, the basis for Foundem’s complaints against Google are starting to come

together. Google had deteriorated Foundem’s PageRank score, making them virtually invisible

for users using Google Search, one of Foundem’s primary sources of traffic. In the meantime,

Google was also making sure that Google’s service Google Shopping had prominent positions on

search results which Foundem would have been featured on absent these algorithmic changes.

The combination of Google’s conduct and their inaction in getting Foundem reinstated in

Google’s search engine results page lead the two founders of Foundem to meet with the EC

about Google’s potential violation of Article 102 of the Treaty on the Functioning of the

European Union (TFEU). Article 102 is the EC antitrust legislation that deals with abuse of

dominance. Soon after the opening of the case in the European Union, the case was picked up by

the United States’ FTC and Canada’s Competition Bureau in 2011 and 2013 respectively.

In both the United States and Canada, these cases were relatively short-lived, with

Google’s acquittal happening in 2013 in the United States, and 2016 in Canada. Both

competition authorities decided that Google’s use of its algorithm to promote Google services,

such as Google Maps and Google Shopping, was a competitive choice which was pursued to

benefit Google’s users primarily, and therefore was efficiency enhancing.5 Since both of these

5 See United States, Federal Trade Commission. (2013). Statement of the Federal Trade Commission Regarding Google’s Search Practices. and Canada, Competition Bureau. (2016). Competition Bureau statement regarding its investigation into alleged anticompetitive conduct by Google.

23 agencies decided that the conduct was not anticompetitive, that was the end of the analysis, unlike the EC’s case against Google.

This section discussed what exactly Google did to draw complaints of anticompetitive behaviour. Google both promoted its own service, Google Shopping while demoting competing comparison shopping services. This conduct was alleged to suck traffic away from competing comparison shopping services, limiting the ability of them to attract advertising revenues, reducing the ability of them to compete against Google Shopping. Both Canada and the United

States did not deem this conduct anticompetitive, while the EC decided that this was anticompetitive.

24 Chapter 3: Case Theories

3.1 Chapter Introduction

This chapter will discuss and explain all the relevant theories required for understanding this case. This chapter begins by discussing the theory behind defining a market. It then goes into an overview of what harms competition authorities should be worried about. There is a brief discussion of some economic and legal concepts such as a Pareto Improvement, a Potential

Pareto Improvement, the Total Surplus Standard, and Type I and Type II errors.

This thesis then goes into a brief overview of relevant barriers to entry in the case, and questions if Bain or Stigler were correct in their definitions of what constitutes a barrier to entry.

Next, there is a discussion of network effects, which is essential to understanding the case. Both indirect and direct network effects are explained, as both may be important in the case.

Next, there is a discussion about whether the characterization of the conduct matters to whether the conduct harms consumers. This thesis argues that the characterization of the conduct does not matter, and leads to misplaced arguments over characterizations of conduct as opposed to the critical question, which is, does the conduct harm consumers? The section continues by explaining scenarios where vertical foreclosure is an exercise of market power and situations where conduct can create or maintain market power. Finally, there is a discussion of the legal thresholds for the potential characterizations of Google’s conduct when it promoted Google

Shopping. Google’s conduct can be interpreted as tying, refusal to deal, or discrimination, and each carries different legal thresholds. As one will see, the characterization of the conduct can affect whether or not a firm is found guilty of anticompetitive conduct as each conduct has different necessary conditions for when it is anticompetitive.

25 3.2 Market Definition

This section will explain the way which economists determine the relevant economic market for a firm. The relevant market definition would be found using the Hypothetical

Monopolist Test (HMT). The HMT asks what geographic and products does a firm have to control to implement a profit-maximizing Small but Substantial Non-Transitory Increase in Price

(SSNIP). Economists do this test to figure out where consumers will substitute when the price of an individual good goes up, therefore signifying the closest substitutes to the good in question.

Implicitly, this answers what goods the firm would have to own, in a particular geographic area for them to be able to increase their price and have it be profit-maximizing. Economists would do this by implementing a SSNIP and check whether or not this increase is profit-maximizing for the firm. Small but Substantial usually refers to a five or ten percent increase in price, and Non-

Transitory usually refers to a year. If the SSNIP is not profit-maximizing, then it implies that one must bring another good or another geographic location into the fold and try the test again. Once the SSNIP is profit-maximizing for the firm, all of the included goods and geographic locations make up the relevant antitrust market for the firm in question. Once a market is defined, one can conclude whether or not a firm has market power in the relevant market, by analyzing the presence of barriers to entry and current market conditions, including the actual substitutes available to consumers, and the possibility of potential entry. Because, if there is no market power in the relevant market, there can be no successful anticompetitive conduct.

When specifically looking at abuse of dominance, the HMT is a test of demand-side substitution, whether consumers are willing to substitute to a relatively similar good if the price goes up, and less so supply-side substitution. This distinction should be made because if a firm is dominant, it likely does not have many firms offering the same good, which is where supply-side

26 substitution comes from. The possibility of demand-side substitution depends on several different parameters, as discussed in Canada’s Competition Bureau’s Abuse of Dominance

Guidelines; the views, strategies, behaviour, and identity of buyers and sellers or other market participants, end-use or functional interchangeability, physical and technical characteristics, switching costs, and price relationship or relative price (Competition Bureau, 2019). If users are very willing to switch to a similar product in the face of price increases, then one would say that there exists plenty of demand-side substitution.

This section laid the groundwork for determining a relevant market for an abuse of dominance case. The first step of an abuse of dominance is to consider whether the firm is dominant in the relevant market. If there is no dominance, then one cannot abuse it. The way to find the relevant market would be to use the HMT to see where consumers would substitute when a firm exercises market power. If a hypothetical monopolist does not find it profit- maximizing to execute a SSNIP, then add another product or geographic market into the fold, and repeat until the monopolist finds it profit-maximizing to implement a SSNIP.

3.3 Standard for Harm

This section will explain how economists demonstrate harm which can come about from monopolists and anticompetitive behaviour, as well as quantifications of this harm. For a competition authority to bring a case against Google they must show that Google’s conduct harmed consumers and not just rivals. While it is possible and likely that by harming rivals, the dominant firm can be limiting the ability of rival firms to constrain their market power, and therefore lead to higher prices for consumers, this could also simply be regular competition.

Rival firms will complain about anything a rival firm does if it reduces their profits, even if what the firm did was something that differentiated their product or created surplus for consumers

27 because of the superior nature of the product. Therefore there needs to be a coherent connection between the harm caused to rivals, and the harm caused to consumers, and if there is not a connection, then the case should be not be pursued.

Not only should competition authorities be looking at the presence of a connection between harm done to rivals and harm to consumers, but they should also be weighing the benefits and costs of the conduct as well. When conduct is engaged in, or policy is enacted, which makes everyone better off and no one worse off, that is called a Pareto Improvement.

However, Pareto Improvements are unlikely to come about in the real world. Therefore economists call a change that makes someone better off, and someone worse off, but less so than the person who is better off, a Potential Pareto Improvement. This idea that society should maximize the entire pie initially, and then squabble over the distribution after the fact is an application of what Robert Bork called, the Consumer Welfare Standard in his 1978 book The

Antitrust Paradox (Bork, 1978). For anyone acquainted with economics, consumer welfare has a specific definition, one which does not align with the idea that Bork had when naming the

Consumer Welfare Standard. If society is maximizing the size of the pie, then the producer surplus should be included too, so the Consumer Welfare Standard should be more aptly named the Total Surplus Standard today. Total surplus is maximized when the marginal firm's willingness to supply and the marginal consumer’s willingness to pay is equal. This is the outcome which maximizes societies wellbeing, and competition authorities should aim for this outcome. When firms create or maintain market power, the market strays from this preferred outcome, as there will be some consumers who would be willing to purchase the good at the marginal cost, but because the price is too high, the trade does not happen. Therefore, competition authorities should aim to maximize total surplus.

28 One worry which is ever-present in competition policy is that under the Total Surplus

Standard, there is the possibility of large transfers from consumers to producers, which still increase total surplus. Some have argued that consumer surplus should be weighted more than producer surplus because competition policy is intended to protect consumers primarily.

However, this may lead to under-investment by firms because they do not receive enough revenue to cover their long-term costs, which will end up harming consumers as producers are no longer providing the best product as they could. Therefore, if a conduct or merger is a Potential

Pareto Improvement, then it should be allowed to happen under the Total Surplus Standard.

However, there could be some form of redistribution post-conduct which can make those who are harmed as well off as before, especially if the group which is harmed consists of consumers.

One way which consumers could also be harmed is when a firms conduct creates or maintains market power.

One important factor to include when thinking about the correct measure of harm is the simplicity of the measure and the likelihood of Type I (false positive) and Type II (false negative) errors. With the complexity of analyzing antitrust cases, there is bound to be mistakes made by economists and judges alike. However, as pointed out by Judge Easterbrook in his 1984 paper entitled The Limits of Antitrust, society should be more concerned with Type I errors than

Type II errors, especially with markets which experience high amounts of innovation. If a firm is lambasted for behaviour later considered to be pro-competitive, then other firms will see this and not engage in that activity, even if it was pro-competitive. This chilling of innovation or business strategy will likely cause dynamic harm down the line, whereas Type II errors will likely be corrected in highly innovative industries by the consistent shifting of the market. In high tech industries, markets are created and rendered obsolete rapidly, and, market power will likely be

29 eliminated as the market shifts. Therefore, it makes sense that a standard is chosen which is the easiest to implement, leading to fewer Type I and Type II errors, but also err on the side of caution and ensure that courts do not falsely claim that a conduct is anticompetitive. So, courts should use the Total Surplus Standard (Consumer Welfare Standard) due to its relative ease to apply, and also because they can err on the side of innovation, which will ensure minimal Type

II errors (Easterbrook, 1984).

3.4 Barriers to Entry and Economies of Scale

Definitions of entry barriers and classifications of what constitutes a barrier to entry have been continually evolving over the past sixty years. This analysis started with Joe Bain in 1956, where he investigated the cigarette and steel industry as well as the manufacturing industry. He was mainly concerned with why pricing may not return to competitive levels and how firms earn supracompetitive returns in the longer run. One of the primary ways that prices would settle down towards competitive levels would be the entry of new firms, which led to his investigations as to how firms reduced entry by setting up barriers to entry. From his investigations, he made several observations about entry barriers and made some conclusions on what he constituted a barrier to entry. Bain mostly pointed to cost advantages, substantial capital requirements, product differentiation, and economies of scale as primary barriers to entry. Thus, his definition of a barrier to entry is “anything that allows incumbent firms to earn above-normal profits without the threat of entry” (Bain, 1956).

Bain’s work leads to work done by Stigler and Demsetz in the 1960s, which builds upon some of Bain’s contributions and disputes other conclusions. Stigler, in his 1968 book entitled

The Organization of Industry, clarified his definition of a barrier to entry as “a cost advantage enjoyed by an incumbent but not by entrants.” Carlton (2005) gives an example where, if an

30 incumbent has to spend one million dollars annually to maintain their reputation, then as long as entrants could replicate this, there should be no asymmetries in the market post-entry (p. 8).

Other examples of structural barriers to entry that can work alone to slow the entry of new competitors into a market would be the presence of switching costs, brand loyalty, absolute cost advantages, and asset specificity (McAfee, Mialon, and Williams, 2004, p. 20). Brand loyalty and switching costs can sometimes lock in users of a particular product because the cost of learning a new product or purchasing the new product can be higher than the increase in welfare from using the new product. For example, one of the barriers to entry in United States v.

Microsoft Corp., (2001) was that users had spent so much time learning the Windows operating system, and the cost to switch to a superior operating system was often more significant than the benefit of a marginally superior operating system (Department of Justice, 2001). This effect locked users into using the Windows operating system and left little room for competitors to enter.

There are also strategies which can be employed by firms when they expect entry in their market. With sequential entry, one can influence what an entrant will do based on the incumbents current level of investment. For example, from Fudenberg and Tirole’s 1984 paper entitled, “The Fat-Cat Effect, the Puppy-Dog Ploy, and the Lean and Hungry Look,” one sees that depending on whether the firms are competing over price (Bertrand) or over quantity

(Cournot) incumbents will make different decisions regarding how much to invest in R&D, and whether to try and foreclose entry. One strategy which is relevant for New Economy firms is the

“Top Dog” strategy whereby the first-mover enters at a massive scale, thereby over-investing upon entering, forcing the entrant in the next period to either match or stay out. However, with economies of scale in the mix, if the entrant does decide to enter, then neither firm will be

31 profitable at the symmetric Cournot level, because they both get too little of the market. The incumbent knows this and exploits the fact that the entrant knows this also, and overinvests to foreclose entry. What the incumbent is then saying is “feel free to enter, however, neither of use will be profitable, so you should probably not enter.” Baker (2003) continues this train of thought, stating that a low-cost incumbent could erect a permanent barrier to entry, maintaining prices above the competitive level, but not high enough where an entrant would find it profitable to enter. Therefore, there exists a first-mover advantage where a market experiences economies of scale and the firms compete over quantity, which can serve as a strategic barrier to entry, which also contradict Stigler’s definition of a barrier to entry.

Since Bain’s time, economists have evolved what constitutes a barrier to entry, as well as explored how each of these barriers interacts with one another. Originally, barriers to entry were defined by Bain (1956) as “anything that allows incumbent firms to earn above-normal profits without the threat of entry.” This definition was adapted by Stigler (1968), who claimed that a barrier to entry was, “a cost advantage enjoyed by an incumbent but not by entrants.” However, as stated by Baker (2003), Stigler won the battle, but Bain won the war for the definition of barriers to entry, as courts acknowledged that absent any Stiglerian barriers to entry, entry should happen, but they also acknowledged that with economies of scale, high sunk expenditures could act as a barrier to entry, contrary to Stigler’s definition. When there exist economies of scale, the incumbent likely will have lower costs than that of the entrant, meaning that the incumbent can profitably raise prices above competitive levels, but keep them below a level such that an entrant could profitably enter, and maintain its market power because it was first in the market. But, if there are constant economies of scale, this strategy will not work as an entrant will be able to enter on even the smallest scale profitably. Since economies of scale exist plentifully in New

32 Economy firms, one can conclude that significant capital investments can serve as a barrier to entry, alongside brand effects and switching costs.

3.5 Network Effects and Externalities

Although network effects were known to exist at the start of the 20th century, it was not until 1980 where Robert Metcalfe presented the idea for the effect between communication devices (Simeonov, 2006). A network effect is when users gain utility from a network based on the number of other users using the network.6 There exist subsets of network effects, the first of which being direct network effects. An example of this is social media today. Consumers use

Facebook because of the number of friends who are on Facebook. Users derive pleasure from having so many of their friends on that network and would be less likely to switch to a social media that has less of their friends on it. The second subset of network effects is indirect network effects, which work similarly but are more subtle. Indirect network effects often arise from two- sided markets where one side of the market sees value in someone joining the other side, as well as implicit value from someone joining their side because it incentivizes more people to join the other side. One example being eBay, where if there are more sellers, more buyers will come, attracting more sellers, and so on. When talking about network effects in the New Economy, one should also mention the positive feedback effects and how they work, as well as tipping, which is how firms often compete in this New Economy.

6 This section’s analysis is based on past research on network externalities done by Church and Gandal, (1993), Church and Ware (1998b), Farrell and Saloner (1985; 1986), Katz and Shapiro, (1985; 1994), Saloner (1990), and Shapiro (1996)

33 Direct network effects are relatively easy to see and explain as, as suggested by the name, they operate directly between users on a network. This effect was originally what executives at the Bell Telephone company saw.7 Theodore Vail, Bell Telephone’s then Chairman, stated that

A telephone—without a connection on the other end of the line—is not a toy or a scientific instrument. It is one of the most useless things in the world. Its value depends on the connection with other telephone[s]—and increases with the number of connections. (Vail, 1908, p. 22)

This acknowledgment that a single phone is useless, and that the value of telephones was increasing with more connected users is the primary conclusion of network effects. During the time Vail was writing about this, network effects were important for only a few industries.

However, with the New Economy, network effects are present in most newly-created markets, and the exploitation of these effects is an essential part of a successful New Economy firm. This has led to much more exploration and understanding of network effects in the past two decades.

While direct network effects are straightforward to see, indirect network effects are more elusive. The best way to explain it is with an example. Operating systems use applications which consumers use to make their user experience better. Operating systems become more valuable with the number of applications which are compatible with it. Application programmers have to choose which operating systems to make their application for. They will choose the operating system which will guarantee them the most users, as well as how many current applications exist since the future user levels depend on current application levels (assuming it is costly to port applications from one operating system to another) (Church, Gandal, and Krause, 2008).

Therefore, the usefulness of an operating system depends on the number of applications, and the

7 See Theodore Vail’s annual report to shareholders, 1908 https://beatriceco.com/bti/porticus/bell/pdf/1908ATTar_Complete.pdf

34 desirability of an operating system for programmers depends on the number of users using the operating system. There is an indirect link between the number of applications and the number of users in this network. Indirect network effects often occur in goods which act as components in a system. They also create a scenario where no one wants to be the first one to join a network in case they choose the wrong system and have to start from scratch on the other more popular system. In the operating system/application example, this chicken-and-egg situation creates an incentive for operating systems to pay for application programmers to write applications for their operating system, to make their operating system more desirable for users, which will, in turn, bring in more applications, which will bring more users, and so on (Gilbert and Katz, 2001, p.

28).

A conclusion that comes from the existence of network effects, whether direct or indirect, is that there exists a positive feedback externality which is only realized if more users join. The more users that join a network, the more valuable the network is seen to be for other potential users, and when these users join, the network looks more valuable for the next group of adapters.

Therefore, consumer expectations of the user base affect the number of users today, as well as future users. The importance of consumer expectations also compounds the incentive for firms to try and incentivize early adopters of their platform by offering lower prices or even paying them to use their product to give other interested consumers an indication that their product will become the standard. Tipping happens when an entrant has stronger network effects than the incumbent, causing a flipping of the market towards the entrant.

With the presence of strong enough network effects and these positive feedback effects, it seems likely that one product becomes the standard which all the users flock to. If the network effects are strong enough, then competition in the market may not tend to exist, and competition

35 instead comes from competition for the market. As described by Economides (2006), there tends to be a significant gap in the market share of the industry leader and the second largest firm in the industry (p. 104). He points to an example discussed before, operating systems as an example of why this happens. The leading operating system has the most applications compatible with it.

This makes it the most valuable operating system for users. Application programmers will make their new applications for the dominant operating system since they know making it for that operating system has the best chance of being sold and used. So, the dominant operating system continues to build its lead in complementary applications, which means that when new users come online, they will likely go for the operating system with the most applications, growing the dominant operating system’s user base. These two intertwined effects are an example of why in the New Economy, industries tend towards large dominant firms, which then leads to competition for the next technology, which will usurp the previous technology.

3.6 Theories of Self-Preferencing

This section starts with the general discussion of raising rivals costs, or reducing rivals revenues literature, of which tying, refusal to deal, and discrimination all fall under. Section 3.6.1 explains a simpler way to analyze anticompetitive conduct as proposed by Krattenmaker and

Salop (1986). Under this analysis, the concern is not about the characterization of the conduct, but the conducts effect on competition, and whether the conduct created or maintained market power. Next, sections 3.6.2, 3.6.3, and 3.6.4 discuss the legal backgrounds of discrimination, refusal to deal, and tying, respectively, and what it takes for a firm’s conduct to be considered anticompetitive in each jurisdiction.

3.6.1 The Economics of Raising Rivals Costs or Reducing Rivals Revenue.

36 Tying, refusal to deal, and discrimination fall under raising rivals’ costs or reducing rivals’ revenues literature. When a firm refuses to deal with a firm, or discriminates against a downstream purchaser, or ties a competitive product to a monopolistic product, the end result of the conduct is that it increases the cost or decreases the revenue of other firms in the market. This sentiment is echoed in Krattenmaker and Salop (1986) where they propose a two-step test to determine whether a conduct is anticompetitive, not by looking at the characterization of the conduct, but the result of the conduct. Krattenmaker and Salop (1986) propose that competition authorities first look at whether the conduct significantly raised rivals’ costs or reduced rivals’ revenues. Next, determine whether the conduct created market power because other firms cannot compete effectively. Rather than focusing on the characterization of the conduct for determining the legality of a practice, determine whether the conduct was anticompetitive from the outcome.

Predicating the legality or illegality of a conduct based on the characterization of the conduct puts pressure on defining the conduct, instead of analyzing the conduct’s effect on competition.

