Committee for the Study of Digital Platforms Market Structure and

George J. Stigler Center for the Study of the Economy and the State

The University of Chicago Booth School of Business

Committee for the Study of Digital Platforms

Market Structure and Antitrust Subcommittee

Report

DRAFT

15 May 2019

Subcommittee members

Fiona Scott Morton (chair), Theodore Nierenberg Professor of Economics, Yale School of

Management

Pascal Bouvier, Managing Partner, MiddleGame Ventures

Ariel Ezrachi, Slaughter and May Professor of Competition Law, The University of Oxford

Bruno Jullien, Senior Researcher, CNRS, Toulouse School of Economics

Roberta Katz, Senior Research Scholar, Center for Advanced Study in the Behavioral Sciences,

Stanford University

Gene Kimmelman, President and CEO of Public Knowledge

A. Douglas Melamed, Professor of the Practice of Law, Stanford Law School

Jamie Morgenstern, Assistant Professor of Computer Science, Georgia Institute of Technology

Acknowledgements

We thank Luigi Zingales and Guy Rolnik for inviting us to spend our time on this interesting and

complex project, and all the wonderful staff at the Stigler Center for the Study of the Economy

and the State. We particularly thank Rachel Piontek for her logistic and document skills, and

Filippo Lancieri for guiding the intellectual work along from the very start. Many thanks for

outstanding help with content go to students Doni Bloomfield, Rachel Cheong, and Steffi

Ostrowski of the Yale Law School. We are grateful to the reviewers at other institutions who

took time to give wise and helpful comments. We additionally want to thank our colleagues in

other nations whose reports came out before ours and provided stimulating ideas and

frameworks. We particularly build on the “Unlocking digital competition: Report of the Digital

Competition Expert Panel” chaired by Jason Furman for the Government of the United Kingdom

and the European Commission’s Special Advisors’ report “Competition Policy for the Digital

Era.”

DISCLAIMER:

The purpose of these preliminary reports is to identify what are the new challenges digital

platforms pose to the economic and political structure of our countries. These reports also try to

identify the set of possible tools that might address these challenges. Yet, there is potential

disagreement among the members of the committees on which of these problems is most

troubling, which tools might work best, whether some tools will work at all or even whether the

damage they might produce is larger than the problem they are trying to fix. Not all committee

members agree with all the findings or proposals contained in this report. The purpose of

these preliminary reports, thus, is not to unanimously provide a perfect list of policy fixes but

to identify conceptual problems and solutions and start an academic discussion from

which robust policy recommendations can eventually be drafted.

Table of Contents

Introduction ....................................................................................................................................4

Executive Summary .......................................................................................................................6

IMPORTANT FEATURES OF THE DIGITAL BUSINESS ENVIRONMENT .................11

The Structure of Digital Markets ...............................................................................................11

An Economic Perspective on Digital Market Structure .........................................................11

Barriers to Entry and Expansion ............................................................................................17

How Big Data Affects Digital Products and the Sale of Advertisement....................................21

What Makes Data Big? And Does Big Data Harm Consumers? ..........................................21

Data Control and Ownership ..................................................................................................28

Digital Platforms are Characterized by Free Services ...........................................................32

PROBLEMS ARISING IN DIGITAL MARKETS .................................................................34

Quality Harms ...........................................................................................................................35

Personal Data and Behavioral Economics .............................................................................35

The Harms from an Advertising-Supported Business Model ...............................................38

Online Exploitation and Addiction ........................................................................................41

The Assessment Problem ..........................................................................................................45

Measuring Consumer Welfare in a Behavioral World ..........................................................45

Assessing the Social Welfare of Advertising .........................................................................46

Harms to Investment and Innovation ........................................................................................46

Rents .......................................................................................................................................46

Harm to Entry, Including Discrimination .................................................................................49

The Practical Consequences of Entry Barriers .......................................................................49

Incumbent Incentive to Leverage Entry Barriers ...................................................................49

Disintermediation and Foreclosure of Potential Platform Entrants ........................................50

Foreclosure of Complements to Capture Rents .....................................................................51

Harm to Innovation ...................................................................................................................53

Competition Promotes Innovation .........................................................................................53

Entry Barriers and Innovation ...............................................................................................53

The Reward of Innovation .....................................................................................................55

Current Trends in Innovation ................................................................................................55

Old Wine in a New Bottle, or a New Reality? ..........................................................................57

SOLUTIONS ...............................................................................................................................59

Will Markets Self-Correct When Competition Problems Arise? ..............................................59

US Antitrust ...............................................................................................................................60

Basic Principles .....................................................................................................................60

Special Challenges Presented by Technology Platforms ..........................................................66

Could a Reformed Antitrust Law be Effective in Dealing with Digital Platforms? .................71

Reform by What Means? .......................................................................................................71

Reform of Antitrust Law Through General Tightening ........................................................73

Specific Areas of Possible Antitrust Reform .........................................................................74

A Competition Court .............................................................................................................78

A Regulatory Partner Could Enhance Effective Antitrust Enforcement ................................78

Regulation .................................................................................................................................79

The Digital Authority ............................................................................................................83

Menu of Regulations .............................................................................................................85

DA-Enforced Remedies for Antitrust Violations ..................................................................95

Aligning other Policies with Competition .............................................................................97

Adjudication Process .............................................................................................................97

CONCLUSION ...........................................................................................................................98

Boxes

Box I: Platforms vs. Brick-and-Mortar Advertising and Targeting ..........................................22

Box II: Why Data Has Increasing Marginal Returns ................................................................27

Box III: Digital Identities ..........................................................................................................31

Box IV: More About US Antitrust Law ....................................................................................61

Box V: EU Competition Law ....................................................................................................65

Box VI: The FCC Model ...........................................................................................................81

Box VII: Real Time Regulation in Financial Services ..............................................................86

I. Introduction

When the global, interactive, many-to-many communications network called the internet became

available for broad public use in 1995, people were overjoyed with the new power at their

fingertips. It is easy to forget that in the early years of the internet, that power was limited to

visiting shops and getting news from the relatively small number of enterprises that had built

websites on the World Wide Web. Over the past 25 years, that power has exploded with head-

spinning velocity: Today, there is no area of human life that has not been affected by the

technological innovations made possible by the internet. We now buy goods and services, do

banking, pay bills, find information, and talk with multiple groups of friends and acquaintances

on the web. The speed, scale, and scope of the internet, and of the ever-more powerful

technologies it has spawned, have been of unprecedented value to human society.

History teaches us that social institutions must adapt after major technological advances. In just

the past century, we saw this after the introduction of the automobile, the airplane, radio and

television, and cell phones, to name but a few examples. New occupations—for example, bus

drivers—arise to replace others—for example, buggy whip producers. At the same time, new

behavioral norms and expectations replace those associated with past technologies—consider the

social and economic changes wrought by the automobile, or the changes in both the rules and

norms of the workplace in the age of mobile devices. As unintended and unforeseen social

problems and harms arose, society responded with governance mechanisms aimed at addressing

the problems or harms without impeding the clear benefits associated with the advances. These

have ranged from the adoption of new social norms to the creation of new laws and regulations.

Examples abound, encompassing everything from crosswalks and traffic lights to legal remedies

addressing unfair competition in the marketplace. Ideally, the goal is to steer technological

advances to ensure widespread benefit without attendant widespread harms—to protect and

preserve innovation and advancement while minimizing harms so that all of society reaps net

benefits. Such a goal often involves government intervention and always involves tradeoffs as

society wrestles with the prevalence and cost of harms and how to balance them against the

prevalence and size of benefits. While often messy, this is a healthy and desirable debate.

This working group came together to address specific problems arising from the digital

platforms’ reach, scale, scope, and use of data. We were asked to examine concerns stemming

from the market structure contemporary platforms have created, and to investigate their

competitive behavior, including the consequences of network effects that can create barriers to

entry for new innovators and entrench incumbents. The global nature of many of today’s

platforms, a result of their scale, scope, and business models, creates novel complexities and

considerations, particularly a concern that the digital platform may be a unique combination of

economic forces that require both new analysis and new public policy. Regulatory authorities

throughout the world are now turning their attention to these same questions. This report

contributes to this international analytical project by providing some of the necessary

frameworks and inputs. We intend it to be a complement to other recent work, as experts across

the world wrestle with how to ensure that markets remain open and healthy, allowing beneficial

technological and social advancements to continue. Many of our conclusions and suggestions

echo the findings of reports that have come out in the past year, and we hope they will be helpful

to those reports not yet released. The list of antitrust experts and agencies working on this

problem includes Australia, the United Kingdom, Germany, the European Commission, France,

Israel, and Japan.

The issues are global in their scope, and these various jurisdictions are all engaged in analyzing

how best to ensure that societies in general, and competition in particular, continue to thrive in

the Digital Age. Our charge was restricted to market structure and competition, while other

committees considered the equally important topics of the impact of digital platforms on politics,

the media, and the nature of privacy. We note that monopolies can concentrate political power,

reduce media plurality, and provide insufficient competition on dimensions such as privacy. In

this way the findings of this report and the others are linked and quite consistent. Digital markets

and platforms have already delivered great benefits to consumers, and the global concerns that

have surfaced relating to actual or potential consumer harms may require action to ensure that

the benefits are not undermined. Our report concludes that with deliberate government action to

protect competition and consumers, the benefits from innovative firms could be even greater and

more equitably spread, ensuring that the public is not short-changed in firms’ pursuit of profit.

Accordingly, this report is offered in the spirit of ensuring a future of continued technological

and economic progress and social well-being as we move further forward into the Digital Age.

See JACQUES CREMER, YVES-ALEXANDRE DE MONTJOYE, & HEIKE SCHWEITZER, EURO. COMM’N, DIRECTORATE-

ENERAL FOR COMPETITION, COMPETITION POLICY FOR THE DIGITAL ERA (Apr. 4, 2019),

http://ec.europa.eu/competition/publications/reports/kd0419345enn.pdf [hereinafter EC

COMPETITION REPORT];

ASON FURMAN ET AL., H.M. TREASURY (U.K.), UNLOCKING DIGITAL COMPETITION: REPORT OF THE DIGITAL

COMPETITION EXPERT PANEL (March 13, 2019),

https://assets.publishing.service.gov.uk/government/uploads/system/uploads/attachment_data/file/785547/unlocking

_digital_competition_furman_review_web.pdf [hereinafter U.K.

COMPETITION REPORT]; AUSTL. COMPETITION &

CONSUMER COMM’N, DIGITAL PLATFORMS INQUIRY: PRELIMINARY REPORT (Dec. 2018),

https://www.accc.gov.au/system/files/ACCC%20Digital%20Platforms%20Inquiry%20-

%20Preliminary%20Report.pdf [hereinafter A

USTRALIA COMPETITION REPORT]; Japanese Ministry of Economy,

Trade, & Industry, Fundamental Principles for Rule Making to Address the Rise of Platform Businesses Formulated,

(Dec. 18, 2018), https://www.meti.go.jp/english/press/2018/1218_002.html; H

EIKE SCHWEITZER ET AL., GERMAN

BUNDESMINISTERIUM WIRSCHAFT UND ENERGIE, MODERNISING THE LAW ON ABUSE OF MARKET POWER: REPORT

FOR THE

FEDERAL MINISTRY FOR ECONOMIC AFFAIRS AND ENERGY (Apr. 9, 2018),

https://www.bmwi.de/Redaktion/DE/Downloads/Studien/modernisierung-der-missbrauchsaufsicht-fuer-

marktmaechtige-unternehmen-zusammenfassung-englisch.pdf [hereinafter G

ERMAN COMPETITION REPORT

SUMMARY]; L’AUTORITÉ DE LA CONCURRENCE [FRENCH COMPETITION AUTHORITY], PORTANT SUR

’EXPLOITATION DES DONNÉES DANS LE SECTEUR DE LA PUBLICITÉ SUR INTERNET [ON THE EXPLOITATION OF DATA

IN THE

INTERNET ADVERTISING SECTOR] (Mar. 6, 2018), http://www.autoritedelaconcurrence.fr/pdf/avis/18a03.pdf

[hereinafter F

RENCH COMPETITION REPORT]; Press Release, The Israel Antitrust Authority invites interested

members of the public to submit their comments, towards an inquiry into competition issues in the digital economy,

Reishut HaTachrut [Israel Antitrust Authority] (Apr. 9, 2018),

http://www.antitrust.gov.il/eng/subject/177/item/35246.aspx.

Executive Summary

a. Attributes

The market structure and antitrust report begins by discussing the characteristics of digital

markets. These markets often have extremely strong economies of scale and scope due to low

marginal costs and the returns to data. Moreover, they often are two-sided and have strong

network externalities and are therefore prone to tipping. If so, the competitive process shifts from

competition in the market to competition for the market. This combination of features means

many digital markets feature large barriers to entry. The winner in these settings often has a large

cost advantage from its scale of operations and a large benefit advantage from the scale of its

data. An entrant cannot generally overcome these without either a similar installed base (network

effects) or a similar scale (scale economies), both of which are difficult to obtain quickly and

cost-effectively.

Additional barriers to entry are, ironically, generated by the very consumers who are harmed by

them. Consumers do not scroll down to see more search results, they agree to settings chosen by

the service, they single-home on one platform, and they generally take actions that favor the

status quo and make it difficult for an entrant to attract consumers. In general, the findings from

the behavioral economics literature demonstrate an under-recognized market power held by

incumbent digital platforms.

The theme that runs throughout the report is the difficulty of entry into digital platform

businesses once an incumbent is established. Whether the entrant is vertical or horizontal, has

succeeded to some degree, is nascent, is a potential entrant, or is a large platform in an adjacent

space, its existence improves consumer welfare. Either the entrant provides more choice,

different features, and a chance of higher quality, or the threat of those outcomes spurs the

incumbent to provide lower prices, higher quality and innovation, and to do so more quickly.

The role of data in digital sectors is critical. Personal data of all types allows for targeted

advertising to consumers, a common revenue model for platforms. The report shows that the

returns to more dimensions and types of data may be increasing, which again advantages

incumbents. Consumer data in the United States is not regulated in any way that gives useful

control or privacy to consumers; and additionally, most consumers have little idea what is being

collected about them and re-sold. One way in which digital platforms often exploit their market

power – and increase their profits – is by requiring consumers to agree to terms and conditions

that are unclear, difficult to understand, and constantly changing, but which give the platform

freedom to monetize consumers’ personal data.

Digital platforms are characterized by free services. “Free” is not a special zone where

economics or antitrust do not apply. Rather, a free good is one where the seller has chosen to set

a monetary price of zero and may set other, non-monetary, conditions or duties. It is possible that

a digital market has an equilibrium price that is negative; in other words, because of the value of

target advertising, the consumer’s data is so valuable that the platform would pay for it. But the

difficulty of making micropayments might lead a platform to mark up this negative competitive

price to zero. As a result, barter is a common way in which consumers pay for digital services.

They barter their privacy and information about what restaurants they would like to eat in and

what goods they would like to buy in exchange for digital services. However, in principle, that

information has a market price that can be analyzed.

b. Harms

Market power, consumer biases and an ad-supported platform model can generate significant

consumer harms. First, market power in advertising markets will result in markups paid by

advertisers. Secondly, while behavioral economists have studied consumer biases and firm

responses in offline markets, these are swamped by what digital businesses can learn by using

high-dimensional, large datasets to explore every nook and cranny of consumers’ many

behavioral shortcomings and biases in real time. Framing, nudges, and defaults can direct a

consumer to the choice that is most profitable for the platform. A platform can analyze a user’s

data in real time to determine when she is in an emotional “hot state” and then offer targeted

sales. These tactics reduce the quality of the zero-price content the user experiences on the

platform.

In addition to de novo entry, platforms fear disintermediation by a partner or complement. If a

platform’s partner is able to directly access and serve the platform’s customers, it might take

them off the platform entirely, reducing the platform’s profit. A platform that has total control of

demand due to control over framing of consumer choices, policies for complements, and

technical standards can steer customers to content and complements of most benefit to it. The

most privately beneficial content might be owned by the platform itself rather than provided by

independent firms that could extract rent or even challenge the platform’s market power in the

future. To the extent that consumers single-home, they may not be aware of such steering, or

may not have competitive alternatives to which they can turn if they are aware.

Today’s platforms understand that in some settings they can obtain higher margins if they either,

make all of the necessary complements themselves, or, position themselves as a mandatory

bottleneck between partners and customers. In particular, digital platforms are often very careful

to maintain complete control over the user relationship so that they do not face any threat of

disintermediation from a complement. These technological and policy choices can be used to

reduce the possibility of successful entry by direct competitor. Other strategies such as exclusive

contracts, bundling, or technical incompatibilities can also be used by platforms to restrict entry

of competitors. Some of these strategies could be violations of existing antitrust law, as discussed

below.

Insufficient competition and entry result in harms to investment and innovation. There is

significant theoretical and empirical research that concludes that anticompetitive creation or

maintenance of market power will cause a reduction in the pace of innovation.

The lessening or

blocking of innovative entry is of particular concern given its value to consumers. A VC will

usually be wary of outright investing in an innovative startup that will implicitly or explicitly

compete head-on with a tech giant. In that case, the best hope might be to be the preferred

innovator of a complement and sell its business to the platform at an early stage. However, if

entry barriers were reduced, the entrepreneur would not have to settle for a small fraction of the

platform’s profits, but could compete for all of them and try to replace the platform. This

possibility would generate a much larger incentive to innovate.

c. Solutions

While some markets may self-correct, the findings of this report suggest that rapid self-

correction in markets dominated by large digital platforms is unlikely.