This norm lends an incentive for both plaintiffs and defendants to argue over the characterization of conduct while ignoring the more important question regarding the conduct’s effect. What

Krattenmaker and Salop (1986) argue is that the classification does not matter, what matters is the effect on competition.

Canada’s Competition Act does precisely this in its abuse of dominance provisions. In the provisions they say that Section 78 enumerates a non-exhaustive list of acts which are deemed to be anticompetitive in applying Section 79 (Competition Bureau, 2019). Section 79(1)(c) requires the conduct in question “has had, is having, or is likely to have the effect of preventing or lessening competition substantially in a market” (Competition Bureau, 2019). There is no mention of conduct and the conditions for them to be anticompetitive; the focus is on the effect.

37 Under this form of competition policy, the first step would be the same as before, determine whether the firm possesses market power in the upstream market. If the firm upstream has no market power, there is likely no significant increase in rivals cost because there are likely multiple reasonable substitutes for the downstream firms to switch to in response to the conduct.

However, if the firm upstream is found to have market power in the input, then the competition authorities should then skip classifying the conduct and go straight to the effect on competition.

If due to the conduct, the upstream firm eliminated the possibility of competition upstream, which either protected its market power or created more market power or softened existing competition upstream, then the conduct should be considered anticompetitive (Whinston, 1989).

Competition authorities should be careful to distinguish between conduct which creates, maintains, or extends market power and conduct which is an exercise of market power. In each jurisdiction, they say that currently possessing monopoly power is acceptable, and they want to find and prosecute conduct which creates or maintains market power. Therefore, conduct which is an exercise of market power is conduct which stems from the existing market power and is not conduct which creates or maintains market power.

Consider this hypothetical situation, the year is 1908, and Ford has just started producing its Model T’s. However, they are all painted black. A wealthy benefactor approaches Henry Ford and asks if it were possible to get a Model T in red. Henry Ford had not thought of this and could not offer this to the benefactor at this moment, so he forms a contract with someone else who can paint the Model T’s red, and resell them for a mark-up. Suppose after a couple of years of this arrangement, Ford becomes able to paint their Model T’s red themselves, and decides to stop the contract with the painter, and reap the benefits of painting the cars themselves. Naturally, the painter is angry and suggests that the refusal to supply cars to him created market power in car

38 painting services. But does this make sense? Nothing has changed in the production of Model

T’s; Ford still has market power in producing them. This is an example of a firm exercising its market power since they were still selling the cars to the contracted painter at the profit- maximizing monopoly price, all they decided was to integrate vertically and to create the value themselves instead of contract someone else to create value on their intellectual property. By

Ford refusing to contract the painter anymore, they did not create market power in the car painting industry, as there was no change in the upstream market which indicates a change in market power. Suppose for example, that instead of licensing the painter to paint the Model T’s red at the start, suppose that Ford decided to fulfill it in house from the start. If Ford had the capabilities to do this from day one, then it would be their prerogative to do so, since they were the one who created the Model T in the first place. Presumably, the reason Ford decided to license out their Model T’s in the hypothetical is that licensing out the painter was cheaper than providing it in-house at that time, or their quality surpassed what Ford could offer at that time.

One would be tempted to say that since there exists less competition downstream due to the discrimination, market power was created downstream. One could see this from the increased concentration downstream because of the conduct, which could lead to higher prices or lower quality downstream. However, this logic is faulty because this outcome stems from the fact that there exists market power upstream, and they exercise their market power by deciding how the downstream market operates (Church, 2018). There is no market power created downstream; an exercise of market power leads to conditions which appear to be the creation of market power.

Of course, this hypothetical could be made to look like a situation where market power is created. Instead, imagine a situation where Ford makes independent dealerships sign an agreement which stops them from selling any other cars than Fords. This agreement could

39 potentially stop smaller car companies from getting their foot in the market, thereby protecting

Ford’s market power in cars. In this situation, Ford used its market power to force dealerships to sign an agreement, which restricts potential competition, allowing Ford to continue its monopoly on cars. Moreover, if because the painters are pushed out of business, General Motors cannot provide the same red cars like Ford, then it also could be said that the conduct could have created market power.

Another way consumers could be harmed by the conduct of a monopolistic upstream firm is if the upstream and downstream products either have independent demand or are complementary to each other. First, if the upstream and downstream demands are separate, then consumers can be harmed by less choice in the downstream market for no benefit to them from the tie (Whinston, 1989). Secondly, if the upstream and downstream are complementary, then the upstream monopolist is not only reducing choice downstream, it is implicitly reducing the benefit of its monopolistic product (Carlton and Waldman, 1998). If either of these scenarios is true, then one can say that the conduct of the monopolist is anticompetitive because it harmed consumers for no pro-competitive reason.

Therefore, to simplify the analysis of conduct which may or may not be anticompetitive, competition authorities should take the steps suggested by Krattenmaker and Salop (1986), where it should first be determined if the firm has market power in the upstream or tying market.

If it is found they do, then look at the conduct, and if it significantly raised costs for rivals and that increase in cost made it so market power was created or protected in either the upstream or the downstream market, then the conduct should be found anticompetitive. This approach will lead to outcomes which will be simpler to analyze and will lead to fewer errors than condemning certain conducts if they meet specific conditions. Finally, competition authorities will have to

40 discern between conduct which is an exercise of market power, and conduct which creates maintains or extends market power, as the latter is conduct which they should focus.

3.6.2 Discrimination Theory.

The legal basis for the discrimination theory in European antitrust law comes from

Article 102(c) of the TFEU, whereby a dominant firm cannot apply “dissimilar conditions to equivalent transactions with other trading parties, thereby placing them at a competitive disadvantage” (European Commission, 2013). The legal basis for the discrimination theory in the

United States comes from the Robinson-Patman Act of 1936 which states, it is unlawful for a seller "to discriminate in price between different purchasers of commodities of like grade and quality" (Robinson-Patman Act, 1936). The act is focused on discrimination, and through case law, one sees that discrimination is interpreted to be a difference in price (Posner, 1987). Finally, in Canada, Section 50(1)(a) of the Competition Act prohibited a supplier from granting better prices or other monetary advantages to one purchaser, which are not available to competing purchasers, in respect to the sale of goods of like quantity and quality (Competition Act, 1985).

This provision has since been repealed, with the Competition Bureau directing firms to the abuse of dominance provisions for all classifications of conduct.

The Robinson-Patman Act and per se illegality of discrimination have been extensively critiqued in the United States by proponents of the Chicago School as protectionism for small business. Because the Robinson-Patman Act does not allow differences in price between firms, this means that quantity discounts could not be allowed, even when it reflected a real cost-

41 saving.8 Not only did Robinson-Patman disallow differences in price, which were reflected in differences in cost, but it also conditioned cartel pricing at the seller level (Posner, 1987). Since firms could not offer discounts for selling more, they kept their prices high, and would not try to undercut rivals for fear of litigation. Legal and economic experts in the United States do not see the Robinson-Patman Act in a very favourable light due to the problems it created, as well as the fact that the good parts of the act regarding predatory pricing were adequately covered by the

Sherman Act (Posner, 1987).

Alongside these concerns of cartelization and protecting inefficient competitors is the lack of economic understanding between the relationship between a dominant upstream provider and downstream providers. As discussed in the previous section, not all conduct by the monopolistic upstream provider, which harms downstream rivals also harms consumers by the creation or maintenance of market power. There are situations where the monopolistic upstream provider does hurt rivals, but the cases should be analyzed on a case-by-case basis as the conduct does not always harm consumers.

In Europe, discrimination cases continue to make it to the CJEU, although most of them constitute discrimination by a vertically-integrated entity against downstream rivals (Cole and

Ponsay, 2018). With the wording of Article 102(c) of the TFEU, the courts are then worried about whether the discrimination resulted in the “hinder[ing of] the competitive position of some of the business partners of undertaking in relation to the others” (British Airways v. Commission,

2007). The courts used an as-efficient-competitor (AEC) test to determine whether the

8 See FTC v. Morton Salt Co., 334 U.S. 37 (1948) where the court found that Morton Salt had violated the Robinson-Patman Act by offering quantity discounts. However, it seems very unlikely that a modern court would find quantity discounts a violation of the Robinson-Patman Act, as the courts now recognize that quantity discounts are not anticompetitive.

42 discriminated firm was as efficient as the dominant company, absent the conduct. If the discriminated firm were as efficient as the dominant firm and was pushed out due to the discrimination, then this would be considered anticompetitive under Article 102(c) of the TFEU

(Cole and Ponsay, 2018). This precedent was updated in MEO v. Autoridade da Concorrência

(2018) where the court had to show that the discriminated firm being put at a competitive disadvantage by the plaintiff's rates of acquisition (Cole and Ponsay, 2018).

Finally, until 2009, Canada had a provision about price discrimination in the Competition

Act (1985). However, in 2009, Section 50 was repealed alongside other provisions in the Act because similar to the United States, these laws were at odds with economic thinking, and were costly to administer (Musgrove, 2009). The repeal of the price discrimination provisions signals similar opinions on price discrimination as the United States, who still have the Robinson-

Patman Act on paper, but has not enforced it in decades.

There is a significant difference of opinion about discrimination on either side of the

Atlantic, where the Europeans are still prosecuting suppliers for discrimination of downstream buyers, and the United States and Canada whose discrimination laws are unenforced and repealed, respectively.

3.6.3 Refusal to Deal Theory.

In Trinko v. Verizon (2004) one sees that firms are not compelled to deal with any other firm, whether that be horizontal rivals or downstream firms, except in a specific set of circumstances set up in Aspen Skiing Co. v. Aspen Highlands Skiing Corp. (1985). The majority had this to say, quoting United States v. Colgate & Co (1919),

43 as a general matter, the Sherman Act does not restrict the long recognized right of [a] trader or manufacturer engaged in an entirely private business, freely to exercise his own independent discretion as to parties with whom he will deal. (p. 8)

Specifically, the Supreme Court had this to say about essential facilities doctrine in Trinko v.

Verizon (2004), “We have never recognized such a doctrine, and we find no need either to recognize it or to repudiate it here” (Trinko v. Verizon, 2004). Therefore, in post-Trinko v.

Verizon (2004) it is suggested that in refusal to deal cases, plaintiffs must show that their case falls under the specificities of the Aspen Skiing Co. v. Aspen Highlands Skiing Corp. (1985) case (Fox, 2005). From this, one sees that the Supreme Court wants to protect the firms right to deal with whomever it pleases, while at the same time denying that there is an essential facilities case to be made.

To see what state refusal to deal jurisprudence in Europe is Petit (2015) turns to some court cases which explains in which scenarios these rules are applied in. He points to three cases which generally follow the same pattern. First, there is a vertically integrated upstream firm which is integrated into the competitive downstream business. Then, the vertically-integrated firm tries to give itself an advantage downstream by way of vertical integration. The most common type of advantage given is the unequal pricing of inputs for the downstream firms. The case which set out the conditions which made up an anticompetitive refusal to deal were laid out in Bronner v. Mediaprint (1998). In this case, the Court of Justice of the European Union (CJEU) decided that the plaintiff need not only show that the conduct is likely to remove all competition in the market, they must also show that the service in itself is indispensable to carry on the firms business and that there are no actual or potential substitutes for the foreclosed input (Akman,

2017). Petit (2015), argues while there is a façade of the essential facility doctrine

(indispensability test), the cases decided by the courts in Europe show that this doctrine punishes

44 inputs for being much less than “essential,” pointing towards Konkurrensverket v. TeliaSonera

Sverige AB (2011) as evidence as to the failure of the European indispensability test (Petit,

2014). The way that the CJEU gets around not meeting the essential facilities threshold is by pursuing the anticompetitive harm through another harm, separate of refusal to deal. The CJEU decided in the TeliaSonera case that margin squeezes9 are a separate form of abuse, distinct from refusal to deal, and therefore, there is no need to pursue the essential facilities threshold.

Some legal scholars have suggested there are scenarios where there is market power created by the refusal to deal. Such as when a firm owns a superior input and when they deny the input to a downstream firm, its revenues are decreased, perhaps to an unsustainable level. The primary example used by authors is Aspen Skiing Co. v. Aspen Highlands Skiing Corp. (1985), where the dominant firm denied to continue a deal where consumers could buy a pass to all four ski hills. It has been suggested that the denial of the input to the other ski hill was found to be anticompetitive for several reasons. First, the arrangement was presumed profitable since the dominant firm engaged willingly to start with, and that the discontinuation of the arrangement presumed that the dominant firm was willing to sacrifice short-term profits to an anticompetitive end (Fox, 2005). Second, the dominant firm was unwilling to sell bulk tickets to the harmed firm, even at retail price (Fox, 2005). Third, the dominant firm denied to even sell to the harmed firm, at any price (Fox, 2005). These three points led the Supreme Court to find that Aspen Skiing’s conduct was that which constituted anticompetitive refusal to deal.

There are also situations where downstream firms cost more to supply than it is worth to the upstream firm, or they are free-riding off the investment of the upstream firm. The primary

9 Margin squeezing is defined as when an firm which is involved in the upstream and the downstream market charges a high price in the upstream market, and a low price in the downstream market, thereby squeezing the profit margins of competitors downstream.

45 example coming from Europe to show this is Bronner v. Mediaprint (1998), where there are two newspapers, one of which is much larger and created the only nationwide delivery system in

Austria. Mediaprint denied the use of their delivery system to the much smaller newspaper. The

Court of Justice then had to figure out whether this exemplified anticompetitive refusal to deal.

One could interpret this action in an anticompetitive way or a cost-saving way. If the denial of the service was to try and stomp out the smaller newspaper, which was growing and potentially could offer its own distribution service which could compete with Mediaprint’s dominant distribution service, then it could be deemed anticompetitive. Nevertheless, one can also imagine that the cost of fulfilling the delivery service could have considerably increased with the addition of another newspaper with potentially different customers, and different pickup locations.

In Canada, refusal to supply falls under the Competition Act, where the Competition

Bureau states that there is no obligation for any business to buy from or supply any other business (Competition Bureau, 2019). However, the Competition Bureau states that the conduct may be anticompetitive if the good or service being denied is both competitively significant and cannot otherwise be feasibly obtained from other sources (Competition Bureau, 2019). If this is the case, then, absent a legitimate business justification for refusing to supply, the Bureau could decide that the refusal was done for anticompetitive ends (Competition Bureau, 2019). The

Bureau does acknowledge that competitively significant inputs are often the result of costly investment and innovation, and forcing firms to share may undermine investment incentives for firms to develop new and beneficial goods and services (Competition Bureau, 2019). From these previous sentences, one can see that Canada has developed itself an unofficial essential facilities doctrine, whereby firms may be forced to deal if the input is deemed essential. This is also seen in recent Competition Bureau decisions where firms with market power upstream were found

46 guilty for anticompetitive behaviour because they refused to deal with some downstream producers, and their input was deemed essential to the operation of the downstream firms

(Church, 2018). Although on record, Canada does not have an essential facilities doctrine, some recent cases have expressed the sentiment that there is indeed an essential facilities doctrine at play in Canada.

All jurisidctions agree that firms generally have a right to deal with the firms they see fit, however, they differ on opinions of essential facilities. The United States’ Supreme Court has refused to recognize the essential facilities doctrine. Whereas, in Canada, the Competition

Bureau has no mention of essential facilities in its Competition Act, however, recent cases have alluded to the existence of essential facilities doctrine. Comparatively, Europe has an essential facilities doctrine (indispensability test), but there is evidence that they have to prescribe that input is essential when there appear to be legitimate substitutes available.

3.6.4 Tying Theory.

Before the 1970s, tying was per se illegal in the United States, which means that if firms were caught attempting to tie their goods with another, they were violating Section 2 of the

Sherman Act. The reason why it was considered per se illegal was that tying “generally serve[s] no legitimate business purpose that cannot be achieved in a less restrictive way,” as stated by

Justice Stevens in United States Steel v. Fortner (1969). Although tying was per se illegal, some thresholds needed to be met for it to be punishable. As summarized by Ahlborn, Evans, and

Padilla (2004) the plaintiffs had to prove that there was substantial market power in the tying market, and therefore the tie would be effective in foreclosing competition (Ahlborn et al., 2004, p. 11). Another barrier the plaintiffs would have to hurdle described by Ahlborn et al. (2004) is

47 proving that consumers were forced to buy the separate tied good (p. 9). The plaintiffs would also have to show that there was significant commerce going to the tied goods because of the tie

(Ahlborn et al., 2004, p. 11). If the revenue accrued from the tied good is small, then it was likely that the tie was not very harmful. Finally, the U.S. Courts agreed that there were exceptions to the per se rule for tying, especially for new industries where firms were tying goods to ensure proper functionality in complex systems (Ahlborn et al., 2004, p. 11).

From the preceding sentence, one sees cracks in the faith of the assumption that tying was always anticompetitive. If there is a reason to tie two goods together to ensure technical operability, then tying cannot always be objectively anticompetitive. This realization was reinforced by Jefferson Parish Hospital Dist. No. 2 v. Hyde (1984) whereby four Justices wanted to head towards a rule-of-reason approach with regards to tying. The Supreme Court realized that in some cases tying may be welfare-enhancing, but were bogged down by past decisions which concluded the per se illegality of tying.

While the Supreme Court started questioning the per se illegality of tying, Chicago

School economists had created models which explained situations where tying could be welfare enhancing. Situations where a firm wanted to ensure quality and operability, lent tying to be an appropriate solution for firms. However, the Chicago School also hypothesized situations where tying could lower production, distribution, and transaction costs, through economies of scope, which had the possibility to enhance efficiency.10 Besides modeling conditions where tying can be welfare improving, the Chicago School also explained why tying with the presence of market power may not be anticompetitive. Using the Single Profit Theorem, a firm which has market

10 See Bork (1978), Stigler (1968), and Posner (1976) for more.

48 power in one market cannot increase its profits by integrating into a second good market.11 If a monopolist cannot increase (or decrease) their profits by integrating, they will not do it except for one of the reasons discussed above. Therefore, the conclusion drawn from this analysis is that if one sees a firm with market power tie their good to a competitive market, it must be for efficiency-enhancing reasons. This analysis was fundamental in showing that tying was not always anticompetitive, and should be looked at in the context of the case when analyzing whether it is anticompetitive or pro-competitive.

While the Chicago School’s analysis showed that tying should be per se legal, there were critiques of the conclusions drawn from the analysis. Whinston (1989) points to the fact that if tying can change the market structure favourably for the monopolist, then there may be an incentive to tie the secondary good to the primary good. This is true if the secondary market experiences economies of scale, so by reducing the quantity sold in the secondary market, they can push up costs and drive down profits of competing firms, until they exit. They are thereby creating market power in the tied market. Additionally, if the upstream and downstream demands are independent or complementary, then consumers can be harmed by the monopolistic upstream firm's attempts to monopolize the downstream market (Whinston, 1989). These harms come about from the elimination of choice for consumers, or in the case of complementary demand, the implicit harming of the upstream product by the elimination of complements downstream

(Whinston, 1989).

Carlton and Waldman (1998) continue with this research into tying with economies of scale and imperfect competition when they look at tying in emerging innovative industries. Their

11 See Whinston (1989) for the formulation of this result

49 model lends evidence that a monopolist in a primary market can have the incentive to tie into the secondary market when the market has significant innovations, and product lifespans are short.

Therefore, a monopolist in a primary market is less worried about competition in the existing market; they are worried about some firm creating a new market which displaces them from their shortlived monopoly. One place that this new market can rise out of is this secondary market. For example, in United States v. Corp., (2001) Bill Gates was less interested in monopolizing the internet browser industry when Microsoft tied their Internet Explorer to their operating system; he was more concerned with the possibility of Netscape “sabotaging our platform evolution." (Gates Dep. 1.)

Europe has similar provisions for when tying is deemed abusive. The tying and tied products should be considered separate products; the firm is dominant in the tying market; the tie does not give consumers a choice to obtain the tying products separate from the tied product; tying forecloses competition (European Commission v. Microsoft, 2004). All four of these conditions must be fulfilled at the same time for the tie to be deemed anticompetitive in Europe.

The final provision that the tie foreclosed competition has two interpretations. First, there is the classical interpretation, where the tie inhibits competition and makes it so the dominant firm can better exercise market power. Secondly, there is a question of whether the tie interfered with normal competitive processes. For example, consider a firm with a dominant product which has multiple complementary downstream markets. Consider if the dominant firm ties a product into one of the complementary markets and forecloses competition in that market. Other firms in other complementary markets will see this, and their investment incentives will be lessened because they notice that the dominant firm can tie into their market and foreclose them as well.

As discussed in European Commission v. Microsoft (2004), Microsoft was tying its Windows

50 Media Player to its dominant operating system. There are many other complementary software which conceivably could be appropriated by Microsoft. Therefore, it was alleged these firms who create software would likely underinvest because they foresee Microsoft appropriating their software and tying it to Windows. This reduced incentive to innovate could also have the effect of maintaining a monopoly as if these complementary firms do not innovate, their product may not overtake existing monopolies, thereby protecting market power. The effects described here are challenging to measure, as researchers would have to accurately hypothesize a world where the tie did not happen and measure the difference in innovation in each scenario.

Finally, in Canada, there are no specific provisions for tying, just as there are no specific provisions for refusal to deal, as mentioned before. All conduct falls into abuse of dominance, where the Competition Bureau analyses whether the conduct significantly raises the costs (or significantly reduces the revenues) of its rivals. Once confirmed, the Competition Bureau must then show that because of the conduct, the dominant firm has created more market power.

Additionally, in 2019 the Competition Bureau released updated abuse of dominance guidelines which mention tying. It says that “tying occurs when as a condition for obtaining or using one product, a firm requires or induces a customer to purchase another product as well”

(Competition Bureau, 2019). It continues, saying that “the Bureau will consider whether the tie excludes competitors in whole or in part by increasing their costs or reducing their revenue”

(Competition Bureau, 2019). So, similarly to refusal to deal in Canada, the worry is not about the classification of the conduct, the worry is whether the conduct significantly harmed rivals ability to compete, and the effect this harm to rivals has on competition.

Opinions on whether tying is anticompetitive have changed and adapted since the 1970s wherein the United States, tying was per se illegal. Since then, the Chicago School has shown

51 situations where tying can be efficiency-enhancing. Meanwhile, other economists such as

Whinston (1989) and Carlton and Waldman (1998) have shown situations where tying can create, maintain, or enhance market power, and has the power to harm consumers. Therefore, while tying can have pro-competitive justification, it can also be a tool for monopolists to protect or create market power. All three jurisdictions are then worried about the effect the tie has on competition, and whether the tie can harm consumers in the long-run by stifling innovation.

52 Chapter 4: The Facts Found by the European Commission

4.1 The EC’s Document Regarding Their Decision in Google Shopping

While the EC published a 216-page document explaining their reasoning behind fining

Google for their conduct, Canada’s Competition Bureau and the United States’ FTC published shorter documents expressing their reasons for closing their investigations. The United States decided that Google’s conduct was pro-competitive as it benefitted consumers primarily. Canada closed its investigation because it did not find evidence that Google’s conduct resulted in a substantial lessening of competition. The United States’ and Canada’s documents did not need to discuss any specifics, as they were closing investigations. However, since the EC decided to pursue Google they were required to show this information as well as why they believed it to be true. Therefore this section will convey the EC’s opinions about market definitions, the presence of barriers to entry and network effects, and the effects of Google’s conduct. Alongside each of these discussions will be explanations as to why the EC believes each of these points to be so.

4.2 EC’s Market Definition

The first step in an abuse of dominance case under the Article 102 of the European

Union’s competition law is to determine whether or not a firm is dominant in the industry which they are being accused of abusing said dominance. The EC decides that the markets which

Google and Google Shopping compete in are general search, and comparison shopping services respectively (p. 28). They go on to say that general search is a distinct market, separate from vertical search engines, social media, or content sites (p. 30). Therefore, since a distinct product than these services, there is limited demand-side substitutability (p. 30). They continue with analyzing the supply-side substitutability by stating that Google has a large current market share on desktop and mobile general search (p. 34). They also go on to say that for a new

53 entrant to enter, they would have to invest a significant amount in the start to try and contest with

Google (p. 34). Therefore the EC concludes there are barriers to entry, which hinder the possibility of future supply-side competition. The EC is then suggesting that one relevant market is general search.

The EC goes on to state that comparison shopping services constitute a distinct economic market. They define comparison shopping services as “specialised search services that: (i) allows users to search for products and compare their prices and characteristics across the offers of several different online retailers (also referred to as online merchants) and merchant platforms

(also referred to as online marketplaces)” (p. 35). They go on to say that these services are distinct from specialized search services (such as flights, hotels, restaurants, or news), online search advertising platforms, online retailers, and offline shopping comparison tools (such as catalogues) (p. 35-36).

The EC qualifies that specialized search services are distinct from shopping comparison services for several reasons. On the demand side, they state that comparison shopping services offer a “selection of pre-existing commercial offers available on the internet for a particular product,” as well as the “tools to sort and compare such offers based on various criteria” (p. 36)

On the supply side, they qualify their statement by saying that specialized search services select and rank results using algorithms to determine the most relevant results for a particular search.

Specialized search services also select from a “pool of relevant suppliers who[m] they have a contractual relationship with” (p. 36). Specialized search services also are more dependent on real-time data (p. 36). The EC Contrasts this with comparison shopping services which “typically employ a commercial workforce whose role is to enter into agreements with online retailers, under which these retailers send them feeds of their commercial offers” (p. 36). The EC

54 concludes this by affirming that there is limited supply-side competition because specialized search services would be hard-pressed to offer services likening to shopping comparison services

“in the short-term without incurring significant costs” (p. 36-37)

The EC also concluded that comparison shopping services were distinct from online retailers for several reasons. First, on the demand side, they suggest that comparison shopping services are an intermediary between users and online retailers (p. 38). The EC suggests that users cannot directly purchase from comparison shopping services, but can from online retailers

(p. 38). On many occasions, online retailers and comparison shopping services are business partners or customers, as they often enter mutually beneficial agreements (p. 38-39). The EC claims that online retailers do not allow users to compare similar goods across different websites and would prefer if users visit no other site than theirs (p. 39). On the supply side, the EC suggests that shopping comparison services offer real-time feeds of many online retailers while sorting based on the query at hand (p. 39). The EC contrasts this with online retailers who have an inventory which they display and for which they offer check out services.

The EC then describes why merchant platforms, such as Amazon and eBay, are not in the same market as comparison shopping services. On the demand side, the EC claims both offer search and filter functionality across retailers and merchants, but they serve different purposes, according to sources which were not disclosed by the EC (p. 40). For similar reasons as online retailers, the EC concludes that merchant platforms and comparison shopping services are separate markets (p. 40). They then add to the differentiators that merchant platforms also mix in their third-party offerings to the user's searches as well as being perceived as multi-brand retailers, and list offers of used products as well as offers from non-professional merchants (p.

45). The EC also points to a study which was conducted in 2013 by themselves along with

55 Deloitte which shows that 91% of respondents said that price comparison services allowed users to compare prices, and 79% responded that the most important aspect of price comparison services is the ability to compare prices (p 45). The EC contrasts this with multi-trader e- commerce platforms (Amazon or eBay), which the majority of respondents (62%) indicated that these platforms’ primary use is to purchase products (p. 45). The EC suggests that comparison shopping services also allow a retailer to remain independent of online merchant platforms, which they often list with, but also compete with. The EC claims that online retailers often would not like to cede the customer interaction and data about its customers to merchant platforms, and using comparison shopping services aids them in this (p. 45).

The EC claims the supply side is similar to the online retailers logic again, although there is a difference in how each is remunerated (p. 47). They show that comparison shopping services are paid on a cost-per-click basis where the retailer pays the comparison shopping service for each click they redirect to the retailer (p. 47). The EC suggests that the merchant platforms are paid a listing fee or a commission on sold products, or if they have their own third-party products, the price of the product, which is different than comparison shopping services (p. 47).

Finally, the EC then explains why comparison shopping services are distinct from offline comparison shopping tools. On the demand side, the EC proposes that offline comparison services lack real-time updating, which is an essential reason why comparison shopping services are considered valuable (p. 54). Additionally, the EC points to the fact that users have much higher costs to follow up with offline comparison shopping tools than online (p. 54). They would have to call or visit the retailer in person to confirm the price and stock. The EC also points to the problematic nature of trying to compare prices across retailers offline. For the supply side,

56 the EC suggests shopping comparison services and offline comparison services differ regarding functionality and business model (p. 54).

The EC’s choice of the relevant geographic market is simple to explain. They decided that both comparison shopping services and general shopping services are national in scope. The

EC deduces that while these websites are available internationally, they often have localized syndicates which are in the local language and are focussed locally (p. 55). They also conclude that language is an integral part of comparison shopping services, as they compare the results of places where users can make purchases, which are likely primarily local and in the primary language (p. 55). Due to this, the EC concludes each service must design its product at a national level to accurately serve each demographic. One part of designing their services is entering into agreements with online retailers, which often operate at a national level as well.

Thus, the EC concludes that the relevant markets for the analysis consist of general search engines, such as Google, and comparison shopping services, such as Google Shopping.

4.3 EC’s Statements to do with Barriers to Entry

The EC puts forward several barriers to entry which exist for general search engines.

They suggest that significant upfront and continued cost, as well as data, the lack of multi- homing, and Google’s reputation act as barriers to entry (p. 62). First, they point to the fact that for a new company to enter, it would require a substantial up-front investment to join the market

(p. 62). This initial cost is shown through evidence submitted by Microsoft, which shows the capital expenditure and investment to do with their search engine, Bing. “Investment in equipment and personnel would likely be the primary costs, [….] [i]n addition, obtaining the large quantity of data necessary to develop an effective [general] search engine would be a significant barrier to entry” (p. 62).

57 Second, the EC suggests that since a general search engine uses data to refine its searches and make better predictions, they require a minimum amount of use to compete viably (p. 62).

The EC claims that generally, the greater the number of queries a general search engine receives, the quicker and better it is at detecting user patterns and figuring out the relevance of results (p.

62). On top of this, general search engines also need queries of uncommon, or “tail” results. The

EC claims that these uncommon queries are a primary differentiator of Google and its competitors (p. 62-63). The EC suggests that Google receives more uncommon queries, which helps them refine their algorithm to help answer these more difficult queries, which competing general search engines cannot answer with current data and usage (p. 62-63).

Thirdly, the EC claims that general search services regularly invest in improving their product and that entrants would, therefore, have to match these investments to remain competitive (p. 63). The EC provides a table of Google’s and Yahoo’s worldwide capital expenditures in their general search service from 2006 to 2015 as evidence for the investment required to compete with Google (p. 64). In the table, one sees Google’s investments rise from

1.9 billion USD in 2006 to 11 billion USD in 2014, while Yahoo’s investment remains relatively flat at under 700 million USD over the same period. The point being made by the EC is that for a competitor to compete with Google, they would have to invest similar amounts in continual improvements to their general search engine.

Fourthly, the EC points to the fact that users are infrequently multi-homing when using general search engines (p. 67). Multi-homing is the act of simultaneously using two or more of the same services which compete with each other. The EC acknowledges the ease of which users can switch from one general search engine to another, but state that this is not how users search online. They reference a 2010 survey which looked at how often Google users in ,

58 , , , and the would use other general search engines. They found that only 12% of users in Germany, Italy, and Spain used two distinct search engine for at least 5% of their online searches (p 66-67). In France, the number was 15%, while the United

Kingdom had 21% of users who used at least two distinct search engines for 5% of their searches

(p. 66-67). The same survey also found that users who indicated that Google was their primary search engine were less likely to multi-home than users who indicated that another general search engine was their primary search engine (p. 66).

The EC next states that the strength of the Google brand also creates a substantial barrier to entry for any potential entrants. The EC suggests that Google’s brand is considered one of the most valuable brands on earth, according to Forbes and WPP, which publishes reports on the most valuable worldwide brands (p. 69). The EC claims that according to internal Google documents, as well as third-party studies, users would likely not decide to switch or multi-home even if Google were degrading its quality of search results (p. 69-70). One third-party study used an eye-tracking experiment which indicated that users would select a higher result on Google’s search engine results page, even if the excerpt below the link were less relevant than a lower result (Pan et al., 2007). Another third-party survey found that users preferred search results that were labeled with Google, even if they were from another search engine altogether (Gesenhues,

2013). The EC then suggests that the brand built by Google has therefore created a barrier to entry, which an entrant likely will not be able to replicate (p. 70). The EC then suggests that it then follows that even if Google’s search quality were to dip below competitors, Google would still command a dominant position in the market because of its reputation and brand strength (p.

70).

59 The EC concludes that upfront and sustained costs are a barrier to entry since if a firm wished to enter, they would have to take the risk to endure upfront costs, as well as continued investment to improve its service. They also conclude that data would pose as a barrier to entry because for a firm to garner a product which competes with Google, they would have to acquire enough data to improve their product to Google’s level. The EC also claims that the fact that users do not multi-home search engines act as a barrier to entry, because it eliminates the need for other general search engines. Finally, they point to Google’s reputation as a barrier to entry.

4.4 The European Commission and Network Effects

Nestled inside the barrier to entry section of the EC’s document is a section about the network effects of general search engines. They comment on the positive feedback effects on both sides of the two-sided platform, which create an additional barrier to entry (p. 64). They suggest the presence of indirect network effects on the advertising side are such that the number of users on a general search engine leads to more value for advertisers who advertise on that general search engine (p. 64). The EC claims that the increase in value is then captured by the general search engine, which will invest the increase in revenue into improving their search engine to bring in more users (p. 64). They conclude that the improvement of the general search engine leads to more users, a higher value for advertisers, and more revenues invested in improving their search engine (p. 64). The EC also finds that direct network effects exist for the general search side, which comes from the fact that users derive benefit from search advertisements (p. 64). They conclude this by stating that since advertisers are willing to bid for

AdWords results lends evidence that some users value the advertisements (p. 64). They conclude that therefore, there is a link between the number of advertisers and the value that some users derive from these advertisements (p. 64).

60 The EC supports this idea by presenting additional factors, such as the fact that since

2007, several companies have exited general search services. Companies such as Yahoo and

Ask.com now rely on Bing and Google services, respectively, to complete their search services

(p. 65). Along with the exit of large companies, there has been a stalling of users and support for several smaller general search services which offer services in several languages (p. 65). The EC also points out there has only been one entrant in the national markets since 2007, which was

Bing, which was likely only able to be brought to market due to the deep pockets of Microsoft

(p. 65). Although Bing did enter, the EC suggests that it has never broken 10% market share in any EEA country (p. 65). The EC also suggests that several small companies have launched worldwide to compete with Google, but most have either given up and pursued more specialized search services or went out of business (p. 65-66). The EC points to several competing general search engines launched from the United States, which only serve English results, such as

DuckDuckGo (p. 66). However, since these are in English only and rely on a third-party to index and crawl the internet, it is unlikely that other general search engines like DuckDuckGo would be able to enter the market and do any better than Bing has faired (p. 66).

4.5 The EC and Google’s Self Preferencing

This section will discuss the reasons why the EC thinks that Google violated Article 102 of the TFEU. The EC states that Article 102 of the TFEU prohibit abusive practices which may cause damage to consumers directly, but also those who harm consumers indirectly through their impact on an effective competition structure (p. 74). The EC goes on to say that the concept of abuse relates to behaviour, which is such to influence the structure of the market, whereas a result of the abuse the state of competition is weakened (p. 74). They contrast this abuse with competition on the merits of the product, which they desire to protect. The same Article also

61 prohibits conduct which tries to extend a dominant position from one market into a separate market through the distorting of competition (p. 74-75). The EC states that to determine whether a dominant firm has abused its position the investigation must determine how the abuse, for example, bars competitors from access to the market, applying dissimilar conditions to equivalent transactions with other trading partners or exploiting a dominant position by distorting competition (p. 75). The EC suggests that the prohibition of the abuse of a dominant position is sustained regardless of the means and procedure by which it is achieved and irrespective of any fault (p. 75-76).

The EC says the following with regards to Google’s conduct with their Google Shopping and Google Universal services. They deem the conduct is abusive because it constitutes a practice outside competition on the merits as it: (i) diverts traffic from Google’s general search engine results page to Google’s comparison shopping service (ii) is capable of having, or likely to have, anticompetitive effects in the national markets for comparison shopping services and general search services (p. 76).

The EC goes on to say that Google positions its comparison shopping service higher on its search engine results page than that of competitors (p. 77). They suggest that this superior rating is not due to the relevance, but due to the forced introduction of a box at the top, which pushes all other advertisements, and organic results further down the page (p. 77). The EC suggests that not only was Google’s service given priority but with the release of several different algorithm changes aimed at combatting spam and irrelevant search results, several of

Google’s competitors were demoted by Google’s algorithm (p. 77). The EC claims that due to the priority given to Google’s service, and the fact that it was prominently displayed regardless of relevance, Google Shopping was never demoted when it conceivably should have been along

62 with the competitors (p. 77). The EC suggests that the decline of comparison shopping services is seen in a graph done by ComScore, which showed that there was a massive decline in the

Sistrix Visibility Index,12 post Panda algorithm change (p. 81). The EC states that these results are consistent across all of the EEA countries when the Panda algorithm update was launched in each country (p. 82-90). They claim these declines in visibility on Google’s search engine results page correspond with a decline in traffic to these comparison shopping services from Google (p.

90). The EC suggests that this decline of ranking of rival comparison shopping services during the launch of the Panda algorithm update in 2011 never did get better for the services, as during the investigation in 2015 and 2016, the average comparison shopping service was displayed on the fourth page of Google’s search engine results (p. 97). They also point to one last way that

Google prioritized its service over competitors was the ability for advertisers with Google

Shopping to attach pictures of their products (p. 103). The EC suggests that since comparison shopping services are limited to what Google gives them, which is only just a link and an excerpt about what the page entails, this again leaves comparison shopping services at a disadvantage (p.

103). The EC suggests that this point is supported by submitted evidence using eye-tracking software, this is quantified by saying that the inclusion of pictures on advertisements increases click-through rates of the advertisement as opposed to a text-only excerpt (p. 103).

The EC concludes that by Google limiting the traffic that comparison shopping services received from Google’s general search engine, Google limited the ability of comparison search services to improve their product (p. 120). The EC claims that traffic allows these comparison shopping services to improve the relevance and usefulness of their results for users, because, like for general search engines, comparison shopping services improve with traffic (p. 120). They

12 The weekly Visibility Index follows changes in rankings of individual sites and a large subset of keywords for Google’s search engine.

63 also suggest that traffic convinces merchants and advertisers to advertise with the comparison shopping service, which will raise revenues, which will improve the usefulness of the service, just like for general search engines (p. 120-121). The EC claims that with traffic, comparison shopping services also will be able to provide suggestions for other search terms, as well as be able to generate more organic user reviews, which helps other users of the service (p. 123).

Therefore, the EC claims that for Google to limit the traffic that these comparison search services receive is to limit a critical input for them, reducing the profitability of providing the service (p.

123-124).

The EC goes on to show that due to Google’s conduct, they reduced traffic to these comparison shopping services. The EC points to the fact that users are much more likely to select the top links on Google’s results page because of the behaviour of users, and the trust they have in Google’s algorithm (p. 126-131).13 While the exact numbers for the comparison shopping services are redacted in the report by the EC, it is shown that the decreases in traffic to select comparison shopping services is between 7% and 92%, of which almost none recovered after their temporary drop in relevance in Google’s search results (p. 129).

The EC also shows that there is a correlation between the trigger rate14 of Google

Shopping on Google’s general search engine, and traffic from Google’s general search engine to

Google Shopping (p. 139-145). The EC suggests that when looking at the numbers for traffic to

Google Shopping and all other comparison shopping services, there is a divergence where traffic to Google Shopping is increasing, but traffic is decreasing for competitors (p. 146-155). The EC

13 See http://www.seodesignsolutions.com/blog/articles/search-behavior-why-ranking-above-the-fold-matters/ 14 The trigger rates are defined as the proportion of queries/keywords for which a particular website or service is displayed (“triggered”) among generic search results on Google's general search results pages (European Commission, 2017)

64 does not offer any specific numbers to show this however, Graphs 45-56 show the relative traffic to Google Shopping compared to other comparison shopping services (p. 146-155).

The EC next goes on to state that the traffic lost from Google is a large proportion of the traffic which these sites receive, and they will likely not be able to replicate the traffic from any other source (p. 162). They publish a large, heavily redacted table, which shows several different groups which have significant traffic coming from Google’s general search engine (p. 163-166).