While US antitrust law has long been flexible in combatting anticompetitive conduct, there is

increasing concern that it has been underenforced in recent years. Antitrust law and its

application by the courts over the past several decades have reflected the now outdated learning

of an earlier era of economic thought, and they appear in some respects inhospitable to new

learning. Antitrust enforcement better suited to the challenges of the Digital Age may therefore

require new legislation.

Technology platforms present particular challenges for antitrust enforcement. Markets tip and the

resulting market power is durable, so even effective antitrust enforcement is unlikely to generate

fragmented markets. Nonetheless, enforcement that protects competition on the merits in the first

stage and prevents exclusionary conduct in the second stage will help ensure that market-

participants make unfettered choices among competing platforms and that entry and innovation

are not inhibited by private rent-seeking.

Economists and lawyers will have to develop tools to explain to courts the role of behavioral

biases in the creation of market power and in their effect on the quality of content. The existence

of zero money prices means that measurement of quality will be critical. The law needs better

analytical tools to take into account the impact of potential and nascent competitors and

competition. Market definition will vary according to what consumers are substituting between,

whether there is competition on the platform between complements, or competition between

platforms, or competition between a platform and potential or nascent competitors regarding

possible future markets. The need to identify the specific anticompetitive exclusionary conduct

and analyze it may raise enforcement costs given all the possible variants of exclusionary

conduct possible in digital markets.

This report details the particular areas where antitrust law will need reform in order to adequately

deliver competition to consumers. The report also suggests the establishment of a specialist

competition court to hear all private and public antitrust cases which would allow judges to

develop some expertise. The committee believes that vigorously enforcing the antitrust laws

under these conditions would be likely to increase entry in digital platform industries,

competition, and consumer welfare. Moreover, such enforcement would likely result in remedies

to restore competition that has already been lost, as well as serve as a deterrent to future

anticompetitive conduct.

However, because technology platforms present the enforcement challenges detailed above, even

effective enforcement may not be enough to generate competitive digital markets in a timely

fashion. Therefore, the report suggests that Congress should consider creating a specialist

regulator, the Digital Authority. The regulator could be tasked with creating general conditions

conducive to competition. The committee also suggests separating out some types of regulation

that will apply to virtually all market participants while other regulation will apply only to

companies with bottleneck power. “Bottleneck power” describes a situation where consumers

primarily single-home and rely upon a single service provider, which makes obtaining access to

those consumers for the relevant activity by other service providers prohibitively costly.

The Digital Authority could routinely collect data on digital transactions and interactions, with

an emphasis on data from businesses with bottleneck power. These data – made public to the

extent possible – would allow policy makers and researchers to assess the performance of the

sector. The DA could have a mandate to create “light touch” behavioral nudges when they will

make markets more competitive. An example of a regulation that would enhance competition is

data portability. The DA could set up rules that allow users to easily port their data from one

service provider to another and monitor compliance. The DA may also promote open standards

in such areas as micro-payments and digital identities. Should Congress request it, the DA could

oversee a mandate for interoperability in any market where market power has become

entrenched and threatens long term harm to competition. The Report also suggests that the DA

could carry out a parallel merger review that would be set up to incorporate necessary antitrust

reforms and modern standards.

Some regulations could apply only to firms that meet the DA’s definition for bottleneck power.

Because the cost of false negatives is high and there is uncertainty, the public interest requires

the DA to take a more interventionist approach in these settings. The DA could have merger

review authority over even the smallest transactions involving digital businesses with bottleneck

power because nascent competition against these entities is very valuable for consumers. Non-

discrimination rules could protect against a complement that is a potential competitor of the

platform itself, or one that operates only on the platform as a rival provider of content.

When a company has been found liable for violating the antitrust laws, part of the current

process is that antitrust authority devises a remedy to restore the lost competition. Data sharing,

full protocol interoperability, non-discrimination requirements, and the unbundling of content

from a platform are all tools that the regulator, in conjunction with the antitrust authority, could

apply and monitor over time in order to restore competitive markets.

II. IMPORTANT FEATURES OF THE DIGITAL BUSINESS

ENVIRONMENT

Digital technologies are a central driver of future prosperity, hopefully delivering waves of

innovation, efficiencies, and consumer welfare. These technologies have revolutionized the way

consumers shop and communicate, the way businesses deliver value, the way people work

together in collaboration, and—the subject of this report—the overall dynamics of competition.

These transformative changes, while immensely beneficial, have also triggered growing concerns

about the power of a small number of firms to control and influence billions of lives. As an

increasing volume and range of commercial activities have been digitalized, society has

witnessed the emergence of certain key platforms and gatekeepers and a shift in market

dynamics.

This section outlines the key features of the digital environment to set the foundation for the

discussion of problems and solutions.

a. The Structure of Digital Markets

i. An Economic Perspective on the Digital Market Structure

Despite the predictions of some early observers that the internet would create competitive—even

perfect—markets,

certain digital firms have been able to acquire significant market positions

and preserve them over time. Many of the most innovative internet-derived digital markets, such

as search engines, social networks, network operating systems, ecommerce, and ride-sharing, are

highly concentrated and have been dominated by one or a few firms for a number of years. The

lack of entry of competitors in these important markets—despite high profits—suggests either

barriers to entry or exclusionary conduct, or both. We first discuss the nature of entry barriers in

digital platforms and why they are difficult for an entrant to surmount.

1) A Unique Setting Subject to Tipping

From an economic perspective, there is no single new characteristic that would make

competition in digital platforms different from more traditional markets. Rather, it is the

coincidence of several factors at a scale that has not been encountered before that makes the

problem unique and requires new analysis of market structure and market power. In particular,

See, e.g., Erik Brynjolfsson & Michael D. Smith, Frictionless Commerce? A Comparison of Internet and

Conventional Retailers, 46 M

GMT. SCI. 563, 563 (2000) (“The conventional wisdom regarding Internet competition .

. . is that the unique characteristics of the Internet will bring about a nearly perfect market.”); Jeffrey R. Brown &

Austan Goolsbee, Does the Internet Make Markets More Competitive? Evidence from the Life Insurance Industry,

110 J.

POL. ECON. 481, 482 (2002) (“The traditional economic view suggests that . . . the Internet should reduce

search costs for consumers and thereby reduce prices and make markets more competitive.”).

the platforms with which this report is most concerned demonstrate extremely strong network

effects, very strong economies of scale, remarkable economies of scope due to the role of data,

marginal costs close to zero, drastically lower distribution costs than brick and mortar firms, and

a global reach.

Markets with these combined features are prone to tipping—a cycle leading to a dominant firm

and high concentration. Digital markets are prone to tipping for two primary reasons. First,

because fixed costs play such an important role in digital markets, these markets feature

especially large returns to scale. Second, many digital markets are driven by network effects that

strengthen large incumbents and weaken new entrants.

When markets are prone to tipping, the competitive process shifts from competition in the

market to competition for the market. In that case, consumers may only benefit from competition

among several firms for the relatively short time period in which the firms compete to be the

ultimate winner of very large economic profits. The winner’s monopoly profits serve as the

inducement for entry and investment. After a market has tipped, a potential rival for the market

can only overcome the incumbency advantage of established networks through significant

innovation. However, even an innovative entrant may not be able to create competition that

benefits consumers in the presence of the factors that led to the tipping, leaving open a role for

public policy to allow for competition for the market. Moreover, a competitor is even less likely

to enter the market if the incumbent platform is able to leverage its powerful position to

disadvantage or exclude potential entrants.

There are many well-known problems that follow from lack of competition, including higher

prices, less innovation, and lower quality in all its forms. Policy may be needed to address the

cause of such symptomatic problems. In the view of this committee, protecting entry for existing

and potential competitors is the most important way to protect or improve consumer welfare in

digital platforms. Initial competition for a market should be conducted on the merits without any

anticompetitive practices, and later entrants should face a level playing field and no exclusionary

conduct as they contest the market. Regulation may be required to prevent incumbents from

erecting improper barriers to entry.

As it is traditionally recognized, the reward for the winner’s innovative activity is the ability to

extract rent from the platform through the exercise of market power.

However, we highlight

three deviations from this principle. First, the winner must have “won” on the merits of its

product, without the use of any anticompetitive conduct. Correcting illegal practices by dominant

firms that have won in a tipped digital market can be difficult, but is necessary or the firm will

earn profits from its illegal behavior—and harm consumers along the way. Second, the reward of

the entire market that propelled the winner to innovate must be available to the next entrant. If

See United States v. Alcoa, 148 F.2d 416, 430 (2d Cir. 1945) (“The successful competitor, having been urged to

compete, must not be turned upon when he wins.”) (Hand, J.).

the incumbent is able to withhold those rents by excluding the entrant, or reduce those rents by

limiting the entrant’s share, then the pace of innovation will slow. The economic literature

demonstrates that vibrant innovation and entry is the most important source of consumer welfare

over time—hence the focus of the report on ensuring that entry functions well both when a

market is created and thereafter.

Third, there are sources of platform rents that society may

determine through appropriate regulation should not be part of the winner’s reward. For

example, investments in complements offered on the platform may be best incentivized and

encouraged to compete if the platform owner is not entitled to exclude them. (As used in this

report, a complement is a good or service offered on or through the platform that increases the

value of the platform to a consumer.) Digital markets at issue today may not have featured free

entry and lack of anticompetitive conduct in the past. This report covers that case by describing

how antitrust can be used to look backward, as well as forward, for antitrust violations. We also

suggest regulation that can look forward to protect entrants, and thereby consumers, in these

other cases.

2) Increasing Returns to Scale

Digital markets are used to exchange information goods and services. Typically, information

goods involve increasing returns to scale because their production requires a fixed cost and no or

little variable cost.

In other words, when an additional user is served, costs do not go up

proportionately. For example, an eBook, once produced, can be distributed at almost no cost to

all users with access to the internet. The same holds for information services that are subject to

fixed design and development costs and fixed maintenance and updating costs; Google can

update Google Calendar for 100 million users with similar fixed expenses as would be needed

for only a fraction of such users.

In contrast to traditional media or cultural markets that have had to incur physical distribution

costs, digital markets are largely able to avoid such costs. The fact that information services can

be delivered to any geographic location with no or minimal cost is one reason for the abundance

of such services. The capacity to reach a large scale at small cost changes the nature of business

growth. While a traditional business often starts with local implementation followed by gradual

expansion through investment as reputation and financial resources increase, many online

businesses aim at rapid large-scale expansion. This rapid growth may reduce the length of the

competition-for-the-market phase, as market winners can establish dominance and begin

exercising their market power quickly. It took only five years for Facebook, the “move fast and

See, e.g., Joseph Farrell & Paul Klemperer, Coordination and Lock-In: Competition with Switching Costs and

Network Effects, in 3 H

ANDBOOK OF INDUSTRIAL ORGANIZATION (Mark Armstrong & Robert Porter eds., 2007),

https://www.sciencedirect.com/science/article/pii/S1573448X06030317; Wen Wen & Feng Zhu, Threat of Platform-

Owner Entry and Complementor Responses: Evidence from the Mobile App Market, (Harvard Business School

Working Paper 18-036, 2018), https://papers.ssrn.com/sol3/papers.cfm?abstract_id=284853.

Increasing returns to scale occur when the average unit cost decreases when sales expand.

break things” company, to go from a million users in 2004, the year of its founding, to more than

350 million users in 2009, when it overtook MySpace for good.

The implications for market structure are well known: increasing returns to scale lead companies

to invest in fixed costs in order to have the best product to attract customers.

Then, with a larger

customer base, the firm can enjoy lower average costs per consumer, allowing it to make an offer

to consumers that is attractive in both quality and price. The increasing returns to scale create

barriers to entry: New firms cannot offer the quality of the incumbent without the same large-

scale operation to pay for the fixed costs. But the firm can only achieve a large scale if quality is

high. Thus, a potential entrant, foreseeing that it will not be profitable at the smaller scale, will

not enter the market to challenge the incumbent.

3) Economies of Scope

Platforms also have powerful economies of scope in the form of their relationships with users

and brands. In digital markets specifically, scale offers an additional advantage. Firms can apply

machine learning to extensive data sets to improve their products and expand their activities into

new areas. Because machine learning yields better insights when it is trained on larger datasets,

firms with access to large amounts of data can raise the quality of their services in ways that

smaller firms cannot. This creates a form of dynamic economies of scale, allowing large firms

with large amounts of data to raise product quality at lower costs than small firms. Firms may

also be able to leverage the data, or the insights due to machine learning, that they receive from

an existing service to enter into an adjacent market with a higher quality product, demonstrating

a novel form of economies of scope. Combining mapping software in a platform that already

offers email, for example, allows that platform to offer a higher quality restaurant

recommendation product. Moreover, firms serving a larger customer base with a greater variety

of products are able to generate more advertising revenue per consumer through more effective

targeting.

The development of machine learning technologies and data analysis is a source of

increasing returns to scale and scope that can contribute to digital market concentration.

See Chloe Albanesius, More Americans Go to Facebook Than MySpace, PC MAG. (June 16, 2009),

https://www.pcmag.com/news/241432/more-americans-go-to-facebook-than-myspace; Ami Sedghi, Facebook: 10

Years of Social Netoworking, in Numbers, G

UARDIAN (Feb. 4, 2014),

https://www.theguardian.com/news/datablog/2014/feb/04/facebook-in-numbers-statistics. See also Steven Levy,

Mark Zuckerberg on Facebook's Future, From Virtual Reality to Anonymity, W

IRED (Apr. 30, 2014),

https://www.wired.com/2014/04/zuckerberg-f8-interview (reporting Mark Zuckerberg noting that “[w]e’ve changed

our internal motto from ‘Move fast and break things’ to ‘Move fast with stable infrastructure’”).

See generally JOHN SUTTON, SUNK COSTS AND MARKET STRUCTURE (1991).

Such economies of scope are leading to a rapid expansion of the capture and use of personal information by firms.

For example, many large firms are starting to expand beyond internet platforms in order to collect more data from

consumers. The rise of voice-assistant products (including Microsoft’s Cortana, Apple’s Siri, Google’s Google

Voice, and Amazon’s Alexa) can enable firms to gather information from offline consumer behavior and then feed

that information into online advertisement-targeting algorithms. One study estimates that major internet platforms,

data brokers, credit card companies, and healthcare data companies derived nearly $76 billion in 2018 from selling

personal consumer information directly or indirectly via targeted ads. Robert Shapiro & Siddhartha Aneja, Who

4) Network Effects

Most information services involve some form of positive network effects, in which consumer

benefit grows as the number of users increases. The simplest network effects flow directly from

interacting with other users, as in social networks or peer-to-peer services, such as eBay or

Venmo, where the more users there are on the network, the richer the users’ experience is likely

to be. (In these examples the platforms choose not to be interoperable, so the network effects

apply to a single firm, rather than an industry.) Multi-homing lessens network effects because a

consumer can enjoy the size of both networks, rather than having to choose one. For example, a

consumer that carries both Visa and American Express credit cards can shop at stores that accept

either card, or both. Many other network effects are indirect, in that they are mediated by a

“complement” to the network. A complement is a good or service that increases the value of

another good or service to a consumer. For example, the ability to not only make calls, but also

to play music on a handset increases the value of the handset to users. Complements today often

come in the form of applications (“apps”) or a specific type of content. The more complements a

platform has, the more popular it is with users. Just as customers of a popular app store receive

an indirect network effect when more and better developers are attracted to app stores with big

customer bases, customers of a widely-used social media site benefit from the many games

designed for that social media site, which in turn are driven by the large number of consumers.

These network effects can also be seen in recommendation systems or driving directions that

exploit larger datasets of users’ purchasing behavior or travel paths to offer higher quality

advice.

Markets with network effects are prone to concentration because consumers benefit from being

on the same network as other users. No one wants to be on their own social media site. However,

when network effects are exhausted relatively quickly (as can occur, for example, in messaging

apps

) or when there is heterogeneity in preferences (e.g. teenagers prefer not to be on the same

social network as their parents

), the market structure may be oligopolistic.

Owns Americans’ Personal Information and What Is It Worth?, FUTURE MAJORITY (Mar. 8, 2019),

https://assets.futuremajority.org/uploads/report-for-future-majority-on-the-value-of-people-s-personal-data-shapiro-

aneja-march-8-2019.pdf.

To send a message to someone, a user only needs that single person to be on a particular messaging app, rather

than everyone they would want to send a message to. This explains why many messaging apps can live alongside

one another—WhatsApp, Snapchat, SMS, and Facebook Messenger all have significant customer bases. See Most

Popular Mobile Messaging Apps Worldwide as of January 2019, S

TATISTA (Jan. 2019),

https://www.statista.com/statistics/258749/most-popular-global-mobile-messenger-apps.

See Mark Sweney, Is Facebook for Old People? Over-55s Flock in as the Young Leave, GUARDIAN (Feb. 12,

2018), https://www.theguardian.com/technology/2018/feb/12/is-facebook-for-old-people-over-55s-flock-in-as-the-

young-leave.

Some indirect network effects are multi-sided: A user of one type (e.g., a buyer) benefits from

the participation of other types of users (e.g., sellers). For instance, in ecommerce platforms,

which intermediate trade between sellers and buyers, a buyer does not directly benefit from the

presence of other buyers but does benefit from the presence of more sellers—who are in turn

attracted by the presence of the buyers. Multi-sided network externalities are prominent on the

internet for two reasons. First, business models like ecommerce are plentiful. Second, and more

importantly, a great deal of activity on the internet is financed by paid advertising on the relevant

site. Advertising-financed services are platforms exploiting two-sided network effects between

advertisers and consumers. Advertising can be more valuable when there are more consumers

viewing a site, while consumers are attracted by content that may be of higher quality when

financed by many advertisers competing for space.