The numbers are quite diverse in their changes over time. Some of the groups see a tiny drop in traffic from Google, as well as a significant increase in traffic from other sources, leading to rapidly increasing total traffic numbers. However, there are other groups which see significant drops in traffic from Google’s general search engine, with no increases in traffic from other sources, and therefore see decreasing traffic throughout Google’s conduct. The EC claims that while the other sources being increased is a good thing, these other sources may be more expensive and less efficient than traffic from Google (p. 167). For example, one of the columns is traffic from AdWords, which is a service which the advertisers have to pay Google when the advertisements are clicked on (p 167). The EC points to the issue that, as mentioned before,

AdWords is also dictated by Google’s algorithm to maintain the relevance of results (p. 167-

176). Therefore, if the comparison shopping services were demoted due to the algorithm updates, then it would be unlikely to be seen as relevant by AdWords as well since they use the same algorithm (p. 172). They conclude that these comparison shopping services could not substitute sponsored results for the organic results, since both would have demoted. The EC states that services would prefer to be included in the free organic search results, as opposed to having to pay to be seen, so while traffic from sponsored results can work, traffic from Google’s general search results would be preferred (p. 176). They also claim that another source of traffic for the

65 groups is advertising outside of Google, trying to build a brand name so that users bypass Google altogether (p. 176-179). This is similar to another column of different traffic sources, mobile applications. The EC suggests that for a user to download and use a mobile app presupposes that the user is aware of the brand and has a previous relationship with the brand (p. 176). The EC concludes that the comparison shopping service will need to build its brand before replacing its traffic with a mobile app. They state that these sources of traffic are costly and may be less efficient uses of money than relying on Google’s organic search results (p. 179).

The next step for the EC was to show how the conduct had the potential for anticompetitive effects in the national markets for comparison shopping services. They suggest that due to the decreased traffic from Google’s general search engine to comparison shopping services, as well as increased traffic to Google Shopping, a rival to these services, the conduct is capable of having anticompetitive effects on the market of national shopping comparison services (p. 179-180). They claim that the conduct reduced traffic to comparison shopping services, reducing the potential for these services to earn revenue, which could cause these services to close down (p. 180). The EC claims that if these services were to close down, then

Google would be able to charge higher fees to merchants to participate with Google Shopping, due to lack of substitutes (p. 180). The EC does not expand on this point, however, assuming that a percentage of these higher fees charged to merchants are passed onto consumers through higher retail prices, then the conduct also increased prices for consumers. The EC adds that along with these price effects for both the merchant and the consumer, there can be dynamic effects going forward due to the conduct (p. 180). The EC suggests that if the decrease in traffic did not force comparison shopping services to shut down, it would still affect the services’ long-term incentives to invest in innovation going forward (p. 180). The EC then asks, if they foresee an

66 increase in traffic coming from somewhere else rather than Google’s search results, (which the

EC determined was unlikely), would comparison shopping services have the same incentives as before the conduct? The EC concludes that not only is this unlikely, but Google also lacks the incentive to improve Google Shopping as they can merely promote it ahead of presumably superior rivals and exploit the trust in their algorithm (p. 181).

Another harm the EC claims is caused by the conduct is the reduction of the ability of consumers to access the most relevant comparison shopping services (p. 181). They suggest because Google’s own vertical search engines are exempt from Google’s ranking through its algorithm, Google can exploit the user's presumed trust in Google and place its services ahead of more relevant organic results (p. 181). They claim users have been trained to believe that the top links on Google are the most relevant results for their query, therefore when a link shows up at the top for some other reason than relevance, the user is no longer receiving the most relevant results (p. 181). The EC suggests that this is compounded by the fact that these universal changes were implemented without informing users that the universal results were ranked differently than organic results (p. 181). Therefore, the EC concludes that the conduct harms consumers directly through the worsening of Google’s general search engine to promote Google services such as

Google Shopping.

The EC then extends the analysis for when the market includes merchant platforms, to show that even with merchant platforms included in the market, the conduct can still harm consumers (p. 183). The EC suggests that if Google’s conduct were to shut out competitor comparison shopping services, they could charge higher prices to merchants to be included in

Google Shopping is unchanged with the new market definition (p. 183). As the EC discussed before, comparison shopping services and merchant platforms are more likely to be business

67 partners in a vertical relationship, which comparison shopping services deliver users to merchant platforms, who then reward comparison shopping services for doing so (p. 183). The EC states that not only do they consider themselves business partners, they often serve different purposes as the EC discussed before (p. 183). So the EC concludes that if Google’s conduct were to foreclose comparison shopping services, the same price effects exist after including merchant platforms in the market because they operate in a different sub-section of the industry and fulfill different services than comparison shopping services.

The EC then gives the results of analysis conducted about the effect of the conduct on the market. They do five different analyses for each country that the conduct was done in. This thesis will only mention the first of the five because it is the most relevant as it deals with the total traffic coming into comparison shopping services from all sources, and it corrects for the limited substitutability between merchant platforms and comparison shopping services. The adjustment reflects that the substitutability is limited to the web pages of these platforms that users visit with a view of offers from several online retailers as an alternative to price comparison services (p.

186). Of all the analyses, this analysis is most representative of the market realities. The EC then states that the conduct had been capable of having, or likely to have had anticompetitive effects on 12% to 87% of the predetermined market from 2011 to 2016 (p. 191-195). That is to say that the conduct is likely to have had anticompetitive effects on 12% to 87% of the national market for comparison shopping services (p. 191-195).

The EC decides that due to the anticompetitive actions done by Google in the national markets of shopping comparison services, the EC decided that Google violated Article 102 of the

TFEU, and therefore were fined for the damages they caused throughout the conduct. A resolution suggested by the EC is that Google was to ensure that all competing comparison

68 shopping services are treated no less favourably than Google’s comparison shopping service on its general search engine results page (p. 204-206).

69 Chapter 5: Comparison Between the Facts Found in this Thesis and The EC

5.1 Chapter Introduction

This chapter begins with the conclusions determined from the theories developed in

Chapter Three about the barriers to entry and the presence of network effects, and what each of these means for Google. Next, this chapter will discuss this author’s preferred market definition for both markets, one of which is a market for general search, and another consisting of “retail searches,” which will be further discussed in Chapter 5.2.

This thesis then asks whether Google’s conduct should be considered economically anticompetitive, based on the test set up in Krattenmaker and Salop (1986). This thesis then concludes that Google’s conduct cannot be considered economically anticompetitive for three reasons. First, Google cannot be said to be denying a vital input to comparison shopping services if there exist legitimate outside ways for them to attract traffic to them. This appears to be the case because looking at a similar market, online travel, one sees that Google’s service Google

Flights struggles massively against incumbents Booking.com and Expedia, who invested heavily in advertising to draw users directly to them, circumventing the “essential input” provided by

Google.

Second, even if Google were able to eradicate comparison shopping services, Google would not be able to better exercise its market power because of the existing competitive constraints, such as Amazon and eBay. The story which was told by the EC was that Google would be able to knock out comparison shopping services, and then be able to charge more to merchants who wished to advertise their goods online. However, because of the EC’s narrow market definition, they did not include Amazon and eBay, who merchants would also be able to advertise with, even with the absence of comparison shopping services. Therefore, even if the

70 conduct were successful in eliminating comparison shopping services, there would exist competitive constraint to Google’s market power, signifying that market power was not created by the conduct.

Finally, the third reason that Google’s conduct should not be deemed anticompetitive is that the conduct is an exercise of existing market power in general search, not conduct which creates or maintains market power. The ability for Google to move downstream and interfere with the market for retail searches on its results page stems from the market power in general search. Google’s general search engine is deemed to be superior to all others, lending it market power, which Google then exercises by entering and promoting Google Shopping. No market power is created downstream by the conduct; it stems from the existing market power upstream.

Next, this thesis goes through the legal necessities of the jurisprudence in each jurisdiction and concludes that Google’s conduct cannot be legally deemed discrimination, refusal to deal, nor tying, based on the legal thresholds put forward in Chapter Three.

Next, this thesis discusses the differences between the results found in this thesis, and the results concluded in the EC’s report. This thesis agrees with the EC that one market consists of general search, but disagrees with the definition of the secondary market. The EC concludes that the secondary market consists of merely Google Shopping and comparison shopping services, while this thesis has argued that it should also include online merchants like Amazon, as well as offline merchants, stepping up their online presence. This thesis generally agrees with the conclusions drawn by the EC with regards to barriers to entry, but objects with the extent that data constitutes a barrier to entry. Next, this thesis also generally agrees with the conclusions drawn by the EC about network effects, again, only contesting the presence of direct network effects.

71 Next, this thesis has some general comments about the EC’s decision on the Google

Shopping case. These disagreements include comments about the continued investment in

Google Shopping, Google’s censoring power, the possibility that comparison shopping services could have experienced a decline from sources other than Google Shopping, the explosion in capital inflows to European e-commerce, which the EC could ignore because of their market definition, the EC’s decision ignoring the benefits of Google’s conduct, and the questionable presence of dynamic innovative harms.

Finally, this section comments on the possibility that, while the case cannot be economically justified, it could be legally justified in Europe. It could be legally justified because of Europe’s relative preference for avoiding Type I (underenforcement) as opposed to Type II

(overenforcement). A similar reason that the decision could be legally justified is that the EC has stated that its purpose is to protect consumers. However, the ECJ has not formally endorsed that interpretation. Moreover, as seen in recent prior cases, it seems that the European competition authorities have focused entirely on the harm done to rivals, merely referencing consumers in passing, as if its primary purpose were to protect harmed rivals. The final reason that this decision could be considered legally justified is that Europe includes a form of anticompetitive abuse, which is called exploitative abuse, or excessive pricing. If one interprets that excessive pricing comes from the exercise of market power, then one could make the connection between exercising market power, and the aforementioned exploitative abuse, opening the possibility for firms to be found guilty for exercising its market power.

This chapter concludes with a discussion about what is next for Google and the EC. Since the closing of this case, several new investigations into other Google services like Google Maps and Google Flights have been opened, with rivals claiming similar issues as the Google

72 Shopping case. If the facts found by the EC in the Google Shopping case permeate to these cases, and there is a fundamental misunderstanding of the market definition, then another incorrect decision could be decided, further harming consumers.

5.1.1 Barriers to Entry in Search Engines.

Horizontal search engines experience large economies of scale, as most of their investment is frontloaded, and the cost to serve another user is negligible. There was, and continues to be, substantial investment in Google’s algorithm and ancillary services which will continue to make Google the leading general search engine. However, with these upfronts costs of creating and updating their search engine, the cost of servicing one additional consumer is virtually zero. Therefore, one would expect to see an exponentially declining average cost curve, where if Google can serve, hypothetically, an infinite number of consumers, their cost curve will be indistinguishable from zero. This point where Google’s average cost curve approaches zero is unknown, but if one hypothesizes about where it approaches zero, something can be said about the nature of competition in general search engines. If the average cost curve is rapidly approaching zero, then one can say that the economies of scale are exhausted quite quickly and general search engines need not have the vast hordes of users of which Google has to compete effectively. However, if the average cost curve continues for a while before approaching zero, then there may be a situation where there are fewer and less effective competitors in horizontal search. Though this thesis can hypothesize that the average cost curve rapidly approaches zero, this thesis lacks the data to confirm or deny the hypothesis.

Economies of scale can act as a barrier to entry alongside large sunk costs. Since Google and other general search engines compete to supply users with answers to search queries, one can think of their competition as Cournot in nature, and assign a strategic substitutionary relationship

73 between Google and other general search engines.15 With this relationship and the fact that there exist economies of scale in general search, there is a strategy which Google has applied, which can act as a barrier to entry. Consider the amount of investment in Google’s general search engine, or its reputation. If someone wished to enter today, they would have to invest a similar level to what Google has spent to develop a technology similar to Google’s level, as well as build a reputation of returning high-quality search results. However, if they were to enter, in a perfect world for them, they would only get half the general search market. The argument can be made that neither Google nor another entrant would be profitable with merely half the market because of the level of investment done by Google. Therefore, by Google knowing that potential entrants who wish to enter would need to do so at a level similar to them, which would be unprofitable to both of them in the best-case scenario for them, Google knows that firms should not enter.

Google has a first-mover advantage, which has led to them entrenching themselves by over- investing in R&D and reputation. The only way that entry would be possible is if a firm can enter on a level which is much smaller than Google, therefore bypassing the large upfront capital costs as a barrier to entry.

Some have argued that the amassed data which Google has from its long history of general search constitute as a barrier to entry. This argument is unconvincing for several reasons.

First, it is hypothesized that data has a marginally decreasing effect on search quality, which is to say that every gigabyte of data improves the quality of the search engine, but less so than the previous gigabyte (Varian, 2018). Second, work done by Chiou and Tucker (2017) suggests that recent data is more important to search quality than older data. That is to say that the older data

15 Strategic substitutes signify that when one firm increases its production, the rival firm will reduce its output in reaction.

74 Google has is likely not improving its current iteration of Google’s algorithm very much. While amassed data may not be a significant barrier to entry, amassing long-tail searches may be a barrier to entry. Long-tail results are uncommon queries, which, help refine a search algorithm by uncovering information about uncommon queries. Since Google has more users than any other general search engine, it likely has more long-tail results, which may prevent entrant search engines from competing effectively. Therefore, it does not seem likely that amassed data is a significant barrier to entry, but long-tail search results could act as a barrier.

There is no denying it; Google has a reputation. When searching the internet, the de facto term used to do so is “Googling.” Google is so synonymous with search that Oxford Dictionary added “Google” to their dictionary in 2006 as a verb which meant “to search for something online.” However, one should question whether or not this reputation is durable through exercises of market power. If Google were to exercise their market power by distorting search results, or something else similarly distasteful for consumers, would users switch to other horizontal search engines? Google has been very stubborn with ensuring that their search engine is preloaded on all new Android phones, as well as paying USD 3 billion in 2017 (Hasselton,

2017), USD 9 billion in 2018 (Segarra, 2018), and USD 12 billion in 2019 (Segarra, 2018) for the right to be the default search engine on iPhone. Google presumably does this because there is a habitual nature to using online search engines. Users either use their URL bar to search, as in the Chrome and Firefox internet browsers, or they may have a bookmark to their favourite search engines, or it may be their homepage as they open their browser. Mobile users will have to make a conscious effort to try and not use Google, leading to similar effects in the mobile market. The use of defaults is just one example of a nudge, a term used in behavioural economics to represent an architectural choice which predictably alters people’s behaviour without forbidding any other

75 options or significantly changing their economic incentives.16 The nudge is important because some users are lazy or ignorant about their choices, and therefore will not change the default option, lending the default option an advantage over other alternatives.

Though there may be a psychological reason why users do not switch away from the default option due to this nudge, it does not change the fact that if users wish to change search engines, they can easily do so. To switch search engines, a user must go to another website and search there. They can even use another search engine to find the search engine they want to switch to! Compare this to the example of relatively expensive switching costs given above with changing operating systems. Users had sunk significant time into developing skills in the

Windows operating system, learning its insides and outs to be more effective in using Windows.

If an operating system came out which was marginally better than Windows, users would be apprehensive about changing operating systems because of the sunk cost of learning Windows, as well as the potential time and effort to develop similar skills in the new operating system.

Another force trapping users into the Windows ecosystem was the prevalence of all the applications which were only available on Windows. Alongside these strong reasons for users to continue to use Windows, multi-homing different operating systems was difficult. Multi-homing is when users concurrently use multiple products which fulfill similar goals. Users of an operating system find it very difficult to multi-home a desktop computer so that they can enjoy benefits of many operating systems, but users of horizontal search, or any website, can open up two internet browser windows and use both for the same search to ensure they find what they are looking for. The ease of switching is seen when Google had a glitch in their algorithm, which

16 See Thaler and Sunstein’s 2008 book entitled Nudge: Improving Decisions about Health, Wealth, and Happiness for a detailed explanation of this phenomenon.

76 flagged all results as malware on January 31, 2009. Due to the glitch Yahoo! queries doubled for the 24 hours when Google was out of commission.17 Since there exist no expensive switching costs for users such as a cost or time to learn a new platform, one can see why Google resorts to trying to make their search engine the default and create a reputation, as these are the only ways which can ensure repeat users and continued support.

The conclusion to draw from this is then to see that barriers to entry do exist for Google, such as economies of scale, a first-mover advantage, brand effects, and nudges which bias users towards using Google. However, barriers that do not exist include older amassed data and switching costs. These barriers make it difficult for new search engines to catch up to Google, which partially explains the failure of Bing and DuckDuckGo to catch up to Google.

5.1.2 Network Effects in Search Engines.

Are direct network effects present in general search? Direct network effects are where more users lead to a more attractive product for new users. Are indirect network effects present in general search? Indirect network effects are where separate sides interact with each other, and bring value to the opposite side.

The presence of direct network effects in horizontal search is unlikely. Generally, the primary concern for users of horizontal search engines is that they have the best results. They do not care about how many users are currently using Google. Contrast this with platforms with known network effects, like Facebook. Users using Facebook want to be connected with their friends, and if they were not there, they would have no desire to be on that platform. Each user

17 See http://googlesystem.blogspot.com/2009/05/googles-competition-is-one-click-away.html

77 on Facebook necessarily makes Facebook a more attractive platform. This is just not true with

Google.

Hal Varian, the Chief Economist at Google, says this about network effects of data for platforms such as Google, “[t]he model is that a firm with more customers can collect more data and use this data to improve its product…. [t]his is essentially a supply-side effect known as

‘learning by doing’” (Varian, 2018). While this is true when looking at only data, the network effect appears when including users too. Users generally prefer to use a better search engine; this is the aspect which search engines typically compete over. Therefore, the best search engine will usually have more users than the second-best, and so on. By attracting more users than the second-best search engine, the best search engine continues to get better at a faster rate than the second-best. This learning by doing explained by Varian turns into an indirect network effect due to the preferences of users for the best search engine. Current users search with Google, making it better, attracting more users as Google gets better, making Google better, and so on. The magnitude of the network effect is unknown but is hypothesized to be diminishing in quality with respect to data (Varian, 2018).

However, since data is often multidimensional (Lycett, 2013), there are likely ways to make a search engine better through more effective use of said data, as opposed to collecting as much data as possible to best reap indirect network effects. That is to say; these firms do not necessarily care for the data itself; they care about what comes from the data. For example, if

Google has a new competitor, which has less data than Google, but they glean more information from said data, allowing the entrant to exploit the same indirect network effects like Google, but with less data. Therefore, there could be some form of competition for data collection and

78 analysis which may lead to an entrant overthrowing an incumbent due to the more effective distillation of data towards search quality (Gal and Rubinfeld, 2017).

Additionally, authors and journalists have suggested that access to vast swathes of data provides Google with a unique advantage against any entrants. However, does this make sense?

Data itself is not a physical asset which, when used by one agent, cannot be used by another agent. An apt comparison by Geoffrey Manne and Ben Sperry (2015) goes to say that, Google knowing someone's birthday does not stop Facebook from knowing that same person’s birthday as well. What this shows us is that while datasets may be exclusive, the underlying data is often nonexcludable and non-rivalrous (Manne and Sperry, 2015). Moreover, there is also a very insignificant cost of collecting and distributing data leading to low costs of entry and exit, as well as a small minimum efficient scale for firms who wish to collect and utilize data (Lambrecht and

Tucker, 2015). There are several commercially available data sets which can serve as a stepping stone for entering firms who wish to enter and form their datasets (Lambrecht and Tucker, 2015).

However, there may be legal or technical reasons why this may not work (Gal and Rubinfeld,

2017). Alongside the presence of commercially available datasets, there is the question of whether or not data is necessary for starting up online. In fact, in the digital economy today, the dominant firms today entered the market with no data, usually assuming that revenues would come after accruing a userbase (Manne, Morris, Stout, and Auer, 2019). The prominence of data only became vital once they had a significant userbase which advertisers would be attracted to.

Once the userbase was there, then Facebook, Google, or Amazon could focus on using its data to give advertisers better ways to advertise to its users. This “sequential entry” is very typical in the

New Economy firms as they aim to fill a niche in the market initially, and then monetize this once the base is present (Evans, 2011). These examples are primarily for firms which are

79 creating a new market but are also valid to an extent for firms which are attempting to enter an already established market.

To support these points about data network effects, Chiou and Tucker (2017) find that longer data retention has little effect on search quality of search engines. That is to say that timely data is likely more necessary for search engines than vast amounts of older data. This is important because one of the reasons why Google is assumed to maintain its market power going into the future is its amassed data. If the effect of older amassed data is small, then there is a much-dampened catch-up effect which other search engines will have to accomplish if they want to compete with Google.

Another feature of network effects is that they work in the reverse direction as well

(Evans and Schmalensee, 2017). That is to say, that if a platform would lose one user, that may induce the marginal user to quit said platform, leading to a cascading effect which leads to massive losses of users on a platform. The effect also happens with indirect network effects.

Using the Google example, if one user decided to search using other search engines, that harms

Google by making other search engines better, which will suck away other users to the other search engine.

The presence of these network effects constitutes another barrier to entry, which potential rivals would have to overcome to compete effectively with Google. Unless a firm has a new insight about how to synthesize less data into a higher quality search engine, then they would have to attempt to exploit network effects in the same way that other general search engines did before. They would also have to somehow deal with the fact that network effects tend to make the large firms larger, as they continue to improve their service because of their current market advantage. Even though network effects can result in the rapid decline of tech firms, as seen with

80 the decline of Myspace, when they are utilized well, they make it very difficult for upcoming firms to compete with incumbents who are continually improving as well.

5.2 Market Definition

This section will explain why this thesis’ chosen market definition consists of general search engines and the secondary market consisting of retail searches, which includes online merchants and retailers. This thesis agrees with other authors, who state that Google is dominant in the market of general search. However, through the use of surveys and anecdotal evidence, one will see why this thesis’ disagrees with other authors and decides that the secondary market which Google is accused of monopolizing consists not just of comparison shopping services and

Google Shopping, but also merchants and retailers.

The use of the HMT and SSNIP to figure out a market works well for traditional markets which sell physical goods for a price, but they are less helpful for markets such as general search engines. Since Google does not charge users to search using its search engine and is unlikely ever to start doing so, one must conceptualize an alternative thought experiment to derive

Google’s market. One way which Google can exercise its market power is by extracting more data than otherwise would be possible with effective competition. Another way Google could exercise its market power is by biasing its results to either promote its services or deteriorate other services on its search engine, which can affect the quality of the search engine. When

Google biases its search results enough that it has made users switch to other means of gaining the information they want, where will they go to do so? The first choice would likely be other horizontal search engines like DuckDuckGo or Bing. However, due to Google’s sizeable advantage in search quality over its horizontal search competitors, this seems unlikely. This is seen with Google’s market share in horizontal search, which is about 75% worldwide (Mangles,

81 2018). What about specialized vertical search engines? That likely depends on the query at hand.

For example, when looking for information about the height of the Calgary Tower, there likely will not be any specialized vertical search to aid with that. But when searching for price comparison services, users are free to use Google to start their search, and then select Google

Shopping or any other service which catches their eye. Users could also bypass Google altogether and go straight to services which they have heard of or have preexisting relationships.

The entire purpose of the online shopping industry is to consummate a sale; therefore, all the different types of services and resources consumers use to reach that final decision should be included in the relevant market. Therefore, this author proposes two markets, one consisting of horizontal search engines, where users prefer to start their search, which Google is dominant due to its superior algorithm and barriers to entry. The second distinct market, where competition authorities are concerned about the extension of Google’s market power is online information for retail products.

So, if Google were to bias its search results to promote Google Shopping and deteriorate rivals services, to who would users substitute? One primary assumption which has tarnished most discussions of the Google Shopping case is the assumption that users will not substitute to looking directly at merchants or retailers when they are comparing prices, they will only use price comparison services. This assumption is completely unfounded and an unnecessary additional segmentation. When someone searches for a retail good on Google, they are met with a Google Shopping box near the top of the results, as well as sponsored search advertisements, as well as the organic search results after. So, even on Google’s main search results page, there are many other places for users to substitute away from Google Shopping. Google Shopping suffers from the same revenue cannibalization as Google’s general search, whereby users can click links

82 which Google does not get paid for instead of the links in which Google does get paid. These links near the bottom may not be price comparison services but could be merchants which users can use to compare prices. Anecdotal evidence shows that 44% of users start their product searches online with Amazon, while 34% start with search engines, and 21% start their search with a specific retailer (Soper, 2015). Google’s search results page is important to competition, but there is also a fight to be the first step for consumers online product searches.

This definition diverges slightly from the market definition found by other authors such as Geoffrey Manne, Will Rinehart, and Joshua Wright who hypothesize the relevant market is a general market for “eyeballs” or users attention.18 One divergence in our definition is that each search objective category, (such as merchandise retail, informational, or services) should be considered a separate market as each will have its separate competitors. For some searches, like informational searches, Google will have market power because they provide the information better than anyone else. However, for some searches, such as flights, Google’s service, Google

Flights has small market shares compared to incumbents like Expedia and Booking.

Analogously, with online retail searches, Google is a small fish fighting against the shark

Amazon. To assume that when users search for products online, they do not consult Amazon seems ludicrous, as they have a vast depository of choices for users to compare. Even Eric

Schmidt, Google’s then-chairman, in 2014, admitted that Amazon was likely Google’s largest rival.19 There are also traditional firms that have tried to build an online presence, such as

Walmart and Target, which can be another source for users to compare prices for goods. For

18 See Manne and Rinehart (2013) and Manne and Wright (2010) for more explanation of their reasoning behind their choice of market definition. 19 See https://europe.googleblog.com/2014/10/the-new-grundergeist.html (“If you are looking to buy something, perhaps a tent for camping, you might go to Google or Bing or Yahoo or Qwant, the new French search engine. But more likely you’ll go directly to Zalando or Amazon, where you can research models and prices, get reviews, and pay for your purchase all at once.”)

83 online retail, these firms often use scrapers to gather information on rivals prices like Google, and other search engines use scrapers to gather information about what is on the web (Finley,

2018). Today online retailers can monitor and set their prices as often as they like to keep up with their rivals, which is likely driving very cutthroat competition between these firms. Leaving theese firms outside the market definition dramatically overstates the market power of Google.

This thesis does not have any definitive evidence as to this market definition, but there exists a study conducted by the EC and Deloitte in 2013 which looks at the shopping behaviour of consumers with regards to shopping comparison services (EC and Deloitte, 2013). One of the findings in this survey was that 47% of users who had never used comparison shopping tools

(around 30% of the population) had not used them because they only purchased products or services from websites they were already familiar with (p. 148). Of the same users who had never used comparison shopping services, 36% of them preferred using search engines, and 34% preferred to compare prices themselves across several websites they already knew (p. 138).

While far from conclusive evidence supporting our market definition, this information provides insights about how users use the internet to shop. Users are heterogeneous; they have differing trust of different retailers and merchants. Perhaps some users would prefer to instead find the information themselves rather than trust price comparison services, as they are generally skeptical of all websites. Alternatively, some users have a preexisting relationship with some online merchant and go to them for their online purchases for convenience.

This finding is later corroborated when the survey asks about consumers perceptions of comparison tools. When asked of the perceived functionalities of shopping comparison services and e-commerce websites, there were subtle differences between the perceived features of both.

91% of users regarded shopping comparison services of being able to compare prices (p. 188).

84 Compare this with only 40% who said this about e-commerce websites (p. 188). However, this was the only significant difference between the perceived functionalities across these services. In between 40% and 60% of those questioned said that both services could be used to compare products, read customers reviews, and buy products. If online shopping is more than just comparing prices, then these services are similar in their perceived functionality by users. Some services may be regarded as better than others in certain aspects of online shopping, such as comparing prices or finding unbiased product information, but for the entire online shopping experience, they are relatively similar.

To build on this point, marketing research done by Forrester has shown that shoppers online look at an average of three websites before making a purchase online (Allan, 2012). This can be interpreted in many different ways, one of which is that consumers wish to compare and contrast results from several different sites before deciding to purchase. For example, a consumer may start their search by comparing similar but differentiated products on Amazon, concurrently reading reviews of each product. They may choose which product seems right for them, and then move onto a price comparison service to compare prices of the specific model they want across several online retailers. Then they may choose a retailer who has checkout and shipping capabilities, and order from them, or perhaps, using the information gathered online to make their purchase in person. Or maybe they decide to merely go to Amazon because they have

Amazon Prime, and do the same process within Amazon’s infrastructure. These examples show that there are many different choices users can pursue when they want to purchase a good online depending on their personal tastes and trade-offs between convenience and finding the absolute lowest price.

85 One can see that comparison shopping services only occupy a small niche in this example, whereby they compare prices across different retailers. But, this functionality is quickly appropriated by retailers who wish to make their consumer's choices easier. For example,

ConversionXL, a marketing blog, suggests that online marketers should be comparing themselves to their rivals on their site. This is because the marketer should be aware that consumers will be comparing competitors when shopping online, and this will give them a chance to show why their choice is superior (Laja, 2012). What this example shows is that there is considerable potential crossover between the functionality of e-commerce platforms and price comparison services.

Therefore, this thesis concludes that the primary market should consist of general search.

However, the secondary market should not merely consist of shopping comparison services and

Google Shopping; it should also include merchants and retailers like Amazon and eBay.

5.3 Theories of Self-Preferencing in Search

This section begins by asking whether Google’s conduct should be deemed anticompetitive based on the test described by Krattenmaker and Salop (1986). This thesis concludes that Google’s conduct could not have created or maintained market power, and could not have harmed consumers. Next, this section questions if Google’s conduct meets the legal conditions for anticompetitive discrimination, refusal to deal, or tying as described in Chapter

Three. This thesis concludes that Google’s conduct does not meet the legal conditions for anticompetitive conduct regardless of the characterization of Google’s conduct.

5.3.1 Raising Rivals Cost and Search.

86 As discussed before, the categorization of the conduct can have a significant role to play in whether or not it is deemed anticompetitive. However, in Canada’s Competition Act, the

Competition Bureau is more concerned with the effects on competition, than classifying the conduct and determining whether it checks the boxes to deem it anticompetitive. Following

Krattenmaker and Salop (1986), the Competition Bureau starts with deciding if the dominant firm is indeed dominant in the tying market or upstream market. Once determining dominance in the upstream market, they would then look to see if the conduct significantly increased the costs of the rivals. If the conduct dramatically increased the costs to rivals, they would finally check if there was a lessening of competition because of the conduct. If the downstream firm cannot compete as effectively after the conduct, which either prevents them from entering the upstream, or harms existing firms upstream, then the conduct may have created or maintained market power.

Interpreting the theory to Google, one can interpret Google’s general search as the dominant undertaking, which could be considered “upstream” of comparison shopping services.

To then assess whether Google’s integration of Google Shopping into its general search results was anticompetitive there must be a significant increase in costs for the downstream firms, as well as an increase in market power from Google’s conduct. First, did Google’s conduct significantly increase the cost of the downstream comparison shopping services? If one looks at the EC’s decision, one will come away with the opinion that, yes, by integrating Google

Shopping into its general search results, the comparison shopping services lost significant traffic from Google’s general search results. But, if they had tried to adapt as opposed to remaining steadfast and insisting that they were owed traffic from Google, their traffic could have been saved. There are other ways to draw traffic to comparison shopping services outside of general

87 search results. If one looks at a similar market, online travel searches, one can see that Expedia and Booking have circumvented Google’s entrance into the market by advertising extensively.

The point of advertising is for users to go directly to Expedia and Booking, and not even need to consult Google at all. However, one sees that Expedia and Booking continue to enjoy a large advantage over Google Flights, even after the integration (Prieto, 2019).

Additionally, if one were to look in the same market as these comparison shopping services, one sees Amazon and eBay. As stated before, Amazon famously does not need to rely on Google to draw users to its website. Through the use of Amazon Prime and other advertising campaigns, it has garnered customers who go directly to them instead of searching Google for products. 48% of users who have prime make at least one purchase a week, compared to 13% of non-prime consumers, and if they are only using Amazon, there is no reason for them to search

Google for their product (Feedvisor, 2019). If users are going directly to Amazon, then Google’s promotion of Google Shopping cannot raise all rivals costs, just those who relied on Google for traffic. This author is by no means suggesting that these comparison shopping services can match

Amazon’s advertising budget, but rather suggests that since Amazon can draw users directly to them, it is feasible for others to replicate it on a smaller scale. If Google’s conduct did not harm other firms in the market, it signals that traffic from Google’s search results page is not essential to the operation of retail searches, there are alternative sources to get traffic.

Moreover, one could say that by promoting Google Shopping, Google is internalizing an externality it created. Before Google Universal, Google would attract users who wanted to find information on the internet and direct them to websites which could answer the queries. This direction of users could amount to an unpriced benefit to these websites since Google attracted users with its superior service but did not reap the full benefit when these other websites got to

88 advertise to the users once Google directed them. Therefore, when Google integrated its general search and its multitude of vertical search engines, it was trying to recapture this uncaptured benefit it had created with the creation of its general search engine.

Did Google’s conduct create market power by way of the softening of competition brought by raising the rival's cost? Again, if one were to read the EC’s decision on Google

Shopping, one would see that if Google’s conduct raised the cost of its rivals by way of the conduct, then Google would better be able to exercise its market power. However, this thesis argues that the EC incorrectly defines the secondary market which Google Shopping competes in, artificially delivering the result that Google can better exercise its market power if the conduct increases the costs of its rivals. When including merchants and retailers in the market, like Amazon and eBay, one sees that even if Google were to completely put comparison shopping services out of business because its conduct was so effective, Google would not be able to charge higher prices to advertisers to advertise with Google. Because merchants such as

Amazon have sponsored products in its results page too, and Amazon was untouched by the conduct since they attract users directly to them, market power was not created by Google conduct. Google would not be able to charge advertisers more to advertise with them; therefore these costs would not be passed into consumers, suggesting that there is no consumer harm due to the conduct.

The final critique of the EC’s decision in Google Shopping is that Google’s conduct is an exercise of market power as opposed to conduct which creates market power. In Chapter Three, a hypothetical situation was presented regarding Model T’s and their potential exercise of market power. One can align the Model T hypothetical with Google’s situation with Google Shopping.

In this case, instead of the market power being in Model T’s, the market power is in general

89 search, as Google has the superior search engine as well as several other barriers to quell entry.

When Google was getting started as a search engine, it did not have the resources to answer users queries itself, so it did its best to direct users towards the right answers using its proprietary algorithm. What sprouted up were many services which were specialized search services which relied on Google’s general search results for traffic. Google realized that all these independent services were getting advertising revenues from what amounted to free-riding off Google’s general search results. As Google gained more expertise and knowledge, it realized it did not need to rely on others to answer users’ queries and developed its vertical search engines, such as

Google Books, Google Maps, and Google Shopping. Therefore, bringing it back to the Model T hypothetical, these independent vertical search engines are the painter of the Model T’s. Both fulfilled a service that the dominant firm (Ford/Google) could not satisfy in its inception.

However, as time went on, and they developed expertise in the field, the dominant firm decided not to allow these independent firms to add value to its superior product and decided to satisfy it in-house. Therefore, analogously to the Model T hypothetical, Google is exercising its market power in general search to reap the value-added that independent firms used to fulfill. Google is, therefore, not creating market power by promoting Google Shopping; it is exercising its market power in general search.

If Google had the capabilities to provide these services from its inception, then it could have chosen to provide it or to let others fulfill it. Google would presumably choose the one which maximized its profits, and the disallowance of anyone else to be in the downstream market would be an exercise of its market power upstream. When Google allowed these downstream firms to free-ride off of its superior search engine and then decided at a later date that they would like to fulfill the market itself, it feels like the action is anticompetitive. Since Google shoves

90 some comparison shopping services out of business by its conduct, it raises concentration downstream, which some may interpret as an increase in market power downstream. However, the change in the market concentration comes from the exercise of market power coming from

Google’s superior general search engine.

However, one can see there are limits to the market power of Google. Even if Google dictated that no one else were to show up downstream other than Google Shopping, Google would not have a monopoly in retail searches. As expressed previously, users often go directly to

Amazon, bypassing Google altogether, signifying the weakness of Google’s power over where users go and see online. Therefore, this point is twofold; if Google wanted to make Google

Shopping the only service that showed up in its search results, this would be an exercise of market power stemming from its market power in general search. Nevertheless, even if Google were to do so, there would exist competition outside of Google’s search results as consumers are drawn to Amazon or eBay directly, bypassing Google’s search engine results page.

There are a couple of previously discussed scenarios in which Google’s conduct could have been considered anticompetitive. First, if Google were trying to put comparison shopping services out of business because they expressed interest and were likely to enter the upstream monopolistic market, then Google would be protecting its market power upstream, and the conduct should be found anticompetitive. This does not seem to be accurate, as there is no evidence that Google was worried about comparison shopping services entering the upstream market, nor any plans by comparison shopping services to move upstream.

Secondly, if by denying the comparison shopping services traffic, it somehow lessened competition between itself and its other competitors upstream. For example, suppose the comparison shopping services were partnered with Bing or DuckDuckGo, and each of them

91 benefitted each other by combining their datasets, and because of Google’s conduct, this partnership had been stopped, and Bing or DuckDuckGo had been harmed. In this situation, by

Google putting comparison shopping services out of business, it weakened competition between its general search engine and its competitors. This is an example of where the conduct is no longer merely an exercise of market power, but conduct which creates or maintains market power of Google in general search, and therefore should be deemed anticompetitive. Again, this does not seem to be accurate, as there does not seem to be any of these partnerships in existence, nor does there seem to be any presence in general search which Google would be overly worried about.

Finally, there is the scenario where there is independent demand upstream and downstream, and the monopolist upstream is trying to monopolize the downstream market with a tie. This scenario does not hold for several reasons. One would be hard-pressed to describe the demands of retail searches and general search as separate. The primary argument of comparison shopping services was that Google used to supply them traffic, which would imply non- independent demand. If the number of internet users was to grow, this would benefit both

Google and comparison shopping services, also suggesting the non-independence of their demand. Therefore, it does not seem that this scenario holds as Google and comparison shopping services seem to have intertwined demand.

The conclusion drawn from this section is that since there are other ways which comparison shopping services could have attracted traffic to their websites outside of Google’s organic search results page, then it could not have had a significant effect on the costs of comparison shopping services. Not to mention, even if Google’s conduct did significantly increased costs for comparison shopping services, there are other services which would have

92 constrained Google’s exercise of market power which would remain in the market, suggesting that the conduct did not create market power. Finally, as seen with the hypothetical, Google’s conduct amounts to an exercise of market power, because they saw the value-added of independent vertical search engines, and decided that they could fulfill that service themselves.

There was no market power created; Google’s conduct was an exercise of its market power upstream in general search. Thus, if the law in either jurisdiction decides that Google’s conduct is anticompetitive because it meets the anticompetitive standards in its jurisdiction, then the law should be considered wrong because the law should reflect the economic reality of the case.

5.3.2 Discrimination Doctrine in Search.

As a reminder, in Europe Article 102(c) of the TFEU decrees that a dominant firm cannot apply “dissimilar conditions to equivalent transactions with other trading parties, thereby placing them at a competitive disadvantage” (European Commission, 2013). Similarly, the Robinson-

Patman Act of 1936 states it is unlawful for a seller “to discriminate in price between different purchasers of commodities of like grade and quality” (Robinson-Patman Act, 1936).

One could interpret the case brought about by the EC on Google as discrimination based on some of the language used by the EC in its report. The EC uses language such as “self- preferencing” and self-favouring” to describe Google’s conduct, which hints they thought of the conduct as discrimination (Akman, 2017). For there to be anticompetitive discrimination in

Europe, there must be at least two distinct transactions which were dissimilarly treated, although they are equivalent (Italy v. Commission, 1963). Once the court confirms that two similar transactions are being conducted dissimilarly due to the seller, the court must then show that the discrimination led to the hindering of some business partners relative to others (British Airways v. Commission, 2007). Finally, the discrimination precedent was updated with MEO v.

93 Autoridade da Concorrência (2018), which stated that the court must prove that the discriminated firm was put at a competitive disadvantage due to the dominant firm.

Conversely, in the United States, the Robinson-Patman Act is almost unenforced because it disallowed differences in prices sold to downstream firms for legitimate reasons, as well as facilitating cartel-like behaviour, as described earlier (Posner, 1987). Canada followed suit and repealed their discrimination provision of the Competition Act in 2009 for similar reasons

(Musgrove, 2009).

Since it seems very unlikely Google would be charged with discrimination in the United

States or Canada, could its conduct constitute discrimination in Europe? Does Google transact with different firms about its placement on Google’s search engine results page? Not directly,

Google’s algorithm chooses placements of results based on several different criteria, such as relevance and timeliness. While it is true that there is a relationship between firms’ website being shown on Google’s results page, one would be hard-pressed to describe the relationship as a transaction. As described by Salinger and Levinson (2013), Google has only a formal arrangement with advertisers on its platform, not with users who come seeking information, and especially not websites which Google’s search results display. Although one could argue that

Google and comparison shopping services have a relationship because they allow Google to crawl and index their website, and Google displays their website for relevant queries, this falls apart for several reasons. This proposed relationship is very lopsided in favour of comparison shopping services, and all other websites which are displayed on Google’s search results.

Comparison shopping services receive what amounts to free advertising when users enter a relevant query, whereas in return Google is allowed to crawl and index their site, which is a benefit to the website since it allows them to be found by Google (Akman, 2017). The use of

94 crawlers amounts to a visit on a website, which no one would argue constitutes a relationship between two entities (Akman, 2017). Not to mention that websites can opt-out of being crawled and indexed by Google, resulting in them not showing up on Google’s results page. What this amounts to is a one-sided relationship which Google receives an infinitesimal benefit for the free advertising it provides for comparison shopping services which happen to show up on Google’s results page.