On multi-sided platforms, one or more sides may be subsidized if their participation attracts

paying customers on the other side. Thus, we may see a monetary price of zero in one part of a

very lucrative business model. Shoppers on eBay can buy without paying fees because eBay

understands that fees would drive them away, whereas their presence draws in sellers who can be

charged fees. Users of Gmail pay a monetary price of zero but allow Google to read their email

so that advertisers can market to them based on personal information. Google can then charge a

high price for the ads. Payment platforms such as credit cards or PayPal similarly charge end

consumers low or negative fees because a large group of consumers bring in retailers who pay

the payment platform for access.

Network effects can lead to consumer-friendly competition at early stages. Economic theory and

market observation indicate that during the phase when competitors are all trying to tip the

market towards themselves, they compete intensely. This phase is characterized by vigorous

competition between firms trying to build market shares and generate bandwagon effects. For

example, Uber and Lyft have hotly contested the market for ride-sharing—and spent billions of

dollars subsidizing riders’ fares along the way.

One 2016 estimate suggested that payments

from Uber customers covered only about 40% of the cost of their rides.

If network effects are

See Shannon Bond & Nicole Bullock, Uber IPO Prospectus Shows Ride-Hailing Revenues Stalled, FIN. TIMES

(Apr. 11, 2019), https://www.ft.com/content/c68d3662-5c76-11e9-939a-341f5ada9d40 (Uber “has lost $12bn from

its operations in total since 2014”); id. (“‘We will not shy away from making short-term financial sacrifices where

we see clear long-term benefits,’ wrote Dara Khosrowshahi, [Uber’s] chief executive.”); Megan McArdle, Uber and

Lyft Are Losing Money. At Some Point, We’ll Pay for It., W

ASH. POST (Mar. 5, 2019),

https://www.washingtonpost.com/opinions/uber-and-lyft-are-losing-money-at-some-point-well-pay-for-

it/2019/03/05/addd607c-3f95-11e9-a0d3-1210e58a94cf_story.html (In 2018, Lyft “lost more than $900 million after

expenses”).

See Izabella Kaminska, The Taxi Unicorn’s New Clothes, FIN. TIMES (Dec. 1, 2016),

https://ftalphaville.ft.com/2016/12/01/2180647/the-taxi-unicorns-new-clothes. Many of the paradigmatic firms of

the dot.com bubble also followed this strategy of subsidizing for growth—often sending the firm to oblivion. See

Greg Bensinger, In Kozmo.com’s Failure, Lessons for Same-Day Delivery, W

ALL ST. J. (Dec. 2, 2012),

https://blogs.wsj.com/digits/2012/12/03/in-kozmo-coms-failure-lessons-for-same-day-delivery.

strong, however, the market will tip in favor of one competitor, who then becomes the

monopolist.

5) Low Marginal and Distribution Costs

Digital goods often have low to zero marginal costs of expansion to another user, as mentioned

above. Distribution costs, one of the major expenses of expanding in a brick and mortar world,

are largely also zero on the internet. Indeed, poor internet infrastructure in the United States and

elsewhere may be the main distribution cost for many digital platforms.

Some digital platforms

do have positive distribution or marginal costs such as a piece of hardware, ecommerce

warehouses, or maintenance of scooters for example.

6) Global Reach

Lastly, many of the digital platforms we discuss in this report have a global reach. Those that are

focused on licensed digital content, rather than user-generated content, must negotiate licensing

and other legal issues in new jurisdictions. Reaching consumers in many languages is another

cost of expanding globally, as is dealing with local transportation regulations. However, the total

cost of expansion is generally lower than in traditional brick and mortar businesses.

ii. Barriers to Entry and Expansion

In this section, we explore the implications of the characteristics discussed above and the

possible effects of the increased market power created by barriers to entry. Outlined below are

some of the drivers that make market power more sustainable and disruption less likely in digital

markets.

1) Cost and Benefit Barriers to Entry

As mentioned above, network effects and scale economies are two significant barriers to entry.

The winner in these settings has a massive cost advantage from its scale of operations, and a

massive benefit advantage from the scale of its data. An entrant cannot generally overcome these

without either a similar installed base (network effects) or a similar scale (scale economies), both

of which are difficult to obtain quickly and cost-effectively. It is possible for an entrant to arrive

with scale and an installed base because it is a competing digital platform. Such entrants may

create effective competition for an incumbent platform that benefits consumers.

See, e.g., Drew FitzGerald, Facebook Looks to Build Underwater Ring Around Africa, WALL ST. J. (Apr. 7,

2019); Michael Sheetz, Here’s Why Amazon Is Trying to Reach Every Inch of the World With Satellites Providing

Internet, CNBC (Apr. 7, 2019), https://www.cnbc.com/2019/04/05/jeff-bezos-amazon-internet-satellites-4-billion-

new-customers.html; Press Release, Think Big With a Gig: Our Experimental Fiber Network, G

OOGLE (Feb. 10,

2010), https://googleblog.blogspot.com/2011/03/ultra-high-speed-broadband-is-coming-to.html.

Barriers to equivalent data resources, a side effect of not having the history, scale, or scope of the

incumbent, can inhibit entry, expansion, and innovation.

The same effects that drive the quality

of digital services higher as more users join—a positive feedback loop—makes the strong

stronger and the weak weaker.

Data feeds the development of algorithmic and AI training

processes that enables more profitable exploitation of consumer attention through advertising. A

data advantage over rivals can enable a company to achieve a virtuous circle of critical

economies of scale leading to network effects, and a competitive balance in its favor, leading to

the gathering of yet more data. A new entrant is likely to experience this in reverse—a vicious

cycle—as it fails to surmount the entrance barrier.

Choosing a business model that is scalable, and has strong economies of scale, is of paramount

importance to creating entry barriers. A social media platform that chose to monitor harmful

content might need to hire many workers to keep ahead of users that game the algorithms. Hiring

workers is costly, slows down the ability to grow, and makes the platform less profitable both in

the short run and in the long run by limiting economies of scale. A platform aiming to tip the

market in its favor quickly will not choose this business model if it can gain consumers at lower

cost.

Because cost and benefit barriers are so powerful, digital platforms have powerful incentives to

pull slightly ahead of any rival. Due to the positive feedback loop described above, a small

advantage can turn into a large one. This powerful incentive to disadvantage a competitor raises

the possibility that some competitors will choose to violate the antitrust laws in order to stay

ahead. In general, to maintain or improve their competitive advantage, incumbents have strong

incentives to limit openness or interoperability and to be averse to data-portability policies. For

example, in a recent case, the upstart ethernet switch firm Arista alleged that the incumbent

Cisco encouraged openness and interoperability while it was dominant, but once a rival mounted

a serious challenge, claimed copyright protection on the previously open user interface, thus

cutting off the entrant from locked-in users.

The key themes below are discussed in greater detail in ARIEL EZRACHI & MAURICE E. STUCKE, VIRTUAL

COMPETITION: THE PROMISE AND PERILS OF THE ALGORITHM DRIVEN ECONOMY (2016). See also Ariel Ezrachi &

Maurice E. Stucke, Digitalisation and Its Impact on Innovation: Report Prepared for the European Commission

(DG Research & Innovation Working Paper, July 2018) (on file with authors).

See Data-Driven Innovation for Growth and Well-Being: Interim Synthesis Report, OECD (Oct. 2014),

http://www.oecd.org/sti/inno/data-driven-innovation-interim-synthesis.pdf, at 29 (citing C

ARL SHAPIRO & HAL R.

VARIAN, INFORMATION RULES: A STRATEGIC GUIDE TO THE NETWORK ECONOMY (1999)).

See Jan Wolfe, Arista to Pay $400 Million to Cisco to Resolve Court Fight, REUTERS (Aug. 6, 2018),

https://www.reuters.com/article/us-cisco-arista-settlement/arista-to-pay-400-million-to-cisco-to-resolve-court-fight-

idUSKBN1KR1PI; Don Clark, Arista Networks Files Antitrust Claims Against Cisco, W

ALL ST. J. (Jan. 25, 2016),

https://www.wsj.com/articles/arista-files-antitrust-charges-against-cisco-1453754642.

2) Barriers to Entry Created by Consumer Behavior

Market power is, ironically, generated by the very consumers who are harmed by it. Consumers

do not replace the default apps on their phones, do not scroll down to see more results, agree to

settings chosen by the service, and take other actions that may look like poor decisions if those

consumers like to choose among options and experience competition. Often the actions needed to

generate choice for the consumer seem trivial, such as a download and installation, opening

another app, or a few clicks. Consumers make these “mistakes” because of inherent behavioral

biases such as discounting the future too much and being too optimistic. The situation is worse

when the information needed to counteract bias is hard to obtain. For example, consumers tend

not to run the same search on a different search engine to compare the results, so they may never

find out the relative quality of the default search engine they use.

Research into behavioral biases and their consequences is several decades old and has already

produced ample and convincing evidence of the nature and existence of these biases on the part

of consumers. In 2002, Daniel Kahneman received the Nobel Prize in Economic Sciences for his

pioneering work on behavioral economics.

Fifteen years later, the Nobel Prize was awarded to

Richard Thaler in recognition of the progress of research in this domain and of the huge amount

of knowledge that has been produced.

In his Nobel lecture, Thaler pointed to the key insight from Kahneman and Amos Tversky:

people exhibit “bounded rationality,” meaning that in a complex world they use consistent rules

of thumb to make predictions and decisions. Thaler called on economists to use “psychological

realism” to improve their understanding of human decision making.

That psychological realism

combined with the economics toolkit produced the field of behavioral economics.

Behavioral economics has had a profound influence in the conduct of economic policy that will

become even more prevalent as more knowledge is digested and applied.

It is of great

relevance for our understanding of internet economics because, as information flows improve

and some physical barriers are removed, human factors are more likely to provide the frictions

that have increasing effects on market outcomes. These frictions in decision-making, and the fact

that consumers can be manipulated to take advantage of their biases, render consumers sticky—

Daniel Kahneman—Facts, NOBEL PRIZE (2019), https://www.nobelprize.org/prizes/economic-

sciences/2002/kahneman/facts.

Richard Thaler—Facts, NOBEL PRIZE (2019), https://www.nobelprize.org/prizes/economic-

sciences/2017/thaler/facts.

Richard H. Thaler, From Cashews to Nudges: The Evolution of Behavioral Economics, 108 AMER. ECON. REV.

1265, 1266-1267 (2018).

See, e.g., Cass R. Sunstein, The Storrs Lectures: Behavioral Economics and Paternalism, 122 YALE L.J. 1826

(2013); Richard H. Thaler & Cass Sunstein, Libertarian Paternalism, 93 A

MER. ECON. REV. PAPERS &

PROCEEDINGS 175 (2003).

that is, people are slower to move to a superior product than they would be absent the

manipulation. This in turn makes demand less contestable and less favorable for an entrant.

One lesson from behavioral economics is that small differences (nudges) in how choices are

presented can have large effects on what choices people make.

A nudge to use a particular

browser as a default, for example, can entrench a platform’s browser. Another lesson is that

consumers overweight their immediate benefit relative to their welfare in the future. A consumer

searching for a solution to a particular problem will be inclined to click or use the first result or

recommendation, rather than searching on another page or scrolling down to examine many

listings. The tendency to choose in this manner entrenches the market power of the platform that

can control the display of content. Similarly, consumers’ preference for instant gratification may

lead them to sign away privacy rights they otherwise say they value.

This allows incumbent

platforms to gather data from these consumers that further entrenches their market position. In

general, the findings from behavioral economics demonstrate an under-recognized market power

held by incumbent digital platforms.

A second way consumers create entrenched market power is by single-homing. A multi-homing

user, for example, checks the price of a ride on both Uber and Lyft each time she needs a car. A

user that single-homes bestows market power on the platform she uses exclusively because

advertisers and other content providers can only get the user’s attention by going through that

platform. While users sometimes have the ability to employ multiple services, there is usually a

convenience cost to doing so. Making multi-homing easier will be a key element in encouraging

competition.

3) Barriers to Entry Created by an Incumbent Rival

Because platform market power is higher when users single-home, platforms try to get users to

do so. Sometimes this results in higher quality services: A mapping service can alert a user that it

will take 45 minutes to get to the airport when it sees a flight entry on the user’s calendar. If the

user buys those services from different platforms, she does not get the alert and experiences

lower quality. However, it is often possible for the data from one service to be read by a rival, so

platforms may also encourage single-homing by preventing interoperability.

See Thaler, supra note 19, at 1283.

See Alessandro Acquisiti et al., The Economics of Privacy, 54 J. ECON. LIT. 442 (2016); Alessandro Acquisti et

al., Privacy and Human Behavior in the Age of Information, 347 S

CIENCE MAG. 509, 510 (2015); Alessandro

Acquisiti, Privacy in Electronic Commerce and the Economics of Immediate Gratification (Fifth ACM Conference

on Electronic Commerce, 2004), https://www.cs.drexel.edu/~greenie/privacy/privacy-gratification.pdf; see also

Americans’ Attitudes About Privacy, Security, and Surveillance, P

EW RESEARCH CTR. (May 20, 2015),

https://www.pewresearch.org/wp-content/uploads/sites/9/2015/05/Privacy-and-Security-Attitudes-

5.19.15_FINAL.pdf, at 17 (reporting that 74% of survey respondents believe it is “very important” to be “in control

of who can get info about you”).

By limiting a rival’s access to data, dominant firms can make exclusive reliance on their service

either inevitable or the clearly best decision. Sharing data or allowing access to certain pieces of

information is often feasible at a technical level, but it is not normally in the interest of the

platform that could lose its users. For example, Google has been able to limit its users from

visiting (multi-homing on) competitors such as Yelp by displaying their information in the

search window directly.

To some extent, the limited switchability to the rival is driven by

asymmetric information in two places. First, consumers are generally not aware of, or attentive

to, the costs of failing to switch; for example, the quality of the content and the extent to which

the platform steers consumers to inferior content or product choices may be hard to see. Second,

there is often no way to compensate the consumer directly with lower prices—as an entrant such

as Warby Parker can do by selling eyeglasses directly to consumers at a lower price—because

the money price of services from a platform is often fixed at zero for both the incumbent and the

rival.

For all of these reasons, digital platform market power can become entrenched. The United

Kingdom, the European Commission, Australia, and Germany have all published reports

concluding that digital platforms’ market power has indeed become entrenched.

Surmounting

the existing barriers to entry created by consumer behavior, cost structure, public policy, and any

past anticompetitive conduct is extremely difficult. This fact has direct effects on consumers:

without entry or the credible threat of entry, digital platforms need not work hard to serve

consumers because they do not risk losing their consumers to a rival.

b. How Big Data Affects Digital Products and the Sale of

Advertisements

i. What Makes Data Big? And Does Big Data Harm

Consumers?

1) Targeted Advertising

Technology firms claim that they perform machine learning on big data, and that doing so gives

them both a competitive edge over rivals and allows them to better tailor their services to their

consumers. How might this work in practice? Consider a search engine with information about a

See Charles Duhigg, The Case Against Google, N.Y. TIMES MAG. (Feb. 20, 2018),

https://www.nytimes.com/2018/02/20/magazine/the-case-against-google.html

See, e.g., Competition Decision, Case COMP/AT.39740—Google Search (June 27, 2017),

http://ec.europa.eu/competition/antitrust/cases/dec_docs/39740/39740_14996_3.pdf. Indeed, as the U.K.

Competition and Markets Authority has found, even in industries where prices exist but there are substantial search

frictions, consumers do not make choices that would give them substantial savings. See note 31, infra, and

associated text.

See U.K. COMPETITION REPORT, supra note 1, at 75; EC COMPETITION REPORT, supra note 1, at 112;

USTRALIAN COMPETITION REPORT, supra note 1, at 35; GERMAN COMPETITION REPORT SUMMARY, supra note 1,

at 2.

given user’s search history, including the fact that this user recently searched for a specific pair

of Nike running shoes. The ad service can more effectively spend advertisers’ budgets by

showing the user ads focused on running-related products (or similar shoes to those the user

searched for). Advertisers will get a greater return on investment because more of their ads will

be shown to users who have demonstrated interest in running products—possibly very similar

running products to the ones being advertised. Moreover, users may actually prefer seeing ads

for other running products compared to more generic ads. The more personalized matching of

advertisement to potential customer would appear to help both the producer and consumer of the

advertisement.

This simplified example of personalized advertisement uses very little information about a

particular user (only that she had a particular search query) and little to nothing about other users

or another augmenting dataset, with the exception of a database of running-related items. Indeed,

this level of personalization could be similarly achieved in many low-tech spaces (e.g., running

products advertised in running magazines, whose readers have expressed interest in running

simply by reading the magazine). What is different about personalized advertisement if the ad

server has much more data at its disposal?

Box I

Platform vs. Brick-and-Mortar Advertising and Targeting

Traditional brick-and-mortar stores and online platforms differ greatly in their advertising and

personalization capabilities.

At the highest level, local grocers tend not to force shoppers to identify themselves when they

shop, rarely verify identification if used, and rarely have the ability to merge purchase history

with other detailed information from other aspects of their customers’ lives to design targeted

advertising. Online retailers, on the other hand, almost always require account creation for

purchasing, verify this information for each transaction, and have direct or easy access to

detailed non-shopping information about their customers.