While not meeting the legal threshold for discrimination because of the questionable transactional nature of the relationship, one can see that Google did discriminate against these services, by promoting its service above others. However, this discrimination takes the form of a firm which produces a good, which emits an externality, which in its inception cannot be captured by them. There are firms which sprout up who use this externality and manage to do well for themselves off the back of this uncaptured externality. One day, the firm whose product is creating the externality develops the ability to capture, if not partially, the externality and does so. The firms which had built its business off the uncaptured externality exclaim that they have been discriminated against. Moreover, they did not try to gain independence from the externality, just expected it to be present forever. This metaphor shows the initial uncaptured externality created from Google’s superior search engine, which, in its inception, could not exploit. But, as they developed expertise, they became able to, and captured it themselves, while the firms which used the externality as a lifeline struggled and claimed they were discriminated against. This metaphor should explain why Google’s conduct should not be considered anticompetitive discrimination.

Since Canada and the United States do not prosecute for discrimination, because Canada removed its discrimination provision, and the United States has not sought a discrimination case

95 in decades, this section focused entirely on Europe. However, this thesis concludes that even in

Europe, which is actively prosecuting firms for discrimination, Google’s conduct cannot be deemed discrimination, even if the language used in the EC’s decision lends evidence that they potentially deemed Google’s conduct discriminatory. It should not be considered discrimination primarily because there were not two similar transactions conducted dissimilarly, and secondarily it can be argued that Google did not put rivals at a competitive disadvantage because they had opportunities outside of Google to attract traffic which they did not attempt to explore.

Additionally, Google’s conduct seemed to be the reclamation of an externality which itself had created and finally was able to capture.

5.3.3 Refusal to Deal in Search.

One interpretation of the relationship between Google and comparison shopping services is that users are drawn to Google because they know it has the answers the user is looking for, and that Google then passes those users off to websites to answer the user’s queries. In this context, one could interpret traffic as an input which Google supplies to websites. One could then understand how comparison shopping services can argue that by Google introducing Google

Shopping, which steals some traffic from the comparison shopping services, is then refusing to supply traffic to the comparison shopping services. How does this story shape up with the facts of the case and the legal standards of refusal to deal in Canada, the United States, and Europe?

In all three jurisdictions, the competition authorities agree that firms have the right to associate and deal with whomever they see fit. Where they differ is in the exceptions to this rule.

In the United States there is no essential facility doctrine, however, if the facts of the case align themselves with those found in Aspen Skiing Co. v. Aspen Highlands Skiing Corp. (1985), then the conduct could be deemed an anticompetitive refusal to deal. In this case, the venture the

96 firms were engaged in was presumed profitable since they willingly engaged in it, and that the discontinuation displayed that the firm was willing to sacrifice short-term profits to an anticompetitive end (Fox, 2005). Second, the dominant firm refused to sell the input in bulk (in this case, skiing tickets) to the plaintiff, even at retail price (Fox, 2005). Finally, they refused to sell the input at any price (Fox, 2005). Do these conditions hold for Google? While at face value, it may seem that Google and the comparison shopping services held a pre-existing relationship, which Google then severed, the reality is that their relationship is rather lopsided and benefitted the comparison shopping services primarily. Google’s only relationship with most websites is that of a casual user who opens a webpage, which is basically what happens when Google crawls and indexes a page. However, comparison shopping services receive what amounts to free advertising from Google as they show up on Google’s search results page for specific queries.

Did Google then give up what was a profitable engagement with comparison shopping services to pursue anticompetitive ends? This too seems tenuous, as Google likely profited more from pursuing Google Shopping than it did freely sending traffic to comparison shopping services as seen in the growth of Google Shopping once it was promoted at the top of some search results

(European Commission, 2017). Therefore, it then seems unlikely that Google’s conduct would then legally constitute anticompetitive refusal to deal in the United States.

In both Canada and Europe, there seems to exist an essential facilities (indispensability test) doctrine which means that while firms have the right to conduct business with anyone they see fit, they may be required to supply if they are found to be essential to the other firms business, and there exist no actual or potential substitutes for the foreclosed input. The input, in this case, is presumably traffic since, in the EC decision, they dedicate a significant portion of the decision to showing that Google’s conduct decreased traffic to the comparison shopping

97 services. However, are there other ways of attracting traffic to comparison shopping services? If one looks to a similar market, travel searches, one will see a similar service Google Flights, which struggles to compete with travel vertical search engines such as Expedia and Booking, to name just two. How is it possible for Expedia and Booking to be able to succeed in a market where Google has promoted its own Google Flights service ahead of them? One hypothesis is that Expedia and Booking have dedicated time and resources to advertising campaigns, whose goal was to draw traffic directly to them, instead of relying on Google. While not entirely desirable for these firms, who would likely wish to spend less on advertising to increase profits, this is just one example of other ways to draw in traffic. While it would be nice to continue to rely on free traffic Google supplies, firms should not be surprised when Google sees potential profits in peripheral markets and decides it would like to get a piece of the pie. Thus, like in similar markets, comparison shopping services could directly draw users to them using advertising campaigns. Nevertheless, they preferred to rely on the free traffic generated by

Google’s search results, which had the potential to be displaced at the drop of a hat. Even in the most extreme example where Google outright denies to list comparison shopping services on its search results, the same point applies, that there are other legitimate ways of directing traffic to themselves outside of Google. Therefore, traffic from Google cannot be indispensable; it is merely nice to have. If there are other legitimate ways of obtaining the input, then the input being denied cannot be considered indispensable, and the conduct should not legally be considered anticompetitive.

Although it appears that Google’s input does not constitute an essential facility, the EC may instead try and find them guilty of a similar abuse called margin squeezing. There is evidence that the EC will instead try and find firms guilty of margin squeezing instead of refusal

98 to deal because it is harder to successfully prosecute for refusal to deal (Petit, 2015). Margin squeezing is when a firm which is present in both the upstream and the downstream sells the input for a high price and the downstream market for a low price. Google should not be found guilty for margin squeezing because of the questionable transactional nature of the relationship between the downstream and the upstream. Google does not “sell” traffic to the downstream, and one would be hard-pressed to argue that Google sells advertisements in the downstream for a low price. Thus, the facts of the case seem not to lend themselves for the EC to find Google guilty of margin squeezing either.

Finally, Canada does not prosecute specifically for refusal to deal; they prosecute for conduct which substantially lessens or prevents competition, which refusal to deal may do. The full analysis of whether Google’s conduct was able to substantially lessen competition was discussed previously, and this thesis concluded that the conduct would not be able to substantially lessen competition. Although Canada does not officially recognize the essential facilities doctrine, recent cases, and the publishing of a non-binding guideline by the

Competition Bureau in 2019 suggests that the essential facilities doctrine may be enforced in

Canada.

There is then a gradient of difficulty in finding a firm’s refusal to deal anticompetitive. In

Europe and Canada, one must show that the refusal to deal lessened or distorted competition, thereby making a rival less competitive. If the input is deemed an essential facility to the operation of the downstream firm, then the upstream firm may be forced to supply to the downstream firm. This thesis has argued that the conduct did not lessen or distort competition, as well that Google’s input, traffic, does not constitute an essential facility for comparison shopping services. If the legal conditions are not met in the more restrictive jurisdictions of Canada and

99 Europe, then it is not met in the United States, which does not have an essential facilities doctrine. However, the refusal to deal could be deemed legally anticompetitive if it meets the narrow conditions as set out in Aspen Skiing Co. v. Aspen Highlands Skiing Corp. (1985).

Though, it seems unlikely that Google’s case falls under the conditions set out in this case.

5.3.4 Tying in Search.

Tying is one of the interpretations of Google’s conduct with its universal search engine.

Google tied its vertical search engine, specifically its comparison shopping service in this case, to its monopolistic general search engine. As discussed before, tying was demonized in the

United States until the Chicago School changed the way lawyers and judges thought about the conduct. They did so by showing in which scenarios tying could be potentially welfare improving, such as when tying was done for security or operability reasons. The question is then, does the tying done by Google fall into this welfare improving metric, or is this tying a way that

Google wanted to protect its current market power? This is relatively difficult to answer due to the moving parts and lack of data, but there are some relevant facts which can offer suggestions of the right direction. First, it is generally accepted that the implementation of universal search by Google did benefit users (Federal Trade Commission, 2013).20 Users gained the ability to get direct and easy answers to their more direct queries, rather than having to parse through related pages to find the answer. When car retailers sell vehicles with tires included, this can be considered a tie, but in many cases, the tie makes consumers better off because they have fewer menial decisions to make to get the result. The result of buying a car is the ability to drive from point A to point B, which the type of tires usually does not play a significant role in getting this

20 The FTC concludes that, “[t]he totality of the evidence indicates that, in the main, Google adopted the design changes that the Commission investigated to improve the quality of its search results, and that any negative impact on actual or potential competitors was incidental to that purpose” (FTC, 2013).

100 result. Similarly, the result of using search engines is to get answers; users do not care who gives them the answer, as long as they get their answer. One could think of Google’s search results page as asking users whom they would like to consult to answer their query, but if Google can answer this query outright, then the answer can be given to users by eliminating one extraneous step.

One relevant model which could be used to explain Google’s conduct in an anticompetitive way is from Carlton and Waldman (2002). Carlton and Waldman explain a model which shows that tying complementary products can be used to preserve and extend monopoly positions. This paper relies on Whinston (1989) which shows that when the secondary market has increasing returns to scale tying to deny minimum efficient scale can be a profitable strategy for monopolists in a primary market, by protecting against the possible overtaking of the primary market by the secondary market. A great way to see this model in action is from the

Microsoft Antitrust case in the United States, where Microsoft tied its internet browser Internet

Explorer to its monopolistic operating system Windows. There is written and spoken testimony saying that this tie was primarily done to foreclose Netscape Navigator, and not to benefit users.

Executives at Microsoft were worried that Netscape would be able to provide a layer of software over top of the operating system, which could replicate and render operating systems irrelevant.

The theory before this pointed towards the fact that operating systems and software were complements, and the presence of more software inherently made the operating systems more valuable to users, and therefore the makers of operating systems should never enter a complementary market to foreclose competition because it makes themselves worse off.

However, when Microsoft is worried about the possibility of a product rising from this market and stealing away their monopoly, they would instead take short-term losses, taking the form of

101 fewer complements for their operating system, for the longer-term gain of maintaining their monopoly position. If this same story held in Google’s case, then this could be a coherent theory of why the tie was undertaken.

There are several reasons why the conduct Google is engaged in with Google Shopping is unlike the conduct engaged by Microsoft. First is the differences in switching costs for Internet

Explorer and Google Shopping. When the Microsoft case was going through the courts, it was rather tricky to acquire and install software on computers as there was plenty of issues that could go wrong. The difficulty of installing new software gave an advantage to Internet Explorer which came preinstalled on Windows operating system, as users were likely to not go through the trouble of installing another browser if they were given one with their operating system. Contrast this with Google Shopping, and one sees that there are multiple different choices on Google’s search results page outside of Google Shopping. The costs of switching from Google Shopping is merely scrolling down the page and selecting another search result. There is no complicated installation or cost of acquiring other retail search results. Another reason Google’s scenario is different than Microsoft’s is the ability of users to get the tied good without the tying good. In

Microsoft’s case, in Windows 98, there was no option for users to uninstall Internet Explorer, which was odd considering it was possible to uninstall other preinstalled features of Windows

98, as well as the fact that it was possible to uninstall Internet Explorer from Windows 95

(Department of Justice, 1999). Contrast to Google; one sees that there is a distinct lack of coercion in the tie Google is accused of. Once again, the search results page shows many different results to each query. Perhaps if Google’s search results consisted of only Google

Shopping and then a blank page, there would be the potential of coercion.

102 Finally, there is a difference in the indirect effects of the tie on innovation. The court in

United States v. Microsoft Corp., (2000) concluded that in the long run, Microsoft’s conduct would cause reduced competition in both browsers and operating systems, which could lead to higher prices or reduced investment in software products for operating systems (Gilbert and

Katz, 2001). One would be hard-pressed to come to that same conclusion with retail searches.

The decline of comparison shopping services does not entail lessening of competition in the relevant market as there exists large players like Amazon, as well as traditional merchants who wish to step-up its online shopping presence too. This is seen in the explosion of e-commerce in recent years, which was conveniently missed by the EC’s narrow market definition (Kucharczyk,

2015). Although one cannot know the counterfactual for this, to determine whether this level of investment is lower than a world where Google did not introduce Google Shopping, the fact that investors were willing to invest signals that they had hope that e-commerce would continue to grow.

Additionally, for one to try to align the facts in the Microsoft case to the Google case, one would have to make the argument that Google was worried about the potential that comparison shopping services would be able to leapfrog Google in general search. As mentioned previously,

Microsoft was anxious about Java and Netscape Navigator rendering the operating system platform redundant because of their ability to lie on top of the operating system. Therefore, their conduct was primarily in protection of its market power in operating systems, less so about monopolizing internet browsers. One would be hard-pressed to assign this same intent to

Google’s conduct, as the possibility of comparison shopping services coming for Google’s market power in general search is absurd.

103 In Europe, four conditions must hold for a tie to be deemed legally anticompetitive. The first three conditions are simple to check off, as they are true or not. The fourth condition is more contentious as it can change depending on different beliefs of current market conditions. As a reminder, the four conditions are: the tying and tied good are considered separate products; the firm is dominant in the tying market; the tie does not give consumers a choice to get the tying product without the tied good; tying forecloses competition (European Commission v. Microsoft,

2004).

Should Google Shopping and Google’s general search be considered two separate products? The presence of firms who only provide comparison shopping services as opposed to both suggests that the two markets are indeed distinct. One would also point out the difference in perceived user functionality between the two services. While one could use both comparison shopping services and general search to search for retail goods, (which the general search would then direct the user to its service or another service), the overlap in functionality does not go the other way. One cannot search comparison shopping services for information on baseball games like they could on general search engines. This too lends evidence that the services should be deemed distinct (Akman, 2017). However, there is also a suggestion that they constitute the same market. When a user inputs a query for a retail good such as “Canon camera,” Google does not know if the user is looking for the history of Canon, looking to buy a Canon camera, or looking for prices of Canon cameras. If in response to that query, Google were not to include results which direct users to merchants who sell Canon cameras (or its own Google Shopping) it likely will not be delivering the most relevant results to its users (Akman, 2017). This suggests that showing items to be purchased in response to a product search is a primary function of general search, not complementary as suggested before. Since there are arguments for each side, for the

104 sake of argument, this thesis will accept that general and product searches are distinct markets because the argument put forward by this thesis does not hinge on this fact being accurate.

The next condition is that Google is dominant in the tying market. There is not much needed to show this condition is fulfilled as it is generally accepted that Google is dominant in general search. The dominance is assumed because of the barriers to entry which exist in general search, such as economies of scale, brand/loyalty effects, and network effects. Google’s dominance is then suggested in its market share, which in Europe is upwards of 90% (European

Commission, 2017). This thesis does not contest this condition.

The third condition is the ability to access the tying product without the tied product. On

Google’s results page, there is a certain lack of coercion pushing users towards a particular link.

The only differentiating factor is the placement, which tells the user something about the relevance of the page to their query. When users enter a product search, they would be met by

Google Shopping at the top, and then the regular ten blue links after. The only force stopping users from looking down the search results page is their bias to do with links which appear higher up. This bias towards results which appear higher contends with the fact Google Shopping only shows up for a handful of queries on Google’s general search. That is to say that users can avoid summoning Google Shopping by carefully selecting their query. Not to mention that the internet is an open system, and users can find their way to comparison shopping services in other ways. One such way would be to use advertising to draw users directly to them, instead of relying purely on free traffic from Google. Although the use of default bias by Google can be thought of as coercion, this thesis will argue that the bias is overwhelmed by the open nature of

Google and the internet as a whole. If users happen to have Google Shopping show up in their search results page, something can be said about coercion. But, since Google Shopping is only

105 triggered some of the time, and the fact that users can find their way to other comparison shopping services without the use of Google, there seems to be a lack of coercion on behalf of

Google, and this condition should not be satisfied.

Even though the tie should not be deemed legally anticompetitive since the third condition is not fulfilled, for the sake of completeness, next follows a discussion of whether the tie foreclosed competition. This thesis argues that the tie will not foreclose competition for two reasons. First, users can find their way to comparison shopping services outside of Google’s search results page. That is to say, that users do not need to use the dominant tying product

(Google general search) to find comparison shopping services, which can nullify the effect of

Google’s dominance in general search. Second, even if comparison shopping services were completely removed from the market because of the conduct, there exist other competitive restraints of Google. If one were to follow the market definition that the EC finds, which consists of Google Shopping and comparison shopping services, then one could see the removal of all competitive restraints because of the conduct. However, as this thesis has argued, the EC’s market definition is too narrow, and it ignored merchants such as Amazon and eBay, as well as brick-and-mortar retailers who are stepping up their online presence. Once these other firms are included in the market definition, one sees the presence of multiple competitive restraints to

Google’s exercise of market power. Therefore, this thesis concludes that Google could not have foreclosed competition with its tie because of the presence of other competitive restraints, and the fact that one need not use Google’s monopolistic general search engine to get to comparison shopping services.

Finally, Canada’s Competition Act worries about conduct that creates market power and specifically does not mention any specific conducts and conditions where the conducts would be

106 deemed anticompetitive. Therefore, the Competition Act is worried about the same effect as

Europe’s final condition of anticompetitive tying, that the tie “forecloses competition.” This paragraph then finds the same conclusion as the last condition of Europe’s tying provisions; that

Google’s conduct did not create market power for two reasons. First, that since there are other ways to drive users to comparison shopping services, outside of Google’s free search results, the comparison shopping services could have circumvented Google’s monopolistic search engine.

Second, even if the tie was strong enough to shut down all comparison shopping services, there exist other services which would restrain Google’s exercises of market power. For both these reasons, Google’s tie could not have created market power, and therefore should not be deemed anticompetitive.

So, in the United States, Google’s tie should not be deemed anticompetitive because it does not follow the findings in the similar case, United States v. Microsoft Corp. (2000).

Similarly, in Europe, Google’s tie should not be deemed legally anticompetitive because there lacks coercion between the tie of Google and Google Shopping, not to mention the suggestion that even if all the prior three conditions held, it does not seem that the tie could foreclose competition. Finally, in Canada, the conduct does not seem to create market power, which would render the conduct anticompetitive.

5.4 Comparison Between Findings of EC and This Thesis

This thesis agrees with the majority of the facts put forward by the EC in their decision, but there are a select few differences in opinion. First, the EC incorrectly defined the secondary market when it only included Google Shopping and comparison shopping services. This thesis has concluded that the secondary market should consist of not only Google Shopping and comparison shopping services, but also online merchants and retailers. By the EC defining the

107 market too narrowly, it incorrectly attributed harm to consumers by Google’s conduct, which is the second difference of opinion. When the secondary market is defined only to be comparison shopping services and Google Shopping, one sees that the EC can tell a story about how when

Google knocks out comparison shopping services, it can exercise market power and harm consumers because there exist no other substitutes. However, when the market is defined correctly, one sees many constraints of Google’s market power, which would protect consumers against exercises of market power by Google.

Lastly, the EC decided that Google’s conduct created market power in the downstream market. Generally, if a firm has a monopoly in the upstream and decides to foreclose a firm in the downstream, the conduct is an exercise of market power stemming from the monopoly upstream. The conduct is anticompetitive if it creates market power by softening competition between the upstream firms, or prevents the entry of a firm into the upstream. Since these two scenarios do not seem to be true, one should conclude that Google’s conduct is an exercise of market power, which competition authorities should not find anticompetitive.

5.5 General Comments about the EC Decision

The main thrust of the argument against the EC’s decision is that Google’s conduct would not be able to create market power because of the presence of other ways to direct traffic to comparison shopping services, the presence of other services which would constrain Google’s exercise of market power even if the comparison shopping services were knocked out, and the fact that the conduct does not create market power but is an exercise of market power. However, there are other points to mention on the EC’s decision. One of the significant assumptions made by the EC was that due to the conduct, Google needed not to invest in future improvements of

Google Shopping because they could coast by on prioritizing Google Shopping. This is not the

108 case as Google has continued to release updates for Google Shopping. For example, in March of

2018, Google released ShoppingActions, which helps users shop using .21 If

Google were content and expected no competition because of their conduct, they would not continue to update Google Shopping. This is consistent with the view of the New Economy given by Economides (2006) which expects that today's dominant firms are continually searching for the next vital market to move to after their current market is irrelevant due to changing consumers preferences. Not only did Google continue to invest in their shopping service, but they also continue to invest in R&D overall, only being beaten out by Amazon in

R&D spending in 2017 (Molla, 2018). Google is thus not acting the way economists tend to believe monopolists act.

One common gripe about Google’s conduct regardless of the anticompetitive nature of it is that Google can censor or demote anything it does not like. That is to say that if Google were to censor something or someone, then they are virtually erased from the internet. This sentiment is echoed in the EC’s decision, where they spend much time showing how irreplaceable traffic from Google is for comparison shopping services. The changes to Google’s algorithms affected all comparison price services equally, Google did not pick and choose which ones would be demoted by their algorithm changes. Also, these comparison shopping services had no formal contract formed with Google; they merely relied on Google for traffic. The primary issue with many of these comparison shopping services is that they put all their eggs in the Google basket.

As stated by Ana Hoffman, a who writes about website traffic, Google makes hundreds of algorithm changes a year, any of which can decimate any business. While the change could be

21 See https://www.blog.google/products/ads/shopping-actions/ (“It’s clear that people want helpful, personal, and frictionless interactions that allow them to shop wherever and however they want -- from making decisions on what to buy, to building baskets, to checking out more quickly than ever before. Put simply, they want an easier way to get their shopping tasks done.”)

109 an accident, having all traffic from one source is likely to go wrong. Think of a hypothetical example where Google tries the same conduct on travel vertical search engines. Google promotes

Google Flights above all other results (which they do) and deteriorates search results for Orbitz,

Booking, and Expedia. These firms expand their advertising using other means to drive users to go directly to them, as opposed to going through Google. One sees that this advertising does work, as currently, Expedia, gets 42.9% of their traffic from direct sources and 44.3% of their traffic from Google.22 The story of Foundem could be wildly different if they diversified their traffic sources and had backup sources of traffic. In online retail searches, it is suggested that users are directly going to merchants such as Amazon at substantial rates, so this shows that traffic from other sources was not unachievable like the EC states.

Another point which the EC dismissed outright from Google was that the decline of comparison price services could have been caused jointly by other services. One crucial assumption is that there was nothing else that was happening over this time period, which would compound this reduction of traffic to comparison price services. Because of the EC’s market definition, they missed an explosion of investment in European e-commerce ventures in 2015. A sample of 500 private e-commerce companies in Europe saw a 4.5 times increase in their capital inflows between 2012 and 2015 (Kucharczyk, 2015). Due to the market definition invoked by the EC, they get to conveniently ignore this boom in investment and competition from European e-commerce, which could have affected the popularity of comparison shopping services too. This potential decline of comparison shopping services would have also affected its PageRank score, which could additionally explain why the comparison shopping services fell in Google’s search results page.

22 See https://www.similarweb.com/website/expedia.com#overview Retrieved Oct. 22, 2018

110 Another issue with the analysis conducted by the EC is that it did not weigh the positives of the changes Google made to its algorithm and search engine. Both the algorithm changes and the implementation of Universal Search were made to make Google a better search engine for users. The algorithm changes are crucial in a search engine as there are nefarious players in the market who wish to subvert Google’s algorithm and end up with high ranking undeservedly.

Without the continual changes to Google’s algorithm, these subversive players would end up with top rankings when it is likely they are not very relevant for the user’s inputted query. These changes are primarily made for the user's benefit, as ad-driven websites compete for users by providing the best services they can.

Google’s blend of its general search engine with its vertical search engines led to greater ease of finding the information which users sought from Google. As stated before, general search was often the starting point for finding who can answer queries for the user. These answers often were answered by vertical search engines which specialized on a subset of information. With the amalgamation of vertical and general search, users no longer had to search many times for the info, as the Universal Search would be able to parse and provide the information required. When users searched for “who wrote Harry Potter” they do not want to see a list of links which directed them elsewhere, they want a name, which was now provided by the general search engine. There is no doubt that the combination of vertical and general search rendered some websites redundant, but if it made the users experience online better, then this is a textbook example of

Schumpeter’s creative destruction.23 Just like the horse-drawn carriage was made redundant by motorized vehicles, websites which provided common information for users were replaced by search engines which could directly answer these queries. The primary point being, that even

23 See Capitalism, Socialism and Democracy (1942) by Joseph Schumpeter for the origins of this idea.

111 though providers of those low-value websites are made worse off, just like the blacksmiths who made horseshoes, consumers as a whole are made better off due to the implementation of new technology.

The value created by search engines like Google is seen in a recent paper by Brynjolfsson et al. (2018) which shows that people would not be willing to part with search engines for a year unless they were paid an average of USD 17,530 a year. Another way to put this is that an average person would rather have access to a search engine for a year than have USD 17,530 in their pocket and not have access to a search engine. The value of search engines seen by users would not be anywhere as high as this if it were not for their sustained investments and innovations. So, for the EC to only see Google’s conduct as a negative, without acknowledging the positives which it creates is a disservice to consumers.

Another concern that is less voiced by the decision that the EC makes, but is nascent in today's discussion of antitrust and New Economy firms is those aforementioned unknown innovation harms. If due to Google’s conduct, they eliminate some of these comparison shopping services, which may have gone on to innovate and develop some new technology or service which would have benefitted consumers greatly, then these costs should be mentioned in the cost-benefit analysis of this case. There is apparent difficulty in this as one cannot see a world where the conduct happened and one where the conduct did not occur; the counterfactual is unknown. Some authors have then argued from this fact that antitrust agencies have then underenforced antitrust laws because they tend to employ short-term price-focused measures when looking at the results of the conduct. They point to the fact that Google was exonerated in the United States and Canada as an example of this underenforcement. A primary reason why this point is suspicious is that the improvement of Google’s general search engine to integrate its

112 vertical search engines with its general search engine was an improvement which primarily benefitted consumers, but which had a side-effect of rendering some websites redundant. Users reach out to search engines for answers to questions, and before the integration of vertical and general search on Google, users were given links in response to questions. Consumers benefitted from faster, more direct answers to relatively simple queries, but this was at the behest of some relatively low-value websites traffic. So, this is then a case of creative destruction where potentially rival firms are harmed, but consumers are benefitted due to the change. A change would be more suspect if there were no tangible benefits to consumers, and the conduct primarily hurt rivals, but this is not true for Google’s conduct.

The point that consumers are the primary benefactors of this change is seen in the fact that the primary complaints of Google’s conduct came from industry rivals, not consumers. From a document created by the creators of Foundem, the vertical search engine in the United

Kingdom, when they filed their complaint against Google, Microsoft had already taken up a complaint against Google.24 As mentioned before, antitrust law is supposed to protect consumers from anticompetitive actions of monopolies, not rivals from superior products. The fact that of nineteen complainants that formally complained to the EC about Google’s conduct, zero were representing consumers at all should be very suspicious (McAllister, 2015). Not only is it questionable that most, if not all, complainants were rivals of Google, but most had non- significant business in Europe during the time of the conduct. Examples of such companies are

Expedia, Nextag, TripAdvisor, and Yelp were formal complainants of Google, presumably vying for an increased presence in Europe that would result from a guilty verdict (McAllister, 2015).

While a fair point would be that if the conduct harmed rivals, forcing some to cease business,

24 See http://www.foundem.co.uk/Foundem_Google_Timeline.pdf (page 18)

113 then this is reducing the ability of rivals to respond to Google’s potentially consumer-harming behaviour—this thesis has previously explained why this story does not hold water.

The determination of Google’s conduct being anticompetitive has more long-term negative effects. If the EC denies Google’s appeal, then the lesson they will have learned is that they must be careful not to harm rivals when they make changes to improve their product. One issue with this is that if Google is made better, it negatively harms rivals, by way of attracting more users to Google. If by improving Google Search, they may be fined for doing so, this dramatically reduces the incentive to improve Google. Who wins by making this change?

Google’s rivals who are free to innovate as they please knowing that Google is likely afraid to innovate due to fears of being fined. Who loses? Consumers lose as Google receives fewer improvements than before. One may argue that this is good, let other search engines catch up to

Google, create more competition. However, they misunderstand that these New Economy firms often become large and dominant for relatively short periods, when a new company shifts the market in such a way which renders the monopoly irrelevant. Therefore, if there was a company which was currently the dominant firm, and that company was fined or lambasted for innovating, other fringe firms would see this and see that by them innovating they may also be fined like this when they are at the top. What this does is teach firms that they can get the government to hurt their rivals because of innovative conduct, incentivizing firms to lobby and pursue legal avenues as opposed to competing on the merits of their products. One should not have to explain why this is an undesirable outcome for society, as firms will end up directing money which could have been used in more productive ways, such as research and development, towards lobbying.

Despite all the evidence which has been laid out in this thesis, the EC’s decision to fine

Google could still be considered legally justified. If the EC is much more sensitive to Type I

114 errors than Type II errors, then the choice they have made could perhaps be understandable. As a reminder, Type I errors in an antitrust context are wrongful anticompetitive prescriptions, while

Type II errors are a failure to prosecute anticompetitive behaviour. Therefore, when one says that if the EC is more responsive to Type I errors, that is to say, that they would prefer to over- enforce, and potentially prescribe anticompetitive behaviour to potentially pro-competitive behaviour. As evidence for this preference, Devlin and Jacobs (2010) point to just a few recent cases in Europe which have shown this preference for over-enforcement. They point to the divergence from the United States reflect differences in experiences and beliefs more than economics. As stated before, from Judge Easterbrook and Robert Bork, the United States has held a different preference about which error they would prefer. As Judge Easterbrook puts it, this preference came from an assumption that “the economic system corrects monopoly more readily than it corrects judicial errors” (Easterbrook, 1984, p. 15). This assumption that monopolies are short-lived and markets are self-correcting may not be perceived to be true based on past experiences in Europe.

Another reason the decision could be legally justified in Europe is that the EC has stated that the goal of its competition law is to protect consumers. However, the ECJ has “not explicitly endorsed this interpretation” (Witt, 2019, p. 37). Alternatively, while the EC has said that competition law is to protect consumers, the courts have in recent cases “primarily or even exclusively assess the conduct’s foreclosure effects and only mention the impact on consumers in passing” (Witt, 2019, p. 11). So similarly to the Google Shopping case, it seems that the competition authorities have focused on showing harm to competitors, and then translating that harm to consumers. This jump in logic from harm to rivals, to harm to consumers is questionable, since rivals can be harmed for a multitude of procompetitive reasons as well. When

115 a firm further improves its product over its rivals, that reduces rivals revenues, but it also benefits consumers because of a better product. Even though the jump from rival harm to consumer harm is questionable, it could be considered legally since “Article 102 only outlaws the exclusion of competitors that are less efficient than the dominant undertaking” (Witt, 2019, p. 42). Therefore, according to the case law, European competition authorities would have to show that the plaintiff is as efficient than the dominant firm, then the exclusionary conduct could be deemed anticompetitive. This provision is a reduction in the scope of Article 102 since it used to consider any exclusionary effects which came about because of the conduct anticompetitive, but it is still far from connecting the harm to rivals to consumer harm.

The final reason why this case could be legally justified, even if it is not economically justified is that Europe has an additional type of abuse of dominance, which is called exploitative abuse (United Brands v. Commission, 1976). This exploitative abuse could more nominally be called excessive pricing (Witt, 2019). One can think of this excessive pricing as an exercise of market power since the firm would be unable to price excessively if it did not possess market power. Therefore, Europe deems exercises of market power which harms consumers as anticompetitive, while Canada and the United States only consider conduct which creates market power as anticompetitive. So, if Europe considers conduct which is an exercise of market power as anticompetitive, and this thesis has determined that Google’s conduct was an exercise of market power as opposed to conduct which creates market power, then Europe could be legally justified to find Google’s conduct anticompetitive. This difference in opinion on exercises of market power could arrive from each authorities beliefs on the markets and worries about equity

(Gal, 2004). In the United States, there is a common belief that monopolies are relatively short- lasting as competition tends to enter and drive down market power over time (Easterbrook,

116 1984). However, the EC seems to take the belief that markets either take too long to correct themselves, resulting in harms to consumers from exercises of market power or that markets do not correct themselves at all over time (Gal, 2004). One can make the argument that allowing exercises of market power enables a firm to get economic profits in the short-run, which can offset the large R&D costs which they have incorporated to develop their product. Also, the short-term economic profit can serve as a motivator for other firms who witness current monopolistic firms to enter and seek the same rents. This same motivation has been used in the patent system, where firms receive a government-protected monopoly for the production and licensing of its intellectual property, which they can then use to charge whatever they see fit, in most cases. Removing this incentive can have the effect of reduced innovation, which dramatically harms consumers in the long-run.

5.5 What’s Next for Google and the EC

When Google introduced Universal Search, it integrated many different vertical search engines into its general search engine, only one of which is Google Shopping. There are many other markets which Google touched and repeated the same conduct in which they may be found guilty for the same conduct. For example, Yelp, a company which crowd-sources reviews and creates directories of local businesses, has argued that Google has abused its dominant position in search to promote Google’s own review and business directory service. The similarities are striking, Google is dominant in general search, and thanks to the integration of Google Maps into

Google’s general search results, Yelp has experienced lessened traffic, which affects the ability of it to compete. Another similar point between these scenarios is the absence of consumer harm.

Considering the hypothetical situation that Google Maps completely wipes out Yelp, will Google be able to charge monopolistic prices for restaurants to advertise locally? This seems rather

117 unlikely, as anecdotal data suggests that more restaurants are advertising with social media companies and influencers to draw in customers (Filloon, 2018). Also, there is the issue that

Yelp (and other vertical search engines) often rely on Google’s organic search results, then when it changes, suggest that there is no other way to bring in traffic. This is not true as there are ways to avoid Google’s ecosystem altogether, and direct users directly to you, such as advertising and usage of phone applications. While having what amounts to free advertising from Google is nice to have, relying on it is a mistake, as Google has no relationship with websites which show up on its search results.

There are multiple examples of integrations which Google has tried, which have failed to be nearly as successful for Google as Google Shopping. For example, there is Google+, as well as , which was meant to provide video conferencing which integrated with

Google Calendars. Each of these examples had been integrated with successful services like

YouTube and Google Calendars, yet they did not manage to retain large market shares. There are many arguments about why these Google integrations failed, but they were still a part of the dominant Google ecosystem, which should have provided them an exclusive advantage over rivals. Zoom Video Communications was able to sweep users away from incumbents Skype for

Business and Google Hangouts, by having a more seamless and integrated service, which stuck with young professionals (OWLLabs, 2018). The story of Google+ is less so one of a new incumbent beating Google, but, Google failing in its integration. Google+ was the social media which integrated with Google accounts and more importantly, YouTube. The integration of

YouTube and Google+ should have provided Google the leverage to compete in social media, but this did not happen. There were several reasons for the failure, as described by former

Google employees, such as Google+ was a solution to Google’s problems, rather than being for

118 users, the move to mobile was too slow and was too much like Facebook (Eedicicco, 2015).

Even with the leveraging ability of YouTube, Google+ could not take off as a success in its own right. These points are brought up to show that Google’s integrations into different complementary markets are not destined to be successful; there are spaces for crafty and innovative entrants to succeed in these markets.

These are only a few examples of Google’s integration brought upon during its Universal update, which integrated its separate vertical search engines into general search results. If one can name a vertical search engine which Google operates in, it can find someone accusing

Google of monopolizing another market. However, as stated before, Google’s conduct with integrating Google vertical search engines into general search can be seen as an exercise of

Google’s market power in general search. When Google was created, it did not have the expertise to offer solutions to queries, so it directed users to websites and services which could.

As more users came online, and as there became an uncountable number of websites, more users went to general search engines to parse the internet for what they wanted to see. Websites which relied on Google’s search results for traffic sprung up, as it was easy to make advertising revenues if Google was directing users to you. Services such as lyric websites and celebrity birthday websites sprung up as users consulted Google for basic answers to easily answerable questions. As Google gained expertise, it decided that it should not allow these low-value websites free-ride off its proprietary algorithm, and decided to tackle answers to these basic questions itself. By doing so, Google made itself able to directly answer more queries, not requiring users to click off to find somewhere else with the easy answer they sought. These websites who had relied on traffic from Google were made worse off since less traffic meant less advertising revenues. However, these links in what Google calls the OneBox, come from the

119 algorithm too, and at least for some prompts, it seems that Google is licensing out the information, paying at least a little bit to the firms who are being shown in the OneBox

(Masnick, 2019).

In this story, there is palpable harm to these websites, but does that translate to harm to consumers? For there to be harm to consumers, one would have to show that by Google eliminating these websites, they could charge more for advertising, which then could be translated to higher prices to consumers. This is tenuous, as there are other places for advertisers to advertise on the internet, not to mention the possibility of offline advertising. This scenario completely ignores the reality that Google being able to answer these basic questions made consumers better off since they directly answer the query instead of siphoning off users to someone else who can tell the answer.

Google’s integration with Google Shopping is one step in its goal to be able to answer any query thrown at it. This ability would be a massive innovation, which consumers would benefit significantly from having. However, if Google cannot answer questions itself, because by doing so eliminates websites which initially relied on Google for traffic, then this benefit will not be realized, leaving consumers worse off. If the EC’s decision is incorrect for Google Shopping, then the same inappropriate facts will likely carry over to the next Google investigation, which has a possibility of coming to the wrong conclusion because of these incorrect facts. The lesson learned from Google is then to stop innovating its general search engine to try and answer questions, and leave it be, which hurts consumers because they will not have the best version of

Google available.

120 Chapter 6: Conclusion

There are several issues with the EC’s case against Google, which, in this authors opinion, has led them to an improper decision in Google Shopping. First, the market definition defined by the EC is too narrow and misses potential restraints to Google’s exercise of market power. This thesis finds the market to not only include comparison shopping services and

Google Shopping but also merchants and retailers such as Amazon and eBay. Considering 44% of users start their search on Amazon when they are searching for retail goods online (Soper,

2015), as well as users generally having similar perceptions of features of all these services, users consider these as similar enough to be in the same market.

Second, from the EC’s inappropriate market definition comes about a flawed theory of harm. The theorized harm to consumers is that by way of prioritizing Google Shopping, Google can strangle the life from comparison shopping services. Once they are disposed of, then Google will be free to charge higher prices to advertisers to advertise on Google Shopping, when then the advertisers will be able to pass on higher prices to consumers. However, this theory of harm relies on the market being narrowly defined, as when Amazon and eBay are included, Google’s potential exercises of market power seem to be sufficiently constrained. Additionally, there is the fact that comparison shopping services should be able to draw traffic in from other sources than

Google’s search results, as many other services have done in the past in response to Google’s integration in similar markets. Therefore, if comparison shopping services could have attracted traffic in alternative ways, circumventing Google altogether, and even if they were foreclosed entirely, other rivals would be able to restrain Google’s exercises of market power sufficiently, then the conduct should not be deemed anticompetitive. Google’s conduct thus should be seen as

121 an exercise of its market power in general search, as it decided instead of allowing other rivals to provide retail information, it could exploit higher profits if it delivered retail searches itself.

Lastly, the conduct should be considered an exercise of Google’s market power in general search, not conduct which creates or maintains market power. Because of Google’s market power in general search, Google has the power to decide how the market operates downstream of its monopoly. If Google finds that it is cheaper to allow downstream firms to operate, then Google will allow it to happen, and charge the downstream firm the “monopoly price” for appearing on its results page. However, if Google decides that it can provide the downstream market better itself, then it is its prerogative to do so, because of the possession of its superior general search engine. While this may look like it creates market power downstream, going from many providers to potentially only Google, the ability to do so stems from its market power upstream. This conduct cannot be said to create market power unless by Google eliminating these comparison shopping services, it has weakened competition upstream in general search, or it eliminated a potential competitor who was looking to compete against

Google upstream. Neither of these two cases holds with Google and comparison shopping services, ergo, Google’s conduct is an exercise of market power, not conduct which creates or maintains market power.

However, even if the conduct should not be considered anticompetitive economically, it could be deemed legally anticompetitive in Europe for three reasons. Europe seems to be rather skeptical when it comes to markets being able to correct themselves of monopolies over time

(Devlin and Jacobs 2010). In the United States, the conventional wisdom is that monopolies are easier to correct for than incorrect government decisions. This comes from the belief that monopolies are naturally unstable, especially in tech and highly innovative sectors, therefore

122 when competition authorities are uncertain about whether a conduct is anticompetitive, they should err on the side of underenforcement, as these monopolies tend to correct themselves over time. Europeans either do not believe this to be accurate or to a lesser extent than Americans.