Local grocery stores are capable of some data collection and personalization. For example, they

normally know that a majority of their customers live relatively close to the grocery store, so

they can rely on their knowledge about the general demographics of the neighborhood

population. The stores may also ask that consumers use a loyalty card to receive discounts,

which then allows them to track per-customer purchasing patterns, offer particular customers

with certain purchasing patterns particular coupons or free products, and see the result of such

offers on a user’s purchase history. Nonetheless, the effectiveness of these methods is limited

in comparison to digital shops. Consumers normally retain the option to buy full-priced

groceries without tracking (pay in cash), or to intermittently switch their cards with other

people they know (for example). More importantly, even if grocers used fingerprints or other

unique identifiers to remove the possibility of anonymity, the data they would track would still

be entirely comprised of grocery purchase history for its customers.

2) Dimensions Along Which Data Can Be Big

In order to discuss the use-cases of large-scale datasets in online personalized advertising, it

helps to define in what sense a dataset can be big. Big data commonly refers to two very

different properties of a dataset: either that the dataset has many people’s data in it, or that the

dataset has a great deal of information about each person in the dataset. For simplicity we refer to

the former as “large population datasets” and the latter as “high dimensional data.” These distinct

ways in which a dataset can be big enable very different uses.

Large population datasets allow the possessor to infer both unknown attributes of current users,

and statistical facts about individuals not currently in the dataset. If the dataset contains many

instances of users’ queries along with their locations at the time they made those queries, an

advertiser could use future users’ locations to help predict what those users might search for or

be interested in seeing advertisements about. Numerous statistical techniques can be used to

show that, for a large population dataset, simple statistics that hold true on the dataset should also

hold true for fresh users, assuming they come from a similar pool as those in the dataset. For

example, if 30% of queries in the database originating from Cape Cod searched for lobster rolls,

there is a 30% chance that a query made by a future user based in Cape Cod will be for lobster

rolls, assuming the user is visiting in a similar time frame that the dataset was gathered in,

arrived at the search engine in a manner similar to other dataset users, and the number of queries

in the dataset originating in Cape Code is sufficiently large. If the dataset has many queries but

Platform vs. Brick-and-Mortar Advertising and Targeting (Cont.)

This limitation does not apply to digital platforms. Online groceries require a customer to make

an account prior to allowing any purchase. Thus, all purchases belong to a particular account,

which also contains identifying information about a customer including their search history,

billing, shipping, and email addresses. This ensures that (potentially) all search history and all

purchases sent to a particular address (or paid for with a certain mailing address) can be collated

and analyzed by the seller and used to design offers or other services for a customer. If this

seller has a broader set of services than just grocery retail (for example, the seller is Google or

Amazon), the information they can collate together with this purchase history might include

email transcripts, calendar information, or search and purchase history of non-grocery products,

amongst others. All of this information can allow for much more aggressive marketing to a

particular consumer, based on many aspects of their lives beyond their historical grocery

shopping habits.

A grocery store can take advantage of hungry shoppers by placing a

representative selection of junk food in the checkout line. A digital platform, by contrast, can

design an individualized tempting “checkout offer” based on that

particular consumer’s

purchase history, current behavior, time of day,

and the emotional content of recent

communications.

fairly few from Cape Cod, statistical techniques will provide lower confidence in their ability to

predict future queries emanating from Cape Cod.

High-dimensional datasets allow for different uses than large-population datasets do. Suppose a

dataset contains only a few users, but each user’s entry contains their entire email history. A

quick read of a user’s recent emails could give an advertiser a very clear window into what that

user might be looking to purchase. The richness of a given user’s data entry can describe in great

detail many facts about her, including her future travel itineraries, plans for large purchases,

information about her career and social networks, and so forth. Looking only at that user's

information, a deft advertiser could likely select any number of products this user would be much

more likely than an average person to find interesting. Hotels in Cape Cod would much prefer to

advertise to a user if her recent emails describe travel plans to Cape Cod. However, if only three

or four users belong to the dataset, an advertiser can learn very little about new users from

studying the rich but poorly-populated dataset; again, statistical techniques have much lower

confidence when the datasets they operate on are small.

The most useful datasets, from the perspective of an advertiser or other service provider, are

large in both senses: they contain rich information about a huge number of people. This is even

more true when the rich information contains different types of data, for example, email,

location, and search queries. A dataset of this type allows the provider to both learn high-level

population statistics (for which it needs a large population in its dataset) and to carefully tailor its

ads to each individual in its dataset (because it has very rich information about users in the

dataset). Even more interestingly, the complexity of population-level statements one can make

from such a dataset increases. This occurs both because each user's data has more dimensions, so

there are more relevant hypotheses to explore, and because as the number of users in the dataset

grows, so does the statistical significance of any particular statement that holds for the dataset.

3) Accuracy (and Utility) May Display Increasing Returns

Entrants could have an opportunity if marginal returns were to fall as providers’ datasets grow,

because the incumbent’s marginal cost of acquiring new data would eventually exceed the

marginal value. But a simple model shows this may not be the case. Given a fixed dataset of a

particular number n of individuals with a certain number k of features (also referred to as

attributes), the dataset owner does not face decreasing marginal returns with respect to either n or

Why might datasets show increasing marginal returns in either the number of features or the

number of rows already present in the dataset? For a formal treatment of this question, please see

the discussion below which works out in detail a particular example in the advertising domain.

This holds even in “natural” settings, where each member of the dataset is drawn identically from a normal

distribution.

Informally, imagine you are trying to sell a service to people who live in Manhattan and are

planning a wedding. Knowing either that a person lives in Manhattan, or that they are planning a

wedding, may make almost no difference in determining their willingness to use the service

because both are very low-probability events, while having both pieces of information together

allows the Manhattan wedding planner to easily pinpoint interested consumers. More generally,

as a firm accumulates more information about more people, its marginal returns to new data need

not fall.

This last point helps us understand why data has increasing marginal returns. Even if the dataset

is large enough that it allows a company to make accurate inferences about a given population,

the company will always benefit from having specific information about a given individual,

allowing it to become more and more confident about what the consumer wants, and to better

tailor its services and ads.

Companies therefore have no incentives to stop looking for and accumulating new pieces of data,

entrenching incumbents with large datasets vis-à-vis entrants with smaller databases. Consumers

on the internet leave numerous traces of their activities across a range of applications (for

example, their location, what they buy, who they talk to, and what they say), and technology

allows platforms to identify and analyze these traces. The amount of data on individual behavior

that can be collected, merged, analyzed, and stored is rising, and the combination of different

dimensions of data generates valuable information about individuals’ tastes and behaviors. As

individuals rely more and more on a platform to organize their lives through their online social,

cultural, or economic activity, their data become more informative about their future choices and

firms are willing to pay to influence those choices. Furthermore, the emergence of the Internet of

Things means that platforms will have access to yet more data generated by home appliances,

cars, and other devices. Indeed, consumers’ devices can now track eye movement, mouse

movement, body movement, and body position. In parallel with the evolution of the internet that

made tracking of billions of individuals possible, advances in data mining and artificial

intelligence have enabled firms to learn more from data than was conceivable a few decades

ago.

Many digital markets have tipped and therefore there are only a few entrenched platforms able to

gather this breadth of data. As Alessandro Acquisti and colleagues concluded in a recent study,

“a few ‘gatekeeper’ firms [will be left] in a position to control the tracking and linking of . . .

behaviors across platforms, online services, and sites.”

When data exhibit increasing returns to

Jaron Lanier, a creator of virtual reality, has warned of the problems associated with a “surveillance economy” in

which users of digital tools and platforms would be enticed to give up personal data in exchange for “free” products

and access. See J

ARON LANIER, WHO OWNS THE FUTURE? (2013). The data, Lanier argues, could then be monetized

by the owners of the platforms and applications, largely through the sale of the data to advertisers and others finding

value in the users’ personal characteristics and proclivities.

Acquisti et al., supra note 22, at 444.

dimensionality and size, platforms that can track many users across those dimensions will have

economies of scale and scope; they will be able to sell more valuable advertising.

Nonetheless, this targeting can also raise the quality of services provided by platforms. When

they can identify individual tastes at fine levels and personalize their services to this taste, they

often improve people’s lives. Search engines can better answer queries or find a nearby

destination, cultural and news websites are able to suggest well-suited content, and ecommerce

websites can improve matching between buyers and sellers. These are all part of the consumer

benefit described previously.

4) Types of Data

The definition of what constitutes a consumer’s data can become complicated. Suppose the

consumer’s data and those of similar consumers allowed the service to infer characteristics about

the consumer (for example, that she has a tattoo) without any action or communication on the

topic by the consumer. Is this piece of information part of the “consumer’s” data, or is it the

intellectual property of the algorithm owner?

The Vestager Report and the Furman Report categorize data as volunteered , observed, or

inferred. Volunteered data is intentionally provided by the user to the service -- for example,

when a user provides their favorite TV shows to a service in order to receive recommendations.

Observed data, such as the history of shows that the consumer actually watches, are

automatically gathered by a service. Some observed data may not be intentionally provided by

users if they do not understand the privacy protections in the service -- for example, location

tracking of a person using a video app. Finally, the service can process volunteered and observed

data to infer additional information about the user or a group of users.

We include (but do not

always distinguish among) all these types of data in this report’s discussions.

See EC COMPETITION REPORT, supra note 1, at 25. Famously, Target claims to have been able to predict a

customer’s pregnancy before she knew about it. See Charles Duhigg, How Companies Learn Your Secrets, N.Y.

IMES (Feb. 16, 2012), https://www.nytimes.com/2012/02/19/magazine/shopping-habits.html.

Box II

Why Data Has Increasing Marginal Returns

For a particular dataset, suppose the first feature is the one the owner wants to predict for

future users. For example, this feature might refer to whether or not a user will book a

particular hotel if shown an advertisement for it. If k=2, and the second feature represents the

home zip code of a user, there might be some limited ability to predict interest in a hotel

based on this zip code being sufficiently far from the hotel. If k=3, consider the case where

the third feature refers to the annual household income of each person.

With both zip code and income available, we can express much more complex prediction

rules for interest in hotel X. For example, we could predict that a user will book the hotel if

 ∙  +  ∙  ≥ ,

for some α, β, γ ≥ 0. If the true relationship between these three features is in fact linear, then

the utility of gathering the income variable will be marginally decreasing in the set of other

(linearly relevant) features.

However, it might be that the best prediction of whether a hotel will be booked is a nonlinear

function of distance and income; perhaps a person will book exactly when they live far

enough away and make at least a certain amount of money:

 ≥  and  ≥ 

where users with distance > α and income > β make up 10% of the distribution. Suppose 40%

of the users have distance > α, and 40% of the users have income > β.

ii. Data Control and Ownership

1) Status Quo

In the United States at present, a consumer has no property or control rights over their data. This

stands in contrast to Europe, where the GDPR sets a standard for what a digital business can do

with a consumer’s data.

Why Data Has Increasing Marginal Returns (Cont.)

Knowing a user’s zip code (and corresponding distance) gives some limited ability to predict

when both the income and distance are sufficiently large, but knowing both the zip code and

income will allow them to perfectly predict the 10% of the population that might have

interest in staying at the hotel. For this reason, the utility of data owners need not be

marginally decreasing in the set of features they have.

To be explicit, fix an advertising firm for the hotel. If the advertising firm purchases the

ability to see each potential advertisee’s zip code, then showing the advertisement to

customers as a function of their distance is now possible. For example, the advertiser could

only show the advertisement to customers whose distance is more than β, increasing the

probability that the ad is shown to an interested party from 10% to 25%. If the firm then

acquired the ability to estimate each advertisee’s income, it could use both features together

and only show the advertisement to customers with both income > α and distance > β.

This would raise the probability that a targeted advertisee would be interested in the ad from

25% to 100%. The two features are otherwise identical, and so it would also be the case that

first acquiring income and then distance would raise the probability of an advertisement

reaching an interested customer from 10% to 25% to 100% if one started with no features,

then added income, then distance. Therefore, the utility increase which comes from adding

distance first is .15 * (value of showing the ad to an interested customer), while the utility

increase from adding distance after first acquiring income is .75 * (value of showing the ad to

an interested customer). That is, the distance feature is more valuable after first learning

about income (and vice versa). This shows that the value of features may not be marginally

decreasing in the set of features accrued thus far.

One can also use the same sort of relationship to argue that the dataset owner's utility need not

be marginally decreasing in n, the number of people in the dataset. Again, suppose one can

only predict interest in the hotel with the conjunction of zip code and income information, but

that this relationship is not a priori known to the dataset owner, and they are instead trying to

learn how to predict hotel interest based on a dataset they have available to them. Using

statistical techniques, it won’t be possible to learn rules of this form (if x and y then z) to a high

degree of accuracy unless the dataset has sufficiently large n. The utility of adding additional

people to the dataset can increase as the dataset grows, depending on the owner's utility for

accuracy of the learned model.

In the U.K., regulators have sought to give consumers useful control over their financial data.

After a lengthy investigation showed that consumers rarely switch banks despite large gains to

doing so, regulators sought to use Open Banking to give people the ability to seamlessly move

between banks.

The Open Banking Initiative allows consumers to obtain, see, and transmit

their banking activity in a standardized and secure fashion to regulated and approved third-party

firms.

This is an example of data portability. Theory suggests that consumers will use that

power to move their business to banks that lower prices and improve services. If Open Banking

causes more competitive outcomes, it may provide a strong model for regulated portability and

interoperability in other markets.

The Vestager Report defines protocol interoperability and

data interoperability as stronger than data portability because they allow continuing

communication between two services. It defines full protocol interoperability as a complete

linking of the two services in a way that reduces network effects. We will return to these

concepts below.

Data intermediaries collect consumers’ information that they then sell to third parties. These

intermediaries may be large websites obtaining the information through their service to

consumers, or data brokers. Data have specific features that make this market unusual. For

example, data are “non-rivalrous” meaning a broker can sell the same dataset to many buyers and

still retain it, unlike, for example, the sale of apples. Once you’ve sold an apple, you’re out an

apple; once you’ve sold a given piece of information to one buyer, you can sell it again to

another buyer. A key feature of data in this context is that data may be either directly shared with

the buyer, or withheld (to prevent the buyer from achieving its own economies of scale and

scope) but embedded into a service the buyer wants, such as targeting advertising.

See Alasdair Smith et al., RETAIL BANKING MARKET INVESTIGATION: FINAL REPORT, U.K. COMPETITION &

MARKETS AUTHORITY (Aug. 2016),

https://assets.publishing.service.gov.uk/media/57ac9667e5274a0f6c00007a/retail-banking-market-investigation-full-

final-report.pdf, at xiii-xiv.

See U.K. COMPETITION & MARKETS AUTHORITY, THE RETAIL BANKING MARKET INVESTIGATION ORDER 2017

(2017),

https://assets.publishing.service.gov.uk/government/uploads/system/uploads/attachment_data/file/600842/retail-

banking-market-investigation-order-2017.pdf; see also O

PEN BANKING LTD., OPEN BANKING: GUIDELINES FOR

PARTICIPANTS (2018), https://www.openbanking.org.uk/wp-content/uploads/Guidelines-for-Open-Data-

Participants.pdf; Rowland Manthorpe, What is Open Banking and PSD2? WIRED Explains, W

IRED (Apr. 17, 2018),

https://www.wired.co.uk/article/open-banking-cma-psd2-explained. Third party firms that customers can elect to

give data access are regulated and approved by the Financial Conduct Authority. See Press Release, U.K. Fin.

Conduct Auth., FCA Finalises Revised Payment Services Directive (PSD2) Requirements (Sept. 9, 2017),

https://www.fca.org.uk/news/press-releases/fca-finalises-revised-psd2-requirements.

There are some signs that Open Banking has seen early success. See U.K. COMPETITION REPORT, supra note 1, at

70.

See Dirk Bergemann & Alessandro Bonatti, The Wild West of Information Markets: What We Need to Know

Before Law and Order Can Rule, V

OX (Oct. 11, 2018), https://voxeu.org/article/wild-west-information-markets;

Dirk Bergemann & Alessandro Bonatti, Markets for information: An Introduction, (Cowles Foundation Discussion

Paper No. 2142, Aug. 29, 2018), https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3240310.

2) Externalities

What can a person do with ownership over her data? The previous section describes two ways in

which data might be used: First, it might be used to tailor services to a given person; second, it

might be used to learn patterns that hold on average for the population from which the dataset

was drawn. The former does not necessarily transfer any knowledge about one customer to the

treatment of another, while the latter aims to learn about some fraction of a dataset and use that

information to affect interactions with future customers. Note that this use of data creates

externalities between consumers. Purchase or travel patterns by one person are used to create

recommendations or suggestions for similar people, where “similar” is determined by machine

learning. For this reason, the value of an individual person’s data is more than the value of using

it to market to them. A customer whose data predicts the behavior of many other consumers is

very valuable to a platform.

3) Deletions

One might simplistically think that “owning” one’s own data gives a person the ability or right to

delete it. The ease with which a customer’s data can truly be deleted depends on the way the firm

has created and is using that data. If a user’s own data is used only to personalize her experience,

deleting her data will mean that the dataset—and the service’s subsequent behavior—has no

trace of the customer ever belonging to the dataset. If, however, the service used the customer’s

data to inform population-wide behavior (by analyzing the dataset including this customer and

storing the results of those analyses), merely deleting the customer’s data from the dataset does

not truly erase her presence—her data will still have a lasting effect on the future behavior of the

service.

For example, suppose the service scans the dataset for one representative user from each zip code

and remembers the purchasing history of each representative user. Even if the service removes

the initial entry corresponding to a particular user in the original dataset, if the service still uses

this set of representative users’ purchasing histories, some of the user’s data can still remain in

the system. While this example may seem contrived, many machine learning methods do some

amount of memorization of some subset of their training data.

Beyond memorization, there are other, more subtle ways in which a user’s data can affect the

long-term behavior of a system and what information the system holds on to. The natural tool to

restrict how much a system’s information is affected by one user’s data is to impose differential

privacy. Loosely speaking, differential privacy restricts the statistics created from the dataset to

not reveal if any particular person or observation is in the dataset.

For example, support vector machines are explicitly recorded as a small number of datapoints on which the model

was learned.

4) Transparency, Choice, and Fraud

According to the Federal Trade Commission, the market for data suffers from a lack of

transparency.

Most consumers have no idea how much information is being collected about

them, sold, and used to make a profit.

One way in which digital platforms exploit their market

power is by requiring consumers to agree to terms and conditions that are unclear, difficult to

understand, and constantly changing. The terminology in these contracts is legal and the

documents are often lengthy; the consequence of the different clauses is difficult to understand

and foresee. Moreover, the user of a device or platform makes a choice to sink investments

(posts, calendars, media, and so on) in a particular platform at a moment in time when a

particular user agreement is in force. After the device or platform updates its terms of service, a

user may be locked in. Having bought a phone, they won’t immediately want to switch; having

built a network of friends, they often won’t want to leave. Thus, the user does not have the same

set of choices as she did the first time she hit “agree,” and some agencies have argued this is no

longer a free choice.

Lastly, a digital platform may describe its data-use policy in its terms of

service and then deviate from that, rendering the initial statement fraudulent.

FED. TRADE COMM’N, DATA BROKERS: A CALL FOR TRANSPARENCY AND ACCOUNTABILITY: A REPORT OF THE

FEDERAL TRADE COMMISSION (2014), https://www.ftc.gov/system/files/documents/reports/data-brokers-call-

transparency-accountability-report-federal-trade-commission-may-2014/140527databrokerreport.pdf.

See id. at 42.

See Press Release, Bundeskartellamt, Bundeskartellamt Prohibits Facebook from Combining User Data from

Different Sources (Feb. 7, 2019),

https://www.bundeskartellamt.de/SharedDocs/Meldung/EN/Pressemitteilungen/2019/07_02_2019_Facebook.html

(determining that Facebook’s take-it-or-leave-it under agreements constitute “abuse of market power”).

See, e.g., Tony Romm & Elizabeth Dwoskin, U.S. Regulators Have Met to Discuss Imposing a Record-Setting

Fine Against Facebook for Privacy Violations, W

ASH. POST (Jan. 18, 2019),

https://www.washingtonpost.com/technology/2019/01/18/us-regulators-have-met-discuss-imposing-record-setting-

fine-against-facebook-some-its-privacy-violations.

Box III

Digital Identities

Perhaps the next major shift in digital competition will be the quest to control the identification

market. Once we create an account with any digital platform, we create a digital identity which

incorporates select data on age, sex, address, email address, preferences, and, frequently, much

more. These digital identities help companies identify and tag users to the data they generate,

be it transactional, social, simple web navigation or even meta data. Digital identities work like

access and tracking mechanisms, allowing a user to surpass a wall in exchange for enabling the

company to link the data to a given dataset (e.g., using a Facebook login to access a website).

Identity data and control is highly valuable and platforms can monetize it in many ways. They

can use it to personalize services and charge subscription fees, provide advertising or market

intelligence, or as a way to increase bottleneck power and charge companies for user access .

The more privacy protection technologies grow and limit widespread data collection online, the

more being the single identification point for users will grow in importance and value.

iii. Digital Platforms are Characterized by Free Services

“Free” is not a special zone where economics or antitrust do not apply.

A free good is one

where the seller has chosen to set a monetary price of zero and may set other, non-monetary,

conditions or duties. Zero is a number, just like 10 is a number. If a competitive price is $10

while the realized price is $15, there is a $5 markup above the competitive price. This is the same

harm the consumer bears when the competitive price is -$5 (at this price the digital platform is

actually paying the consumer for her data and information in addition to providing her with

services) and the realized price is $0.

Barter is a common way in which consumers pay for digital services. They barter their privacy

and information about what restaurants they would like to eat in and what goods they would like

to buy in exchange for digital services. The platform then sells targeted advertising, which is

made valuable by the bartered information. But, in principle, that information has a market price.

It is not easy to see if the value of any one consumer’s information is exactly equal to the value

of the services she receives from the platform. However, many digital platforms are enormously

profitable, and have been for many years, which suggests that in aggregate we do know the

answer: the information is more valuable than the cost of the services. The economics literature

has modeled this setting and is able to define a data markup.

The current inability to use both positive and negative prices for digital goods means that the

policy discussion cannot focus on dollars alone as the unit of cost. Rather, digital platforms

should be analyzed using both price and quality. “Quality-adjusted price” is a metric often used

by economists in this situation. If a platform’s price is fixed at zero and the quality of the service

improves, then its quality-adjusted price has fallen. Conversely, if a platform’s price remains

zero but its quality falls, its quality-adjusted price has risen. When the price is fixed at zero, it is

possible to track quality-adjusted price over time: the movement in quality accurately reflects

quality-adjusted price.

Online platforms offer many services for zero monetary price while they try to raise participation

in order to generate advertising revenue. Free services are prevalent on the internet in part

See, for example, the current head of the U.S. Department of Justice’s Antitrust Division describing the challenge

faced by his division of defining markets when goods are free. Makan Delrahim, U.S. Assistant Att’y General, “I’m

Free”: Platforms and Antitrust Enforcement in the Zero-Price Economy, Address at Silicon Flatirons Annual

Technology Policy Conference at The University of Colorado Law School (Feb. 11, 2019),

https://www.justice.gov/opa/speech/assistant-attorney-general-makan-delrahim-delivers-keynote-address-silicon-

flatirons (arguing, without regard for quality effects, that traditional antitrust market definition cannot work because

“[w]e cannot look at the effects of a five percent increase in price because five percent of zero is still zero”). See

also id. (arguing that “[c]hoosing variables for measuring market shares also can be more complicated where shares

of revenue is not an option”).

See Dirk Bergemann et al., The Design and Price of Information, 108 AMER. ECON. REV. 1 (2018).

The European Commission has noted the importance of assessing product quality in zero-price settings. See

Samson Y Esayas, Competition in (Data) Privacy: ‘Zero’-Price Markets, Market Power, and the Role of

Competition Law, 8 I

NT. DATA PRIVACY L. 181 (2018).

because internet firms can harness multi-sided network externalities. While the low price can be

a blessing for consumers, it has drawbacks for competition and market structure in a world where

institutions have not arisen to manage negative prices. Because there is currently no convenient

way to pay consumers with money, platforms are able to mark up the competitive price all the

way to zero. This constraint can effectively eliminate price competition, shifting the competitive

process to quality and the ability of each competitor to generate network externalities. Depending

on the context this may favor or impede entry of new products. For example, entry will be

encouraged when a price of zero leads to supra-competitive profits, and impeded when a zero

price prevents entrants from building a customer base through low price. Moreover, unlike

traditional markets where several quality layers may coexist at different price levels (provided

that some consumers favor lower quality at low price), markets where goods are free will be

dominated by the best quality firm and others may compete only in so far as they can

differentiate their offers and target different customers. This strengthens the firm’s incentive to

increase quality through increasing fixed costs in order to attract customers (known as the Sutton

sunk cost effect) and further pushes the market toward a concentrated market structure.

It is a puzzle that, to date, no entrepreneur or business has found a way to pay consumers for

their data in money. For example, a consumer’s wireless carrier could aggregate micropayments

across all manner of digital destinations and apply the credit to her bill each month. It may be

that adverse selection, transaction costs, and coordination difficulties create too large a barrier

for today’s entrepreneurs, though technical solutions like a verifiable digital identity, mentioned

above, would combat adverse selection.

Furthermore, a carrier that could bargain effectively

with platforms on behalf of its subscribers for high payments would likely gain subscribers.

Notice that an easy method to pay consumers, combined with price competition for those

consumers, might significantly erode the high profits of many incumbent platforms. Platforms

likely have no economic incentive to work diligently to operationalize negative prices.

See SUTTON, supra note 7.

See ERIC A. POSNER & E. GLEN WEYL, RADICAL MARKETS: UPROOTING CAPITALISM AND DEMOCRACY FOR A

JUST SOCIETY (2018).

III. PROBLEMS ARISING IN DIGITAL MARKETS

The changing market dynamics, outlined above, raise doubts about the market’s ability to ensure

ongoing innovation and competition. Increased concentration levels, market power, network

effects, and control over data and analytics have in many digital markets tipped the market in

favor of the incumbents.

The theme we return to throughout this report is the difficulty of entry into digital platform

businesses once an incumbent is established. Whether the entrant is vertical or horizontal, has

succeeded to some degree, is nascent, or is a potential entrant, its existence improves consumer

welfare. Either the entrant provides more choice, different features, and a chance of higher

quality, or the threat of those outcomes spurs the incumbent to provide lower prices, higher

quality and innovation, and to do so more quickly.

By focusing on this feature of digital platforms, we are highlighting the cause of the market

power, not its result. Absent entry barriers of the type discussed above, the tremendous amount

of profit available in these markets would stimulate entry. Protecting competition in these

markets requires protecting competitors. Entry and potential entry create more competitors (in

expectation), and that increase depends on competition working effectively so that a meritorious

entrant can successfully dethrone the incumbent.

The categories of economic harms to consumer welfare from digital platforms are the standard

ones: price, quality, and innovation. The report will primarily emphasize quality and innovation

harms due to their greater complexity and generality. For example, by excluding competitors,

dominant firms do not need to innovate as hard as they otherwise would be required to keep their

customers. Likewise, when platforms do not face competition, they will be able to reduce

quality, for example, by decreasing privacy protections, without losing customers or revenue.

When a service reduces quality without lowering price, it is raising quality-adjusted prices,

which harms consumers. For example, if a phone service were to lower the quality of service but

keep monthly fees the same, it would have raised quality-adjusted price. Because many digital

services are purchased with barter, the monetary price paid by consumers is zero, and quality-

adjusted prices cannot be directly seen the way a nominal price can be. On the advertiser side of

the platform, where monetary prices are charged, harms to competition exist in the more

ordinary form of higher markups for ads and other services. These markups are eventually paid

by consumers because they are built in to the prices of the goods and services that are advertised

online.

See Agustín Reyna, The Psychology of Privacy—What Can Behavioural Economics Contribute to Competition in

Digital Markets?, 8 I

NT. DATA PRIVACY L. 240 (2018).

a. Quality Harms

i. Personal Data and Behavioral Economics

1) Behavioral Economics

As discussed above, behavioral economics helps improve our understanding of real consumer

choices and suggests that consumer exploitation is common. There are a number of systematic

consumer biases that, when incorporated into economic analysis, affect outcomes and welfare.

For instance, individuals can be subject to salience effects, putting excessive weight on the most

salient information. Confirmation bias can lead them to change their preferences to conform with

past choices.

Consumers are often biased toward the status quo even when it is no longer

optimal.

Perhaps the most important consumer biases are impatience and lack of self-control.

The former refers to the discounting of any payoff that occurs further in the future than the

present. The latter is closely related and refers to the extent to which individuals fail to resist

short-term impulses in order to achieve long-term goals. The literature in behavioral antitrust

argues that status quo, salience, and impatience are the most relevant for antitrust analysis.

Platforms that analyze their consumers’ behavior can exploit these biases by framing choices to

make certain information salient, designing a status quo that is profitable, inducing addictive

behaviors, generating sales through impulsive consumption, and exploiting consumers’

disinclination to search. These strategies are common in the brick and mortar world. For

example, the candy aisle in supermarket check-out lines, or the rug store that has a special 50%-

off sale every day.

2) Using Machine Learning to Take Advantage of Consumers

There are now decades of economic research demonstrating consumer bias and firm responses in

offline markets.

For example, gyms offer subscription memberships rather than pay-per-visit

Confirmation Bias, WIKIPEDIA (Mar. 22, 2019), https://en.wikipedia.org/wiki/Confirmation_bias (“Confirmation

bias is the tendency to search for, interpret, favor, and recall information in a way that confirms one's preexisting

beliefs or hypotheses.”); see, e.g., Thaler, supra note 19, at 1266 (“People guess that in the United States today gun

deaths by homicide are more frequent than gun deaths by suicide, although the latter are about twice as common.

The bias comes because homicides are more publicized than suicides, and thus more ‘available’ in memory.”).

William Samuelson & Richard Zeckhauser, Status Quo Bias in Decision Making, 1 J. RISK & UNCERTAINTY 7, 8

(1988) (“[D]ecision makers exhibit a significant status quo bias. Subjects in our experiments adhered to status quo

choices more frequently than would be predicted by the canonical model.”).

See Amelia Fletcher, The EU Google Decisions: Extreme Enforcement or the Tip of the Behavioral Iceberg?,

OMPETITION POL. INT. ANTITRUST CHRON. (Jan. 2019), https://www.competitionpolicyinternational.com/wp-

content/uploads/2019/01/CPI-Fletcher.pdf.

See, e.g., DANIEL KAHNEMAN & THOMAS GILOVICH, HEURISTICS AND BIASES: THE PSYCHOLOGY OF INTUITIVE

JUDGMENT (2002); Nicholas Barberis & Richard Thaler, A Survey of Behavioral Finance, in HANDBOOK OF THE

ECONOMICS OF FINANCE (René M. Stulz & George Constantinides eds., 2003); see also, e.g., Daniel Kahneman,

Jack L. Knetsch, & Richard H. Thaler, Anomalies: The Endowment Effect, Loss Aversion, and Status Quo Bias, 5 J.

ECON. PERSP. 193 (1991).

knowing that members will not come as often as they anticipate,

and credit cards offer teaser

rates knowing that consumers plan to have no debt in 6 months’ time.

However, the strategies

firms have used to date are swamped by what digital businesses can learn by using high-

dimensional, large datasets to explore every nook and cranny of consumers’ many behavioral

shortcomings and biases in real time. This strategy is of serious concern when combined with an

understanding of the digital platform business model discussed above. The platform’s goal is to

use its knowledge of consumers, combined with its market power (and the resulting lack of

consumer choice), to extract profit. Some of that profit may come from value creation, or

“expanding the pie,” but some will come from transfers to the platform from both the consumer

side and the advertiser side. Because individuals are subject to behavioral biases, consumers are

vulnerable to a platform’s exploitative behavior. Additionally, when individuals are workers, as

they often are in the “gig” economy, these tools can be used to advantage the platform against

the worker.

Digital businesses not only have more information than traditional firms, but they have more

variations of products or services and the ability to control the environment and the timing of

choices and offers. For example, a firm can hide a component of a good’s price to let consumers

discover prices only once they have invested significant time and effort into buying the good.

This strategy can be differentially employed depending on the consumer’s past willingness to

pay. Framing, nudges, and defaults can direct a consumer to the choice that is most profitable for

the platform. A platform can analyze a user’s data in real time to determine when she is in an

emotional “hot state” and offer a good that the user would not purchase when her self-control

was higher. Consider a supercomputer tracking a consumer, via her cell phone, around the town

until she is tired and frustrated in some way, and at that moment presenting her ads and

information about junk food. This type of exploitation could depend on input from devices such

as eye-tracking sensors, the ability of AI to understand the emotion expressed in texts and email,

and all the other data the platform has about the consumer combined at a very large scale. This

same tactic can be used to gain advantage against an independent contractor, e.g. a driver, whose

behavior and location can be tracked for long periods.

In addition, machine learning applied to

big data may help differentiate well-informed and sophisticated consumers or workers from

poorly informed or more naïve consumers, raising the possibility of further exploitation of those

least prepared to resist it.

Internet firms make frequent use of digital defaults, framings, and nudges. When a user is

signing up for a new service, the sign-up form may automatically check the box that permits the

service to send the user emails.

A user can opt-out of receiving emails by unchecking the box,

See Stefano DellaVigna & Ulrike Malmendier, Paying Not to Go to the Gym, 96 AMER. ECON. REV. 694 (2006).

See Oren Bar-Gill & Ryan Bubb, Credit Card Pricing: The CARD Act and Beyond, 97 CORNELL L. REV. 967

(2012).

https://www.nytimes.com/interactive/2017/04/02/technology/uber-drivers-psychological-tricks.html

This is no longer permitted in the European Union. See Regulation (EU) 2016/679 of the European Parliament and

the Council of 27 April 2016 on the Protection of Natural Persons with Regard to the Processing of Personal Data

but doing so is harder than sticking with the default. The results of a search that fits onto one

page may all be sponsored, whereas finding the organic links requires paging down. Homo

economicus is hardly influenced by defaults—to a rational agent, scrolling down or unchecking a

box is trivial—but real people are influenced. Nudges are not unique to digital products; for

example, an employer offering a default health care plan to its employees nudges employees

towards choosing that plan—employees can select a different plan, but doing so is harder than

sticking with the default.

What is noteworthy, however, is the platform’s detailed,

personalized, minute-by-minute control over their interface. This control enables platforms to

create a façade of competition, choice, and autonomy when in fact users are being directed with

behavioral techniques.

With big data and machine learning, firms are able to understand and manipulate individual

preferences at a scale that goes far beyond what is possible in traditional markets. This capability

is qualitatively new. The environment is characterized by extreme asymmetries of information

and analytical capacity between the platform and the user. This enables firms to charge higher

prices (for goods purchased and for advertising) and engage in behavioral discrimination,

allowing platforms to extract more value from users where they are weak.

The problem is only

growing; platforms continue to make investments to extract data, encourage stickiness and

addiction, and promote ever-greater use, in order to run data analytics and enable more precise

targeting.