Thus, it could make sense that in a case where there is much unknown, that Europeans err on the side of over-enforcing, as they are skeptical of the transient nature of monopolies.

Next, the decision could be considered legally justified is that while the EC’s stated objective is to protect consumers, it is also intent on protecting other competitors, as seen in recent cases.25 In several recent cases, the ECJ and EC have focused almost entirely on how the defendant's conduct has harmed the plaintiffs business, and then weakly translated the harm to rivals to harm to consumers. The fact that the ECJ and EC have been defending rivals instead of primarily consumers makes sense when the ECJ has “not explicitly endorsed this interpretation” that the primary goal of Article 102 is to protect consumers (Witt, 2019, p. 37). Therefore, if a secondary or equally important goal is to protect rivals, then this case could make sense because

Google harmed comparison shopping services when it introduced Google Shopping. However, the connection between rival harm and consumer harm is tenuous for the same reasons mentioned above. If the ECJ and EC were worried about harm to rivals, then the decision could make sense legally.

Finally, the last reason this case could make sense legally justified is that the EC also finds exercises of market power as anticompetitive, as seen in United Brands v. Commission

(1976). Canada and the United States deem the exercise of market power as a necessary part of the innovative process, where firms innovate, they are allowed to reap the benefits of its superior product until their patent is up, or someone innovates them out of the market. The only conduct

25 See for example, Commission v. Motorola, 2014, Commission v. ARA, 2016, and Commission v. Baltic Rail, 2017

123 which is then considered anticompetitive is conduct, which creates market power. As discussed,

Google’s integration can be thought of an exercise of market power because of its superior general search engine, which then rendered some websites and services irrelevant. Therefore, if the conduct could be considered an exercise of market power, and Europeans find exercises of market power anticompetitive, then Google’s conduct could be construed as anticompetitive.

Thus, the decision by the EC could be believed legally justified based on the current provisions of European jurisprudence and case law. However, the decision cannot be economically justified, as it does not create market power for Google, nor does it harm consumers. Consequently, this is why Canada and the United States disagreed with the EC’s decision to fine Google.

124 References

American Telephone & Telegraph. (1909). Annual Report of the Directors of the American Telephone & Telegraph Company to the Stockholders [Press release]. Boston, MA: Geo. H. Ellis, Printers:.

Ahlborn, C., Evans, D. S., & Padilla, A. J. (2004). The antitrust economics of tying: a farewell to per se illegality. The Antitrust Bulletin, 49(1-2), 287-341.

Akman, P. (2017). The theory of abuse in Google Search: A positive and normative assessment under EU competition law. U. Ill. JL Tech. & Pol'y, 301.

Arriola, B. (2017). Google's Big Daddy Update: Algorithm and Infrastructure Improvements. Retrieved from https://www.searchenginejournal.com/google-algorithm-history/big-daddy- update/

Aspen Skiing Co. v. Aspen Highlands Skiing Corp., 472 U.S. 585 (1985)

Baker, J. B. (2003). Responding to developments in economics and the courts: Entry in the merger guidelines. Antitrust LJ, 71, 189.

Bain, J. S. (1956). Barriers to new competition: their character and consequences in manufacturing industries (Vol. 329). Cambridge, MA: Harvard University Press.

Bork, R. (1978). The antitrust paradox: A policy at war with itself.

Bill Gates Deposition Transcript – Part One [Interview by Houck]. (1999, November 2). Retrieved from https://www.washingtonpost.com/wp- srv/business/longterm/microsoft/documents/gatespart1.htm

Blake, T., Nosko, C., & Tadelis, S. (2015). Consumer heterogeneity and paid search effectiveness: A large‐scale field experiment. Econometrica, 83(1), 155-174.

British Airways v. Commission, Case C-95/04 P (European Court of Justice, 2007)

Bronner v. MediaPrint, Case C-7/97 (European Court of Justice, 1998)

Brynjolfsson, E., Diewert, W. E., Eggers, F., Fox, K. J., & Gannamaneni, A. (2018, November). The Digital Economy, GDP and Consumer Welfare: Theory and Evidence. In ESCoE Conference on Economic Measurement, Bank of England (pp. 16-17).

Bureau. (2019, March 7). Abuse of Dominance Enforcement Guidelines. Retrieved from https://www.competitionbureau.gc.ca/eic/site/cb-bc.nsf/eng/04420.html

Carlton, D. W. (2005). Barriers to entry (No. w11645). National Bureau of Economic Research.

Carlton, D. W., & Waldman, M. (1998). The strategic use of tying to preserve and create market power in evolving industries (No. w6831). National Bureau of Economic Research.

125 Chiou, L., & Tucker, C. (2017). Search engines and data retention: Implications for privacy and antitrust (No. w23815). National Bureau of Economic Research.

Church, J. (2018). The Lamentable Rise of an Expanded Essential Facilities Doctrine in Canada: The Troubling Economic Foundations of the Toronto Real Estate Board Decision. CCLR, 31, 122.

Church, J., Gandal, N., & Krause, D. (2008). Indirect network effects and adoption externalities. Review of Network Economics, 7(3).

Church, J. R., & Ware, R. (2000). Industrial organization: a strategic approach (pp. 367-69). Boston: Irwin McGraw Hill.

Cole, M., & Ponsay, J. (2018, April 23). EU Court's Analysis of "Competitive Disadvantage" in Rare Price Discrimination Case. Retrieved from https://www.covcompetition.com/2018/04/eu- courts-analysis-of-competitive-disadvantage-in-rare-price-discrimination-case/

Competition Act, R.S.C., 1985, c. C-34

Creswell, J. (2018, September 03). Amazon Sets Its Sights on the $88 Billion Online Ad Market. Retrieved from https://www.nytimes.com/2018/09/03/business/media/amazon-digital-ads.html

Demsetz, H. (1968). Why regulate utilities?. The Journal of Law and Economics, 11(1), 55-65.

Demsetz, H. (1982). Barriers to entry. The American economic review, 72(1), 47-57.

Devlin, A., & Jacobs, M. (2010). Antitrust error. Wm. & Mary L. Rev., 52, 75.

Dixit, A. (1980). The role of investment in entry-deterrence. The economic journal, 90(357), 95- 106.

Eadicicco, L. (2015, April 26). Why Google failed, according to Google insiders. Retrieved from https://www.businessinsider.com/what-happened-to-google-plus-2015-4

Easterbrook, F. H. (1984). Limits of antitrust. Tex. L. Rev., 63, 1.

Edelman, B., Ostrovsky, M., & Schwarz, M. (2007). Internet advertising and the generalized second-price auction: Selling billions of dollars worth of keywords. American economic review, 97(1), 242-259.

Economides, N. (2006). Competition policy in network industries: an introduction. The new economy and beyond: Past, present and future, 5, 96.

European Union, European Commission, Consumers, Health and Food Executive Agency. (2013). Study on the coverage, functioning and consumer use of comparison tools and third- party verification schemes for such tools.

126 European Union, European Commission. (2017, December 18). CASE AT.39740 Google Search (Shopping). Retrieved from http://ec.europa.eu/competition/antitrust/cases/dec_docs/39740/39740_14996_3.pdf

Evans, B. J. (2011). Much ado about data ownership. Harv. JL & Tech., 25, 69.

Evans, D. S. (2009). The online advertising industry: Economics, evolution, and privacy. Journal of economic perspectives, 23(3), 37-60.

Evans, D. S., & Schmalensee, R. (2017). Network Effects: March to the Evidence, not to the Slogans.

Fee, P. R., Mialon, H. M., & Williams, M. A. (2004). What is a Barrier to Entry?. American Economic Review, 94(2), 461-465.

Filloon, W. (2018, November 16). Yelp's Heyday Is Over. Retrieved from https://www.eater.com/2018/11/16/18094979/yelp-stock-plunge-future-viability-competition- google-instagram-twitter

Finley, K. (2018, July 25). The Secret Internet War Over Bots. Retrieved from https://www.wired.com/story/scraper-bots-and-the-secret-internet-arms-race/

Fortner Enterprises, Inc. v. United States Steel Corp., 394 U.S. 495 (1969)

Fox, E. M. (2005). Is There Life in Aspen after Trinko-The Silent Revolution of Section 2 of the Sherman Act. Antitrust LJ, 73, 153.

Gal, M. S. (2004). Monopoly pricing as an antitrust offense in the US and the EC: Two systems of belief about monopoly?. The Antitrust Bulletin, 49(1-2), 343-384.

Gesenhues, A. (2013, November 17). Study: Many Searchers Choose Google Over Bing Even When Google's Name Is On Bing's Results. Retrieved from https://searchengineland.com/users- prefer-google-even-when-155682

Gilbert, R. J., & Katz, M. L. (2001). An economist's guide to US v. Microsoft. Journal of Economic perspectives, 15(2), 25-44.

Goldfarb, A., & Tucker, C. (2011). Online display advertising: Targeting and obtrusiveness. Marketing Science, 30(3), 389-404.

Google. “How Search Works.” YouTube, YouTube, 4 Mar. 2010, www..com/watch?v=BNHR6IQJGZs.

Google Privacy | Why data protection matters. (n.d.). Retrieved June 4, 2018, from https://privacy.google.com/how-ads-work.html

Hahn, R. W., & Singer, H. J. (2008). An Antitrust Analysis of Google's Proposed Acquisition of DoubleClick. AEI-Brookings Joint Center Related Publication, (07-24).

127 Hardy, Q. (2012, October 29). Microsoft Renews Relevance With Machine Learning Technology. Retrieved from https://www.nytimes.com/2012/10/30/technology/microsoft-renews-relevance- with-machine-learning-technology.html?ref=technologyand_r=0

Haselton, T. (2017, August 14). Google is paying Apple billions per year to remain on the iPhone, Bernstein says. Retrieved from https://www.cnbc.com/2017/08/14/google-paying-apple-3- billion-to-remain-default-search--bernstein.html

Haucap, J., & Heimeshoff, U. (2014). Google, Facebook, Amazon, eBay: Is the Internet driving competition or market monopolization?. International Economics and Economic Policy, 11(1-2), 49-61.

Hayes, B. (1987). Competition and two-part tariffs. Journal of business, 41-54.

Italian Republic v Commission of the European Economic Community. Case 13-63, (European Court of Justice, 1963)

Jefferson Parish Hospital District No. 2 v. Hyde., 466 U.S. 2 (1984)

Joe, R. (2015, January 30). Google CPC Declines As Company Misses Earnings Estimates. Retrieved from https://adexchanger.com/investment/google-cpcs-decline-as-company-misses- earnings-estimates/

Konkurrensverket v. TeliaSonera Sverige AB. Case C-52/09, (European Court of Justice, 2011)

Krattenmaker, T. G., & Salop, S. C. (1986). Anticompetitive exclusion: Raising rivals' costs to achieve power over price. Yale LJ, 96, 209.

Kucharczyk, J. (2018, April 24). Towards a 'Law of the Platform'? A Regulatory Perspective. Retrieved from https://www.project-disco.org/european-union/113015-towards-a-law-of-the- platform-a-regulatory-perspective/#.V4zfjiN95FQ

Laja, P. (2019, March 06). People Comparison Shop, Stupid. Retrieved from https://conversionxl.com/blog/people-comparison-shop-stupid/

Lambrecht, A., & Tucker, C. E. (2015). Can Big Data protect a firm from competition?. Available at SSRN 2705530.

Liedtke, M. (2018, October 08). Google Plus to close after bug leaks personal information. Retrieved from https://business.financialpost.com/pmn/business-pmn/google-plus-to-close-after- bug-leaks-personal-information

Long, D. (2007, February 1). Masterclass: The Revolution Masterclass on behavioural targeting. Retrieved May 31, 2018, from https://www.campaignlive.co.uk/article/masterclass-revolution- masterclass-behavioural-targeting/628568

Luchetta, G. (2013). Is the Google platform a two-sided market?. Journal of Competition Law and Economics, 10(1), 185-207.

128 Lycett, M. (2013). ‘Datafication’: Making sense of (big) data in a complex world.

Mangles, C. (2018, January 30). Search Engine Statistics 2018. Retrieved from https://www.smartinsights.com/search-engine-marketing/search-engine-statistics/

Manne, G., Morris, J., Stout, K., & Auer, D. (2019, January 7). Understanding Competition in Markets Involving Data or Personal or Commercial Information (FTC Hearings, ICLE Comment 7). Retrieved from https://laweconcenter.org/resource/icle-comments-competition-in-markets- involving-consumer-data/

Manne, G. A., & Sperry, B. (2015, March 26). Debunking the Myth of a Data Barrier to Entry for Online Services. Retrieved from https://truthonthemarket.com/2015/03/26/debunking-the-myth- of-a-data-barrier-to-entry-for-online-services/

Manne, G. A., & Wright, J. D. (2010). Innovation and the Limits of Antitrust. Journal of Competition Law and Economics, 6(1), 153-202.

Masnick, M. (2019, June 18). Dumbest 'Gotcha' Story Of The Week: Google, Genius And The Copying Of Licensed Lyrics. Retrieved from https://www.techdirt.com/articles/20190617/13335342414/dumbest-gotcha-story-week-google- genius-copying-licensed-lyrics.shtml

McAllister, N. (2015, April 27). REVEALED: The 19 firms whose complaints form EU's antitrust case against Google. Retrieved from https://www.theregister.co.uk/2015/04/27/eu_google_antitrust_complainants/

MEO v. Autoridade da Concorrência. Case C-525/16, (European Court of Justice, 2018)

Microsoft Corp v. European Commission. Case T-201/04, (European Court of Justice, 2004)

Molla, R. (2018, April 09). Amazon spent nearly $23 billion on R&D last year - more than any other U.S. company. Retrieved from https://www.vox.com/2018/4/9/17204004/amazon-research- development-rd

Molla, T. T. (2017, July 24). Google is still growing 21 percent - and its cloud segment is growing even faster. Retrieved March 11, 2018, from https://www.recode.net/2017/7/24/16020840/alphabet-google-business-stock-earnings-growth- cloud-revenue-profits-q2-2017

Moore, S. (2018, March 29). Amazon Commands Nearly Half of Consumers' First Product Search - BloomReach - Powering Enterprise CMS and Optimizing Commerce Experiences. Retrieved from https://www.bloomreach.com/en/blog/2015/10/amazon-commands-nearly-half-of- consumers-first-product-search.html [https://perma.cc/LVD9-F6W9]

Musgrove, J. B. (2009, Summer). Most Significant Changes to Competition Act in 20 Years Barely Noticed. Retrieved from https://mcmillan.ca/101473

129 Nadel, D. (2019, March 19). The 2019 Amazon Consumer Behavior Report. Retrieved from https://fv.feedvisor.com/rs/656-BMZ-780/images/Feedvisor-Consumer-Survey-2019.pdf

Pan, B., Hembrooke, H., Joachims, T., Lorigo, L., Gay, G., & Granka, L. (2007). In Google we trust: Users’ decisions on rank, position, and relevance. Journal of computer-mediated communication, 12(3), 801-823.

Perrin, N. (2018, September 19). Amazon Is Now the No. 3 Digital Ad Platform in the US. Retrieved from https://www.emarketer.com/content/amazon-is-now-the-no-3-digital-ad- platform-in-the-us

Petit, N. (2014). Price Squeezes with Positive Margins in EU Competition Law: Economic and Legal Anatomy of a Zombie. SSRN Electronic Journal. doi: 10.2139/ssrn.2506521

Petit, N. (2015). Theories of self-preferencing under Article 102 TFEU: A reply to Bo Vesterdorf. Available at SSRN 2592253.

Pope, D. (1983). The making of modern advertising. Basic Books.

Posner, R. (1987). The Robinson-Patman Act: Federal Regulation of Price Differences. Books.

Ratcliff, C. (2014, October 30). What is Google AdWords and how does it work? Retrieved June 4, 2018, from https://econsultancy.com/blog/65682-what-is-google-adwords-and-how-does-it-work

Ratliff, J. D., & Rubinfeld, D. L. (2010). Online advertising: Defining relevant markets. Journal of Competition Law and Economics, 6(3), 653-686.

Ratliff, J. D., & Rubinfeld, D. L. (2014). Is there a market for organic search engine results and can their manipulation give rise to antitrust liability?. Journal of Competition Law and Economics, 10(3), 517-541.

Raff, A., and Raff, S. (2018, January 21). The EC's Google Search Case A Timeline of Significant Events. Retrieved June 8, 2018, from http://www.foundem.co.uk/Foundem_Google_Timeline.pdf

Renda, A. (2015a). Antitrust, Regulation and the Neutrality Trap: A Plea for a Smart, Evidence- Based Internet Policy. CEPS Special Report, 104.

Renda, A. (2015b). Searching for harm or harming search? A look at the European Commission’s antitrust investigation against Google (Vol. 118). CEPS.

Robinson-Patman Act, (1936). 15 U.S. Code § 13

Rohle, T. (2007). Desperately seeking the consumer: Personalized search engines and the commercial exploitation of user data. First Monday, 12(9).

Rosenberg, E. (2019, March 12). How Google Makes Money (GOOG). Retrieved from https://www.investopedia.com/articles/investing/020515/business-google.asp

130 Rubinfeld, D. L., & Gal, M. S. (2017). Access barriers to big data. Ariz. L. Rev., 59, 339.

Salinger, M. A., & Levinson, R. J. (2015). Economics and the FTC’s Google Investigation. Review of Industrial Organization, 46(1), 25-57.

Segarra, L. (2018, September 29). Google to Pay Apple $12B to Remain Safari's Default Search: Report. Retrieved from http://fortune.com/2018/09/29/google-apple-safari-search-engine/

Simeonov, S. (2006, August 1). Metcalfe's Law: more misunderstood than wrong? Retrieved from https://blog.simeonov.com/2006/07/26/metcalfes-law-more-misunderstood-than-wrong/

Soper, T. (2015, October 07). Amazon's dominance of online shopping starts with product searches, study shows. Retrieved from https://www.geekwire.com/2015/amazon-dominates-the- online-shopping-world-survey-shows/

State of Video Conferencing: Owl Labs. (2018). Retrieved from https://www.owllabs.com/state-of- video-conferencing

Stigler, G. J. (1983). The organization of industry. University of Chicago Press Economics Books.

Stucke, M. E., & Grunes, A. P. (2016). Introduction: Big Data and Competition Policy. Big Data and Competition Policy, Oxford University Press (2016).

Sullivan, D. (2016, November 11). Google Launches "Universal Search" and Blended Results. Retrieved from https://searchengineland.com/google-20-google-universal-search-11232

Sullivan, D. (2016, November 11). What Is Google PageRank? A Guide For Searchers and Webmasters. Retrieved from https://searchengineland.com/what-is-google-pagerank-a-guide-for- searchers-webmasters-11068

Sullivan, D. (2017) Google Forecloses On Content Farms With "Panda" Algorithm Update. Retrieved from https://searchengineland.com/google-forecloses-on-content-farms-with-farmer- algorithm-update-66071

Teevan, J., Dumais, S. T., & Horvitz, E. (2010). Potential for personalization. ACM Transactions on Computer-Human Interaction (TOCHI), 17(1), 4.

United Brands Company and United Brands Continentaal BV v Commission of the European Communities (1978). Case 27/76, European Court of Justice

United States, Federal Trade Commission. (2013, January 3). Statement of the Federal Trade Commission Regarding Google’s Search Practices. Retrieved from https://www.ftc.gov/sites/default/files/documents/public_statements/statement-commission- regarding--search-practices/130103brillgooglesearchstmt.pdf

United States v. Microsoft Corp., 253 F.3d 34 (D.C. Cir. 2001)

131 U.S. v. Microsoft., Civil Action No. 98– 1232 (TPJ), Court’s Finding of Facts, U.S. District Court for the District of Columbia, November 5, 1999.

United States v. Colgate & Co., 250 U.S. 300 (1919)

Varian, H. (2018). Artificial intelligence, economics, and industrial organization (No. w24839). National Bureau of Economic Research.

Verizon v. Trinko, 540 U.S. 398 (2004)

Wagner, K., and Molla, R. (2018, February 12). Facebook lost around 2.8 million U.S. users under 25 last year. 2018 won't be much better. Retrieved from https://www.recode.net/2018/2/12/16998750/facebooks-teen-users-decline-instagram-snap- emarketer

Whinston, M. D. (1989). Tying, foreclosure, and exclusion (No. w2995). National Bureau of Economic Research.

Witt, A. C. (2019). The European Court of Justice and the More Economic Approach to EU Competition Law—Is the Tide Turning?. The Antitrust Bulletin, 64(2), 172-213.

132