The economic literature suggests that competition by itself cannot resolve the issue raised by the

exploitation of behavioral biases or poor consumer information. This is because staying

profitable in a competitive environment may force firms to exploit behavioral bias to achieve

maximal profitability. Firms abstaining from doing so may be driven out of the market. Rather,

competition causes a shift of surplus to wiser consumers; profit from exploitation of biased

consumers is used to compete for well-informed consumers. For consumers who can guard

against exploitation, there are therefore significant gains to be had from competition. But this

and on the Free Movement of Such Data, and Repealing Directive 95/46/EC (General Data Protection Regulation),

2016 O.J. (L 119/1), § 32 (“Silence, pre-ticked boxes or inactivity should not . . . constitute consent.”).

Thaler, supra note 19, at 1283. Default settings have very large effects, for example, in voluntary organ donation

decisions. Countries with an op-in default generally have the vast majority of the nation’s adults enrolled to donate,

while countries with default opt-out decisions see the opposite result. See Eric J. Johnson & Daniel Goldstein, Do

Defaults Save Lives?, 302 S

CIENCE 1338 (2003).

See ARIEL EZRACHI & MAURICE E. STUCKE, VIRTUAL COMPETITION: THE PROMISE AND PERILS OF THE

ALGORITHM DRIVEN ECONOMY (2016) (comparing the façade of choice and autonomy on digital platforms to the

Truman Show).

For a discussion of the use of online price and behavioral discrimination, see Maurice E. Stucke & Ariel Ezrachi,

How Digital Assistants Can Harm Our Economy, Privacy, and Democracy, 32 B

ERKELEY TECH. L.J. 1239 (2018);

Maurice E. Stucke & Ariel Ezrachi, Alexa et. al., What Are You Doing with My Data?, 5 C

RITICAL ANALYSIS L. 148

(2018).

Haley Sweetland Edwards, You’re Addicted to Your Smartphone. This Company Thinks It Can Change That,

IME (Apr. 13, 2018, 10:28 AM), http://time.com/5237434/youre-addicted-to-your-smartphone-this-company-

thinks-it-can-change-that.

will not be true for all people; some will be taken advantage of under perfect competition if

consumer protection regulations are insufficient. Hence, while this report focuses on the

competition problems created by powerful platforms, related issues raise broader consumer

protection concerns that cannot be solved through greater competition.

ii. The Harms from an Advertising-Supported Business Model

1) Market Power Leads to Markups

Access to high-quality data, scale, and scope has allowed a few large platforms to offer efficient

targeting services for advertising and to dominate the advertising market. Facebook, Google,

and, increasingly, Amazon act as gatekeepers to the online advertising market. The three

platforms provide access to billions of users, as well as a data-rich environment, essential for

modern online advertising. Being in control of the data and the assets—users—creates market

power. That market power generates a profit margin which, for ad-supported platforms, comes

from the sale of advertising. One of the characteristics of the digital advertising environment is

its opacity: major platforms are able to leave bidders and publishers in the dark with respect to

the true success, costs, and profits from placement of advertising. This can be exacerbated when

the platform also supplies buyer or publisher tools and analytics. Opacity is partially a

consequence of market power. Market power can be used to discourage, or even prevent, multi-

homing by buyers. One additional concern is that this opacity may give firms not only the

normal ability to exercise market power in intermediation but also to engage in fraud.

Lack of

transparency also undermines buyers’ ability to measure the effectiveness of digital advertising

and therefore to understand its true value relative to price.

A report by the French Competition Authority estimated that publishers received 40% of

advertiser sales, with intermediaries collecting the rest.

What justifies such a large markup for

intermediaries? Without detailed study we cannot know for sure. However, reasons likely

include the fact that very few platforms can target customers (who may be single-homing) using

detailed and accurate data, so advertisers cannot take advantage of competition. The platforms do

See, e.g., Suzanne Vranica & Jack Marshall, Facebook Overestimated Key Video Metric for Two Years, WALL ST.

J. (Sept. 22, 2016), https://www.wsj.com/articles/facebook-overestimated-key-video-metric-for-two-years-

1474586951; Mike Shields, Facebook Says it Found More Miscalculated Metrics, W

ALL ST. J. (Nov. 16, 2016),

https://www.wsj.com/articles/facebook-says-it-found-more-miscalculated-metrics-1479303984; Cyrus Farivar,

Advertisers Allege Facebook Hid the Fact That No One Watches Video Ads, A

RSTECHNICA (Oct. 17, 2018),

https://arstechnica.com/tech-policy/2018/10/advertisers-allege-facebook-hid-the-fact-that-no-one-watches-video-ads

(reporting allegations that Facebook’s “average viewership metrics were not inflated by only 60-80 percent; they

were inflated by some 150-900 percent”).

This in turn obscures the size of the platforms’ markup. See SELECT COMMITTEE ON COMMUNICATIONS, HOUSE OF

LORDS, 1ST REPORT OF SESSION 2017–19, HL PAPER 116, U.K. ADVERTISING IN A DIGITAL AGE 14, § 29 (2018),

https://publications.parliament.uk/pa/ld201719/ldselect/ldcomuni/116/116.pdf; David Pidgeon, Where Did the

Money Go? Guardian Buys its Own Ad Inventory, M

EDIA NEWSLINE (Oct. 4, 2016),

https://mediatel.co.uk/newsline/2016/10/04/where-did-the-money-go-guardian-buys-its-own-ad-inventory.

FRENCH COMPETITION AUTHORITY REPORT, supra note 1, at 40, § 82.

not sell the data to advertisers, but promise to place the ad in front of the requested demographic.

The advertiser lacks transparency or any ability to learn about its customers and potential

customers. Moreover, a platform that operates in the advertising placement business holds the

keys to a second black box through its control of the pricing process.

The integration of the

business of running the price-discovery mechanisms as well as tools for each side of the platform

enables opacity, which helps maintain market power. When a platform bundles services such as

advertising placement and return-on-investment analysis with advertiser data such as completed

purchases, it can further enhance its market power.

The platform generally never shares the

data with advertisers or publishers, but keeps it to itself, preventing disintermediation by a brand

or publisher, further sustaining its large profit margin.

2) Incentives Created by Markups

Ad-supported platforms’ high markups provide a powerful reason to try and keep users online

for another minute in order to show more ads. These profits push platforms to design their firms

around “engagement”—an obsession with keeping users on their system for as much time, and

with as much attention, as possible.

As much of the behavioral literature cited above shows,

advertising-supported digital businesses can use consumer biases to hold people’s attention in

ways that ultimately harm them. Early empirical work has found such effects in social media. In

a recent working paper, Hunt Allcott and colleagues found that Facebook users who were paid to

leave the site for four weeks wound up with higher subjective well-being than similarly situated

people randomly assigned to a group not offered the payment.

Users who took the break from

Facebook had a “large and persistent reduction in Facebook use after the experiment,” along

with reduced political polarization and news knowledge.

For the platform, engagement serves

two reinforcing purposes. First, the more time a user spends on a platform, the more the platform

knows about her. Second, the longer a user is on the platform, the greater its income from ads

and services. Thus, the more time a user is on the platform, the more ads the platform can sell,

and the more it can charge per ad.

For example, Google’s ownership of the ad platforms formerly known as DoubleClick and AdWords. See Todd

Spangler, Google Killing Off DoubleClick, AdWords Names in Rebranding of Ad Products, V

ARIETY (July 27,

2018), https://variety.com/2018/digital/news/google-doubleclick-adwords-rebranding-1202859088.

See generally Einer Elhauge, Tying, Bundled Discounts, and the Death of the Single Monopoly Profit Theory, 123

ARV. L. REV. 397 (2009).

See, e.g., Mark Bergen, YouTube Executives Ignored Warnings, Letting Toxic Videos Run Rampant, BLOOMBERG

(Apr. 2, 2019), https://www.bloomberg.com/news/features/2019-04-02/youtube-executives-ignored-warnings-

letting-toxic-videos-run-rampant (“The company spent years chasing one business goal above others: ‘Engagement,’

a measure of the views, time spent and interactions with online videos.”).

Hunt Alcott et al., The Welfare Effects of Social Media (NBER Working Paper No. 25514, Apr. 2019),

https://web.stanford.edu/~gentzkow/research/facebook.pdf.

Id. at 1.

This financial incentive may explain the growing use of tracking and the creation of ecosystems

that are based on the ability to manipulate the user into staying longer on the platform.

This

opens the door to exploitative tactics or content as a method of increasing engagement.

Furthermore, a second element to the large margin earned by a platform may be the low cost of

the purchased content that keeps users on the platform. An important question (addressed more

fully by the Media committee) is whether the prices for that content are set competitively. In a

bargaining environment, the market power of the platform may allow it to negotiate a price for

content that is below competitive levels, potentially creating dynamic harm to input providers.

The financial incentive created by a large markup can lead to anticompetitive behaviors as well

as exploitative ones. Platforms may seek to reduce interoperability and awareness of outside

options. For example, platforms may exclude certain services or increase friction in accessing

third parties’ services. High search and switching costs are used to “lock in” users and reduce the

ability of competitors to access those users. Platforms may adopt strategies to reduce multi-

homing to obtain more market power over their users. We will return to this theme below.

3) Resulting Quality of Content

As discussed in detail below, the cheapest way to keep users on a platform is to present content

as being more attractive than users thought at the moment when they were ready to leave the

platform, or to make departure more costly in other ways, so that users stay longer. (See a similar

discussion in the Privacy Report.) The platform is essentially degrading the quality of the content

offered in a way that present-biased human beings find engaging. As discussed further below,

content that instantly engages most effectively is content that generates outrage, not necessarily

content that is truthful or thoughtful. Simple strategies such as more advertising minutes per

minute of content also lower quality. Low-quality content represents an increase in the quality-

adjusted price of platform services experienced by consumers.

This business model, and the opacity that accompanies it, makes it difficult for either advertisers

or consumers to realize they are being charged a markup and makes entry into these advertising

markets extremely difficult. Because the advertiser side is where a platform earns its revenue,

this is obviously a key entry barrier. The question of how an enforcer or regulator might enable

entrants to overcome this barrier to entry is an important part of the discussion in Section III.B of

the report.

See Maurice E. Stucke & Ariel Ezrachi, The Subtle Ways Your Digital Assistant Might Manipulate You, WIRED

(Nov. 29, 2016, 8:00 AM), https://www.wired.com/2016/11/subtle-ways-digital-assistant-might-manipulate.

See Maurice E. Stucke & Ariel Ezrachi, Looking Up in the Data-Driven Economy, COMPETITION POL’Y INT’L

(May 15, 2017), https://www.competitionpolicyinternational.com/looking-up-in-the-data-driven-economy.

4) Welfare and Efficiency

When analyzing the impact of online advertising and the behaviors it creates, one should keep in

mind that antitrust aims to promote effective competition for the benefit of consumers.

Competition authorities tend to treat advertisers as any other customers of a service. The premise

is that market power in advertising channels impedes the ability of brands to reach consumers

and inform them about their products. Such a treatment of the welfare from advertising relies on

the implicit view that i) the role of advertising is solely to inform consumers; and ii) competition

in the advertising market leads to efficiency. The economic literature tells us that both

assumptions are questionable, as does observation of digital markets.

In competition policy,

when assessing the impact of a decision or a regulation involving advertising, it is important to

focus on the final impact on consumers and market efficiency rather than on the advertisers’

surplus.

On the one hand, targeted advertising to wise and well-informed consumers is welfare-improving

insofar as it allows advertisers to send the right information to the right people, improving their

choices and fostering competition among suppliers. On the other hand, in the modern economy

this simple model becomes more complex because of the cost to the consumer, namely loss of

privacy. There is an open empirical question as to whether the tradeoff is worthwhile to

consumers.

We discuss the problem of measuring welfare in a setting where consumers are

manipulated below.

iii. Online Exploitation and Addiction

1) Human Reward Systems

Digital platforms’ manipulation of their users—in part designed to get users addicted—is getting

a considerable amount of attention from the public. These actions lower the quality of a platform

and harm users.

As discussed above, some platforms have deliberately incorporated features that feed human

“reward” centers into their products to induce users to give more and more of their time—and

data—to the platform. These tools are designed for scale—they become even more valuable the

more traffic they carry and the more users they garner—and hence the competition among

producers has been described as a competition for eyeballs.

Because the digital tools and

See, e.g., Anderson, S. P., Waldfogel, J., & Stromberg, D. (Eds.). Handbook of Media Economics, vol. 1A (2016).

Elsevier; Goldfarb, A., & Tucker, C. Digital economics. Journal of Economic Literature, 57(1), 3-43 (2019).

See, e.g., Elizabeth Aguirre et al., Unraveling the Personalization Paradox: The Effect of Information Collection

and Trust-Building Strategies on Online Advertisement Effectiveness, 91 J.

RETAILING 34 (2015); Justin P. Johnson,

Targeted Advertising and Advertising Avoidance, 44 R

AND J. ECON. 128 (2013).

See, e.g., Doni Bloomfield, High-Speed Ad Traders Profit by Arbitraging Your Eyeballs, BLOOMBERG (Nov. 7,

2014), https://www.bloomberg.com/news/articles/2014-11-07/high-speed-ad-traders-profit-by-arbitraging-your-

eyeballs.

networks have been designed for use at scale, and because there is relatively little cost and

considerable benefit associated with adding more users, producers want as much user

engagement as they can get. The creators of digital products have benefited from social science

and neuroscience findings that concern, for example, how certain colors or mechanics can feed a

user’s dopamine, much as nicotine does.

The Center for Humane Technology, which calls the

problem the “hijacking of our society,” describes the issues as follows:

Facebook, Twitter, Instagram, and Google have produced amazing products that have

benefited the world enormously. But these companies are also caught in a zero-sum race

for our finite attention, which they need to make money. Constantly forced to outperform

their competitors, they must use increasingly persuasive techniques to keep us glued.

They point AI-driven news feeds, content, and notifications at our minds, continually

learning how to hook us more deeply—from our own behavior.

This business model is based on acquiring a large volume of data to generate income, and it has

led to unprecedented investment in addiction—ensuring continuing use of the interface. From

diaper apps,

to rewarded ads (in which viewers are given in-app rewards, such as another “life”

in a video game, in return for watching an ad)

internet firms harness their knowledge of

second-to-second individual responses alongside social science learnings about human biases for

maximum attention.

Researchers do not yet know the full extent of the harm (or potential

harm) that may come from this sort of addiction and manipulation, but they are now actively

engaged in relevant research.

What is known is that a business that depends on users staying

online to watch ads and have their preferences harvested will focus its resources on keeping

See Edwards, supra note 56.

Ctr. for Humane Tech., Our Society is Being Hijacked by Technology, HUMANE TECH (2018),

http://humanetech.com/problem.

Edwards, supra note 56.

Rewarded Ads: A Win for Users, Developers, and Advertisers, GOOGLE ADMOB (2018),

https://admob.google.com/home/resources/rewarded-ads-win-for-everyone.

The video app TikTok may have taken this approach to its logical conclusion. See John Herrman, How TikTok is

Rewriting the World, N.Y.

TIMES (Mar. 10, 2019), https://www.nytimes.com/2019/03/10/style/what-is-tik-tok.html

(“TikTok assertively answers anyone’s what should I watch with a flood . . . [the app] has stepped over the midpoint

between the familiar self-directed feed and an experience based first on algorithmic observation and inference. . . .

It’s an algorithmic feed based on videos you’ve interacted with, or even just watched. It never runs out of material. It

is not, unless you train it to be, full of people you know, or things you’ve explicitly told it you want to see. It’s full

of things that you seem to have demonstrated you want to watch, no matter what you actually say you want to

watch.”).

See, e.g., Daria J. Kuss et al., Problematic Mobile Phone Use and Addiction Across Generations: The Roles of

Psychopathological Symptoms and Smartphone Use, 3 J.

TECH. BEHAVIORAL SCI. 141 (2018); José De-Sola

Gutiérrez et al., Cell-Phone Addiction: A Review, 7 F

RONTIERS PSYCHIATRY 175 (2016) (noting “a consensus about

the existence of cell-phone addiction, but the delimitation and criteria used by various researchers vary”). The field

is still nascent and there is no consensus that internet addiction exists or is a distinct psychological condition. See

Francesca C. Ryding & Linda K. Kaye, “Internet Addiction”: A Conceptual Minefield, 16 I

NT. J. MENTAL HEALTH

& ADDICTION 225 (2018) (criticizing the cavalier use of the term “internet addiction” without a firmer grasp of the

problem); Kristyn Zajac et al., Treatments for Internet Gaming Disorder and Internet Addiction: A Systematic

Review, 31 P

SYCH. ADDICTIVE BEHAVIORS 979 (2017).

users online—for example, with intelligent and flexible algorithms.

The algorithm will learn

from, and respond to, these basic human preferences, thereby delivering—perhaps along with

good content—a large quantity of low-quality content.

2) Examples

There are many examples of exploitation of consumers who have sunk costs in a platform and

may not be fully informed. For example, app designers have enticed children into playing free

games that are built around in-app purchases, leading children to make large purchases without

parental knowledge or permission. The U.K.’s consumer protection agency, the Office of Fair

Trading, found that such purchases could be pricey: “A My Little Pony game, for example,

offer[ed] users a virtual ‘mountain of gems’ for a real-life £69.99.”

Regulators around the

world have cracked down to force companies to refund large payments that children make on

their parents’ phones.

More disturbing examples of low-quality content are YouTube

recommended videos that lead the viewer towards false or dangerous content.

Prior to having

these patterns made public and criticized, a Google search about the earth’s geology would lead

to a chain of recommendations that resulted in “flat earth” content;

YouTube would offer

Beginning in 2014, Twitter followed Facebook in centering around algorithm-selected content rather than simply

displaying tweets from accounts a user chooses to follow in reverse chronological order. See Will Oremus, Twitter’s

New Order, S

LATE (Mar. 5, 2017),

http://www.slate.com/articles/technology/cover_story/2017/03/twitter_s_timeline_algorithm_and_its_effect_on_us_

explained.html.

See Paul Lewis, ‘Fiction is Outperforming Reality’: How YouTube’s Algorithm Distorts Truth, GUARDIAN (Feb.

2, 2018), https://www.theguardian.com/technology/2018/feb/02/how-youtubes-algorithm-distorts-truth; Paul Lewis

& Erin McCormick, How an Ex-YouTube Insider Investigated its Secret Algorithm, G

UARDIAN (Feb. 2, 2018),

https://www.theguardian.com/technology/2018/feb/02/youtube-algorithm-election-clinton-trump-guillaume-chaslot;

see also Aim of the Project, A

LGOTRANSPARENCY,

https://algotransparency.org/methodology.html?candidat=Francois%20Fillon&file=ytrecos-presidentielle-2017-06-

10.

Hilary Osborne, OFT Warns Free Online Games Pressure Children Into In-app Purchases, GUARDIAN (Sept. 26,

2013), https://www.theguardian.com/money/2013/sep/26/free-online-games-children-apps.

See Press Release, U.K. Competition & Markets Authority, CMA Refers 3 Games to the ASA (June 4, 2015),

https://www.gov.uk/government/news/cma-refers-three-childrens-online-games-to-the-asa; Press Release, Fed.

Trade Comm’ n, Apple Inc. Will Provide Full Consumer Refunds of At Least $32.5 Million to Settle FTC

Complaint It Charged for Kids’ In-App Purchases Without Parental Consent (Jan. 15, 2015),

https://www.ftc.gov/news-events/press-releases/2014/01/apple-inc-will-provide-full-consumer-refunds-least-325-

million; Press Release, Euro. Union, In-app Purchases: Joint Action by the European Commission and Member

States is Leading to Better Protection for Consumers in Online Games (July 18, 2014), http://europa.eu/rapid/press-

release_IP-14-847_en.htm.

See Bergen, supra note 62.

See Casey Newton, How YouTube Helps Flat-Earthers Organize, VERGE (Feb. 20, 2019),

https://www.theverge.com/interface/2019/2/20/18232524/youtube-flat-earth-recommendation-algorithm-conspiracy;

Press Release, Google, Continuing Our Work to Improve Recommendations on YouTube (Jan. 25, 2019) (noting

that the company was planning to “begin reducing recommendations of borderline content and content that could

misinform users in harmful ways—such as videos promoting a phony miracle cure for a serious illness, claiming the

earth is flat, or making blatantly false claims about historic events like 9/11”); see also Kevin Roose, YouTube

Unleashed a Conspiracy Theory Boom. Can it Be Contained?, N.Y.

TIMES (Feb. 19, 2019),

https://www.nytimes.com/2019/02/19/technology/youtube-conspiracy-stars.html.

teenage girls interested in diets videos about how to get anorexia, and so forth.

It is important

to realize that this content is not chosen by human curators at the platform. Rather, the algorithm

learns what content people will click on, and what content will cause them to stay on the

platform longer, through many millions of small experiments; that is the content that is suggested

and viewed. Exploitation and addiction caused by the optimization of the platform is a harm to

consumers because they are likely watching lower-quality content than they would choose if they

were fully informed in advance about how the content is chosen, or perhaps if they had

alternative platforms to choose among.

iv. Privacy

Another worry is that the privacy of consumers on digital platforms is violated; this too is a

decline in the quality of the product. For example, Facebook recently announced that it will

merge the infrastructures of Facebook Messenger, WhatsApp, and Instagram. As the New York

Times noted, “[t]he integration plan raises privacy questions because of how users’ data may be

shared between services. WhatsApp historically required only a phone number when new users

signed up. By contrast, Facebook and Facebook Messenger ask users to provide their true

identities. Matching Facebook and Instagram users to their WhatsApp handles could harm those

who prefer to keep their use of each app separate.”

Germany’s competition regulator responded

to this announcement by prohibiting Facebook from combining data from different sources (such

as WhatsApp or Instagram) with data from Facebook.com without a user’s explicit and voluntary

consent.

Facebook’s eagerness to get third-party apps connected to its network has led to mass

data leaks, exposing sensitive information from hundreds of millions of people.

And Facebook

is hardly alone.

See Nicholas Tompson, When Tech Knows You Better Than You Know Yourself, WIRED (Oct. 4, 2018),

https://www.wired.com/story/artificial-intelligence-yuval-noah-harari-tristan-harris/ (reporting Tristan Harris, head

of the Center for Humane Technology, arguing that “[t]he problem is [YouTube] doesn't actually care about what

you want, it just cares about what will keep you next on the screen. The thing that works best at keeping a teenage

girl watching a dieting video on YouTube the longest is to say here's an anorexia video”).

Mike Isaac, Zuckerberg Plans to Integrate WhatsApp, Instagram and Facebook Messenger, N.Y. TIMES (Jan. 25,

2019), https://www.nytimes.com/2019/01/25/technology/facebook-instagram-whatsapp-messenger.html.

See Bundeskartellamt, supra note 37.

See, e.g., Data on 540 Million Facebook Users Exposed, BBC (Apr. 4, 2019),

https://www.bbc.com/news/technology-47812470; Millions of Facebook Passwords Exposed Internally, BBC (Mar.

21, 2019), https://www.bbc.com/news/technology-47653656.

See, e.g., Douglas MacMillan & Robert McMillan, Google Exposed User Data, Feared Repercussions of

Disclosing to Public, W

ALL ST. J. (Oct. 8, 2018), https://www.wsj.com/articles/google-exposed-user-data-feared-

repercussions-of-disclosing-to-public-1539017194; Chaim Gartenberg, Twitter Advising All 330 Million Users to

Change Passwords After Bug Exposed Them in Plain Text, V

ERGE (May 3, 2018),

https://www.theverge.com/2018/5/3/17316684/twitter-password-bug-security-flaw-exposed-change-now.

b. The Assessment Problem

i. Measuring Consumer Welfare in a Behavioral World

Each of these issues results in fundamental difficulties in applying standard antitrust analysis—

which is related to the assessment of the welfare effects of various practices or of a merger—to

digital markets. Adequate measures of volume, quality, and consumer surplus may be difficult to

obtain. The number of users choosing a certain option may not reflect their true preferences if the

platform can make that choice a default that is difficult to see or to change. The number of clicks

on ads may not correlate with greater welfare if higher volume of clicks is obtained by exploiting

lack of self-control and addictive behaviors.

Given the prevalence of behavioral effects in the digital economy, the measurement of consumer

welfare must be carried out very carefully. As we have mentioned, behavioral economics is now

a well-established discipline that can help sort different online behaviors and business practices.

Incorporating this knowledge into the legal practice’s toolbox may help develop better measures

of output and quality.

We caution, however, that the legal structure of US antitrust law is not

well set up to accommodate this complexity as it opens the door for judges to weigh all manner

of social concerns as well as traditional economic effects. We see two approaches that might be

more fruitful. First, the fastest route to more accurate measures of welfare might well be for a

digital regulator to limit the business models that serve harmful content, the way regulators limit

the harm from mortgages by restricting debt to income ratios, the harm from prescription drugs

by requiring access through a physician, and the harm from automobiles by requiring airbags and

crash tests, to name a few examples. If platforms had little or no incentive to deliver harmful

content to consumers, the standard consumer welfare toolkit would be more accurate in this

market. Second, harmful content is also, from the viewpoint of the consumer’s long run or ex

ante self, low-quality content. Analytical paradigms in antitrust analysis commonly deal with low

quality and quality-adjusted prices, and could incorporate the role of exploitative content in this

way.

Another reason to be pessimistic about measuring traditional surplus concepts is related to the

barter nature of the exchange: Users barter attention and personal data for services. With a “free”

service, consumers are paying for any expansion of activity with their attention to content. When

facing a zero-money price, and when quality is difficult to observe, consumers are not receiving

salient signals about the social value of their consumption because the price they believe they

face does not reflect the economics of the transaction, and they are ignorant of those numbers.

For a view of the role behavioral economics can play in legal practice, see THE OXFORD HANDBOOK OF

BEHAVIORAL ECONOMICS AND THE LAW (Eyal Zamir & Doron Teichman eds., 2014).

ii. Assessing the Social Welfare of Advertising

Behavioral economics also calls for careful use of standard economic surplus measures when

applied to the supply of advertising. Unlike supply of goods, higher value of advertising need not

imply higher social value. As discussed above, when advertising efficacy relies on psychological

nudges that bring people into consumption they would consciously avoid otherwise, more

advertising may well mean less consumer welfare.

Moreover, while price competition directly benefits consumers, competition through persuasive

advertising may waste resources.

To see this point, consider the following example: Two firms

share the market equally without advertising. If one then uses advertising and shifts the demand

toward that firm (perhaps by some psychological effect), it will then have a larger share of the

market. However, if both firms decide to advertise, they will continue to share the market

equally. If the advertising cost is not too large, both firms will advertise in order to preserve their

market share. Therefore, despite advertising spending, the equilibrium sales and consumption are

the same as without advertising. Here, allowing for advertising only induces wasteful spending,

which in our setting is captured by the platform as profit. This is an extreme example, but it

highlights that a higher volume of advertising—and greater platform profits—may not indicate

an increase in social welfare in the same way we normally think a higher output of, for example,

shoes, would indicate.

c. Harms to Investment and Innovation

i. Rents

Successful platform-style strategies pre-date the internet. For example, the key to Microsoft’s

business was the orchestration of ecosystem-wide innovation—the personal computer industry—

to benefit its own core offering as well as complements provided by third-party partners.

The

strategy we highlight in this section of the report is a successful platform’s choice of how much

rent to expropriate from these complementors. For example, Microsoft turned PC hardware—an

essential complement to an operating system—into a commodity business with the main

exception of the microprocessor made by Intel. However, PC applications software—again an

Although subtler, the same point has been shown to hold in the economic literature about informative advertising

that conveys useful information about products such as existence, location, and other product characteristics. See

Kyle Bagwell, The Economic Analysis of Advertising, in 3 H

ANDBOOK OF INDUSTRIAL ORGANIZATION

(Mark Armstrong & Robert Porter eds., 2007), https://doi.org/10.1016/S1573-448X(06)03028-7; Simon P.

Anderson & Régis Renault, Advertising Content, 96 A

MER. ECON. REV. 93 (2006). In this case advertising brings

real value to consumers. But competition to steer consumers from other suppliers and/or to preserve market share

may still lead to excessive spending on advertising.

See ANNABELLE GAWER & MICHAEL A. CUSUMANO, PLATFORM LEADERSHIP: HOW INTEL, MICROSOFT, AND

CISCO DRIVE INDUSTRY INNOVATION (2002) (detailing the characteristics of these types of platforms).

essential complement to the OS—remained (in part) an area where other firms could enter,

compete, and earn profit.

The level of market power attached to a successful platform is so high that it often gives the

platform owner the ability to expropriate almost the entire surplus available on its platform. And

the ability to add that surplus to its existing core profit is the incentive to do so. For example,

Microsoft’s actions to favor Word over WordPerfect expropriated surplus from an existing

software complement to its operating system. Importantly, in this example, WordPerfect was not

a potential entrant into, or substitute for, a PC operating system, so there was not an exclusionary

theme. The distinction between complementors that could disintermediate the platform and are

therefore potential rivals, and those that cannot is critical for the antitrust analysis below. When

the platform owner takes steps to disadvantage complementors who have no market power, the

complementors may resist by using non-market (political and legal) strategies. Such

complainants were part of the antitrust cases against Microsoft. We see this response to current

platforms by firms in Europe today.

Chamath Palihapitiya, a venture capitalist, has quoted Bill Gates as arguing that a platform exists

whenever “the economic value of everybody that uses it exceeds the value of the company that

creates it.”

In that world, complementors earn rents. By contrast, Ben Thompson, a tech

journalist, argues that Facebook and Google are what he calls aggregators—firms that

completely control the relationship between suppliers and users.

His point is that this control

allows the aggregator to exercise market power over one side of the platform, control access by

the other side, and extract all the rents. This emphasis on the creation of complementor rents and

their distribution or expropriation is a theme that runs through this Report.

When and how a platform appropriates the rents of its complementors varies and may engender

different responses from complementors. At root, the reason the complementor is there in the

first place is because the platform originally needed content to attract consumers, so it invited the

complementors onto the platform. Businesses selling widgets on ecommerce sites, games on

social media, and mapping apps on handsets, are all examples of complements that were critical

to successfully launching a platform. The complementors make these investments thinking they

will obtain a return, and that expected return leads to efficient levels of investment. If

investments were made knowing that the returns would be zero, there might well be dynamic

harm because the platform would not be able to get off the ground. (Or alternatively, the owner

of the platform would have to vertically integrate into many applications and provide them

Interview by Semile Shah with Chamath Palihapitiya, CEO, Social Capital (Sept. 16, 2015),

http://haystack.vc/2015/09/17/transcript-chamath-at-strictlyvcs-insider-series.

Ben Thompson, Tech’s Two Philosophies, STRATECHERY (May 9, 2018), https://stratechery.com/2018/techs-two-

philosophies (contrasting platforms, such as Microsoft and Apple, which “need 3rd parties to make them useful and

build their moat through the creation of ecosystems,” with aggregators, such as Facebook and Google, which “attract

end users by virtue of their inherent usefulness and, over time, leave suppliers no choice but to follow the

aggregators’ dictates if they wish to reach end users”).

itself.) A significant source of discontent today seems to come from complementors who

invested believing that they could capture surplus with a good product and who are later

expropriated by the platform. A second discontented group are the complementors that existed in

the brick and mortar world and now have no choice but to use a dominant platform, so they have

no bargaining power despite providing valuable products or services.

If the expropriation is accomplished using a tool that is anticompetitive, it may violate

competition laws. In the United States, this antitrust violation can be established when the

complementor may be able to disintermediate the platform and is therefore a potential rival, for

example, Netscape and Windows.

If such a move is not a violation of competition law but

violates the expectations of market participants, it may be viewed as unfair. On the other hand, if

the platform has significantly improved quality or engaged in innovation, market participants

may think the platform fairly earned those rents.

Digital platforms have a variety of rent-extraction strategies that seem to be loosely related to the

social scrutiny they are receiving.

For example, the news industry had no choice but to use

Facebook. But Facebook has been reluctant to share any of its profits with news companies—and

its market power has meant it hasn’t had to.

Interestingly, when content providers have market

power, the platform cannot always extract rents, even when it has a high market share. Take the

hypothetical case of a fragmented travel platform industry in a geography where the airline

market is concentrated. In that setting airlines could disintermediate a travel site by withholding

their participation in the site, or by favoring their own sales channels. The equilibrium division

of platform rents in this case will favor the powerful content providers, the airlines, rather than

the platforms.

Market participants observe that Facebook has slowly commoditized most

companies which supply attention to its users—from news and content markets to games and

apps, companies’ profits deriving from Facebook have slowly diminished as Facebook

appropriated most of the gains.

These trends may be part of the reason there is growing

attention to Facebook’s business model today.

The antitrust theories of harm that cover this case are discussed below.

See, e.g., Josh Dzieza, Prime and Punishment: Dirty Dealing in the $175 Billion Amazon Marketplace, VERGE

(Dec. 19, 2018), https://www.theverge.com/2018/12/19/18140799/amazon-marketplace-scams-seller-court-appeal-

reinstatement (discussing the fear Amazon sellers have of the firm and its ability to shut them down with little or no

explanation); Izabella Kaminska, Amazon (sub)Prime?, F

IN. TIMES (Mar. 20, 2019),

https://ftalphaville.ft.com/2019/03/20/1553085361000/Amazon--sub-Prime-.

See Ben Thompson, The Buzzfeed Lesson, STRATECHERY (Jan. 28, 2019), https://stratechery.com/2019/the-

buzzfeed-lesson (“[C]ontent suppliers are absolutely commoditized: Facebook doesn’t need to do anything to keep

them on the platform, because where else will they go? Might as well keep the money for itself.”).

For a related setting, see Fiona Scott Morton et al., Benefits of Preserving Consumers’ Ability to Compare Airline

Fares, (unpublished manuscript, 2015) (on file with authors).

See, e.g., Thompson, supra note 94; Pete Brown, More Than Half of Facebook Instant Articles Partners Have

Abandoned It, C

OLUM. J. REV. (Feb. 2, 2018), https://www.cjr.org/tow_center/are-facebook-instant-articles-worth-

it.php; Harrison Weber, Facebook Kicked Zynga to the Curb, Publishers Are Next, V

ENTUREBEAT (June 30, 2016),

https://venturebeat.com/2016/06/30/facebook-kicked-zynga-to-the-curb-publishers-are-next.

Today’s prominent tech companies most likely learned from older tech businesses that those

older businesses had “lost,” or shared, a too-high percentage of profits with their ecosystems.

While older businesses were continually confronted with questions about what to own versus

what to cede to partners on their platform, without today’s technological advancements

(including advanced data analytics and greater computational power), older platforms were not

as swift and flexible in capturing value as current digital platforms can be. The increased scale

and scope of control has provided modern digital platform owners with increased power over

their ecosystems. Today’s platforms understand that they can obtain higher margins if they either

make all of the necessary complements themselves or position themselves as a mandatory

bottleneck between partners and customers—leading to many platforms taking a significant

commission on sales on their platform or extracting value through barter of information. In

particular, today’s digital platforms are very careful to maintain complete control over the user

relationship so that they do not face any threat of disintermediation. These choices can be used to

reduce the possibility of successful entry by an innovator in the platform’s space. The next

section turns to this problem of entry.

d. Harm to Entry, Including Disintermediation

i. The Practical Consequence of Entry Barriers

When evaluating entry as the main source of competition against a platform, as discussed above,

the entrant has a significant disadvantage relative to the platform. Venture capitalist investors

will often evaluate a startup based on its ability to either access or build enough data swiftly

enough, all with the aim of reaching enough insights to take advantage of all the forces discussed

above. In other words, a new entrant starved of data, quantitatively and qualitatively speaking,

relative to a tech giant, is at a significant competitive disadvantage, and investors will be unlikely

to invest if they view that data deficit as insurmountable. Although it can be attractive for a VC

to invest in a firm that may be acquired by Facebook, Google, or Amazon, the road to a

successful acquisition is fraught with danger, most notably the ability of any of these three giants

to replicate a specific feature, functionality, or business model should other considerations such

as intellectual property rights, team quality, defensibility, or time to market not weigh against

such copycat strategy. Additionally, investors do not always fare well in these acquisitions.

ii. Incumbent Incentive to Leverage Entry Barriers

There is growing evidence that conglomerate digital platforms are in an advantaged position to

stop or block entry by more focused rivals when compared to traditional businesses. Companies

like Alphabet, Amazon, and Facebook operate in multiple business verticals (for example, mail,

See, e.g., Rachel Kraus, Amazon Bought Eero for $97 Million and Employees Still Got Screwed, MASHABLE (Apr.

5, 2019), https://mashable.com/article/amazon-eero-wifi-router-sale/#HN0veV33jPqc (“Ultimately, thanks to a ‘last

in, first out’ philosophy, Eero’s Series D investors, led by Qualcomm, will recoup 84 percent of their investments.

The seed round and Series A-C investors will all get back 31 cents on the dollar.”).

maps, and search), collecting different dimensions of data on a consumer (for example, identity,

location, and purchase intent) which give faster intelligence on competitive threats and new

chinks in the platform’s competitive armor. These companies can then derive superior insights

into what firms they should block, which they should buy, and how they should grow

strategically. This gives the platform an advantage over a rival entrant considering the same set

of opportunities, and increases their abilities to exclude such rivals. A rival platform with similar

economies of scope, data insights, and installed base may be a more formidable entrant.

If large digital platforms have both the incentive and ability to purchase and block entrants that

compete with them, or might compete in the future, the question is whether they have done so.

The evidence that platforms have bought a series of potential competitors in recent years is

anecdotal but fairly robust.

For example, many observers believe that Instagram and WhatsApp

might well have been serious competitors for Facebook.

The evidence that platforms have

blocked potential entrants is likewise anecdotal

100

and was used in the Android and AdSense

cases the European Commission has brought against Google.

101

More formal research in this

area is essential.

iii. Disintermediation and Foreclosure of Potential Platform

Entrants

In addition to de novo entry, platforms fear disintermediation by a partner. If a platform’s partner

is able to directly access and serve the platform’s customers, it can overtake the platform.

Disintermediation can also occur through commoditizing services so that one side (normally the

end consumer) is willing to substitute away from the platform—leading to a loss of profits.

The threat of disintermediation has important implications for the analysis of market entry and

foreclosure. Modern platforms have an incentive to regularly thwart companies that compete with

them for user demand. A platform that has total control of demand can steer customers to content

and complements it owns rather than to those provided by independent firms that might challenge

its market power. And because of potential harm to competition and complements, US regulators

have often been tasked with preventing discrimination, foreclosure, and similar strategies in

specific industries.

102

Without a great deal of insight into the technology and the strategy of the

platform, foreclosure will be difficult to observe by outsiders. Technological tools such as mobile

hardware technologies, advances in computing power, communications technologies, application

See U.K. COMPETITION REPORT, supra note 1, at 49, tbl.1A.

Tim Wu, The Case for Breaking Up Facebook and Instagram, WASH. POST (Sept. 28, 2018),

https://www.washingtonpost.com/outlook/2018/09/28/case-breaking-up-facebook-instagram.

100

American Tech Giants Are Making Life Tough for Startups, ECONOMIST (June 2, 2018),

https://www.economist.com/business/2018/06/02/american-tech-giants-are-making-life-tough-for-startups.

101

Press Release, Eur. Comm’n, Antitrust: Commission Fines Google €1.49 Billion for Abusive Practices in Online

Advertising (Mar. 20, 2019), http://europa.eu/rapid/press-release_IP-19-1770_en.htm.

102

See Lina Khan, The Separation of Platforms and Commerce, 119 COLUM. L. R. (forthcoming 2019),

https://papers.ssrn.com/id=3180174.

programming interfaces (API), cloud computing technologies, and data analytics enable a strategy

of keeping the attractive content from establishing a relationship with the user and thereby possibly

entering as a competitor. Platforms have bluntly moved to prevent disintermediation and have

engaged in foreclosure to block potential rivals. For example, Facebook acted to suppress the

growth video-capture-and-sharing app Vine when Vine attempted to link its users to their

Facebook friends.

103

Facebook CEO Mark Zuckerberg personally approved the decision to prevent

Vine users from finding friends on the app via Facebook.

104

Exclusive contracts and loyalty contracts can also be used to achieve exclusion. For example, a

long-term contract that requires an advertiser not to use an entrant can foreclose demand from

that entrant, leading to exit. An exclusive contract with a global reach can prevent an able niche

competitor from growing larger and obtaining economies of scale. Bundling of services by the

incumbent platform can be designed to exclude entry or foreclose existing rivals. Contracts

between platforms and advertisers that allow for individual negotiation can protect an incumbent

from losing individual targeted sales to an entrant without requiring the incumbent to lower its

prices across the board. An incumbent platform with market power will often have the incentive

and ability to undertake these strategies and thereby preserve its profit.

iv. Foreclosure of Complements to Capture Rents

One critical place to control the relationship is platform access. Amazon and Facebook regularly

make decisions over which app or vendor is able to sell or is denied access to their stores and

customers. Platforms often have a financial incentive to steer customers to particularly profitable

products and can use the power of defaults and ordering to accomplish that effectively. Vendors

operate in a risky environment where the platform’s whims can determine its future as much or

more than consumer satisfaction. If it chooses, a platform can steer demand elsewhere and the

vendor loses access to its customers—because these customers are another company’s users who

single-home and buy from the default choice at the top of the page. Provided the consumer

continues to find the totality of the platform experience positive, this bargaining power allows

the platform to dictate business terms. The Vestager Report describes the setting clearly:

Other platforms impose rules and institutions that reach beyond the pure matching

services and shape the functioning of the marketplace and, potentially, the relationship

between the various platform sides, e.g. by regulating access to and exclusion from the

platform, by regulating the way in which sellers can present their offers, the data and

APIs they can access, setting up grading systems, regulating access to information that is

generated on the platform, imposing minimum standards for delivery and return policies,

103

See Adi Robertson, Mark Zuckerberg Personally Approved Cutting Off Vine’s Friend-finding Feature, VERGE

(Dec. 18, 2018), https://www.theverge.com/2018/12/5/18127202/mark-zuckerberg-facebook-vine-friends-api-block-

parliament-documents.

104

See id. Several years after buying Vine, Twitter shut it down. See Casey Newton, Why Vine Died, VERGE (Oct. 28,

2016), https://www.theverge.com/2016/10/28/13456208/why-vine-died-twitter-shutdown.

providing for model contracts, imposing price controls and MFN clauses, etc. Such rule-

setting and “market design” determine the way in which competition takes place.

105

The way competition takes place determines the level of profit achieved by the platform and

each complement. In and of itself, a platform setting the terms of trade, quality levels, services,

and so on may not be problematic if the purpose of the change is to “grow the pie” in a way that

complementors view as fair (i.e., not involving expropriation). For example, Amazon, Facebook,

or Google know in real time which products are sold to whom, at what price, and which

packaging or incentives work, which may drive their rules. However, if these rules become

opaque and uncertain or the insights gleaned from an app or vendor are biased or used against it

in an asymmetrical manner, then the rule changes may not be about increasing everyone’s

revenue, but about moving a larger share to the platform.

Merchants or vendors can find themselves banned, demoted in search results, or required to bear

higher costs without the ability to move to a competing platform because either there is none or

because the customers single-home, will not depart the platform because of the loss of one

vendor, and cannot be reached elsewhere. The EC’s Android case describes the disadvantage

faced by independent apps that compete with the Google apps that are included in the mandatory

bundle.

106

Twitter has vertically integrated into video streaming by foreclosing the rival service

Meerkat.

107

Likewise, by selling logistics services to many of its sellers, Amazon gains an

advantage when it wishes to launch a store brand. It can analyze the data from its rivals to

develop an entry plan against those sellers. It is important to measure whether, and how much,

quality increases with these strategies.

It is not clear what profits such complementors expected or achieved from the platform

relationship, nor if there was significant relationship-specific investment required. Vendors may

be less likely to enter the market at all, or to innovate, if they know they must distribute through

a particular platform and their most successful products will be quickly copied. Businesses that

could grow on a platform and increase the platform’s attractiveness to consumers will be

unwilling to invest if their profits are not secure, and this may be a source of dynamic

inefficiency. However, platforms have an incentive to attract good complements in order to

105

EC COMPETITION REPORT, supra note 1, at 60.

106

The recent Spotify complaint against Apple in Europe is another example. See James Vincent, Spotify Files

Antitrust Complaint Over ‘Apple Tax,’ V

ERGE (Mar. 13, 2019),

https://www.theverge.com/2019/3/13/18263453/spotify-apple-app-store-antitrust-complaint-ec-30-percent-cut-

unfair.

107

See Feng Zhu, Friends or Foes? Examining Platform Owners’ Entry Into Complementors’ Spaces, 28 J. ECON. &

MGMT. STRAT. 23, 23 (2018) (“Meerkat, a mobile app that enabled Twitter users to broadcast live video streaming

to their followers, vanished after Twitter acquired its competitor Periscope and cut off Meerkat’s access to Twitter’s

social graph.”); see also Chris Welch, Tumblr ‘Truly Disappointed by Twitter’s Decision’ to Block Friend-Finding

On the Blogging Service, V

ERGE (Aug. 22, 2012), https://www.theverge.com/2012/8/22/3261270/tumblr-removes-

twitter-find-people-you-know. Twitter acquired Periscope, a competing startup, and then decided to shut out

Meerkat. On October 2016, Meerkat was shut down. Currently, Periscope is fully integrated with the Twitter

platform and claims 1.9 million daily users (last updated and released info as of 2017).

attract users. This incentive limits the platform’s desire to expropriate complementor rents under

some circumstances.

e. Harm to Innovation

i. Competition Promotes Innovation

There is significant theoretical and empirical research that concludes that anticompetitive

creation or maintenance of market power will cause a reduction in the pace of innovation.

108

This result is intuitive in the sense that firms “run faster” when they face competitors; competing

firms will try to offer a better product on any dimension consumers care about, including

innovation. Engaging in successful innovation is certainly both feasible and common for a large

platform with its enormous collection of data and other assets. These companies routinely spend

large sums on R&D, launch new products and services, and are more able than other competitors

to derive superior insights into how they should innovate based on the data collected from

aggregating demand and advances in machine learning and advanced data analytics. However,

the relevant counterfactual is whether the pace of innovation would be faster if platforms faced

more robust competition.

ii. Entry Barriers and Innovation

The lessening or blocking of innovative entry is of particular concern given its value. A VC will

usually be wary of outright investing in an innovative startup that will implicitly or explicitly

compete head-on with a tech giant. Given the tech incumbents’ ability to block or foreclose a

threatening entrant, the chance of successful entry is tiny. VCs would rather invest in businesses

that are creating new categories or solving common technical issues. Take Google’s search

engine as an example. To our knowledge there is only one search engine that has reached a

successful market size as a standalone business: Duck Duck Go. Google search is the dominant

firm in this category with high entry barriers; despite the enormous size of this market, VC

investors apparently do not want to fund an entrant. By contrast, VCs are attracted to startup

teams that solve a specific issue for Google (and have a chance of buyout) as opposed to funding

a team that wants to compete head to head. This dynamic leads to a self-fulfilling prophecy.

Reduced VC investment due to the inability to enter successfully in fact causes less entry; and

less investment also causes less differentiated innovation in the sector. This can be seen

increasingly by evidence of platform acquisitions.

109

Despite very high and stable profit margins,

markets like social media and search have faced little entry.

108

For a comprehensive survey of the innovation literature, see the literature summarized in Giulio Federico, Fiona

Scott Morton & Carl Shapiro, Antitrust and Innovation: Welcoming and Protecting Disruption (NBER Innovation

Policy and the Economy, Working Paper, 2019) (on file with authors).

109

The number of potential competitors purchased by the tech giants is large. For example, Amazon has purchased

Zappos, Fabric, CDNow, Quorus, Audible, Goodreads, and Quisdi; Facebook has acquired WhatsApp, Instagram,

As discussed above, large tech platforms can combine enormous financial resources with data

resources. Digital platforms have an ability to produce free cash flows at a speed and level that is

entirely new because of the combination of almost zero marginal cost, instant distribution, and

global reach. This advantage creates both the incentive and the ability for the digital platform to

outspend, to out-invest, or to acquire incumbents or new competitors. Incumbents have the

incentive and ability to stand in the way of possibly disruptive innovation. With deep pockets,

they can purchase possible future disruptors in order to align the path of innovation with their

strategies or otherwise control it. This story is widely believed to be the reason that Facebook

purchased Instagram and WhatsApp.

110

Alternatively, platforms may create “kill-zones” around themselves.

111

For example, Facebook

and Twitter’s aggressive API foreclosure, acquisition of competitors, and copying of new

services have boosted their market power.

112

Facebook even acquired a mobile phone monitoring

app, Onavo, which allows it to spot up-and-coming rivals and buy them or snuff them out.

113

While investment in innovation will continue, the type of innovation that will be funded will be

broadly determined by the incumbent and its strategies. Disruptive innovation in markets that are

characterized by high concentration levels and network effects is likely to be reduced compared

to a competitive market. One of the few sources of entry in digital platforms comes from rival

platforms that enter each other’s markets, as these large firms are more able to overcome entry

barriers of all kinds.

FriendFeed, and tbh; Google has bought Nest Labs, DoubleClick, YouTube, Waze, AdMob, Teracent,

BeatThatQuote.com, Admeld, and Tenor. See List of Mergers and Acquisitions by Amazon, W

IKIPEDIA (Apr. 12,

2019), https://en.wikipedia.org/wiki/List_of_mergers_and_acquisitions_by_Amazon; List of Mergers and

Acquisitions by Facebook, W

IKIPEDIA (Mar. 25, 2019),

https://en.wikipedia.org/wiki/List_of_mergers_and_acquisitions_by_Facebook; List of Mergers and Acquisitions by

Alphabet, W

IKIPEDIA (Apr. 12, 2019),

https://en.wikipedia.org/wiki/List_of_mergers_and_acquisitions_by_Alphabet. At times, these firms have barely

bothered to disguise their anticompetitive purpose. See Jacob Kastrenakes, Facebook is Shutting Down a Teen App it

Bought Eight Months Ago, Along with Two Others That Languished, V

ERGE (July 2, 2018),

https://www.theverge.com/2018/7/2/17528896/facebook-tbh-moves-hello-shut-down-low-usage.

110

Tim Wu, The Case for Breaking Up Facebook and Instagram, WASH. POST (Sept. 28, 2018),

https://www.washingtonpost.com/outlook/2018/09/28/case-breaking-up-facebook-instagram.

111

Google’s, Facebook’s, and Amazon’s annual conferences “held to announce new tools, features, and acquisitions,

always ‘send shock waves of fear through entrepreneurs,’” according to one investment firm, and “[v]enture

capitalists attend to see which of their companies are going to get killed next.” The Future of Tech Startups: Into the

Danger Zone, E

CONOMIST (U.K. edition) (June 2, 2018), at 61, http://weblogibc-co.com/wp-

content/uploads/2018/06/The_Economist_UK_Edition_-_June_02_2018.pdf.

112

See notes Error! Bookmark not defined.-Error! Bookmark not defined., supra; see, also, e.g., Erin Griffith,

Will Facebook Kill All Future Facebooks?, W

IRED (Oct. 25, 2017), https://www.wired.com/story/facebooks-

aggressive-moves-on-startups-threaten-innovation.

113

See Deepa Seetharaman & Betsy Morris, Facebook’s Onavo Gives Social-Media Firm Inside Peek at Rivals’

Users, W

ALL ST. J. (Aug. 13, 2017), https://www.wsj.com/articles/facebooks-onavo-gives-social-media-firm-inside-

peek-at-rivals-users-1502622003l.

iii. The Reward for Innovation

Entrepreneurs may expect a low payoff to developing a free-standing product because of entry

barriers and exclusionary conduct by the incumbent platform. In that case, its best hope is to be

the preferred innovator of a complement and sell its business to the platform at an early stage.

This source of financial reward is often cited as a reason why large platform acquisitions are

good for society. Certainly, a reward of this type (a share in the platform’s business model and

flow of rents) is better than no payoff and will stimulate some level of innovation. However, it is

important to see that this incentive is distorted relative to that of the competitive context.

Suppose public policy could reduce entry barriers and prevent anticompetitive exclusionary

conduct towards entrants. In that case, the entrepreneur would not have to settle for a small

fraction of the platform’s profits, but could compete for all of them and try to replace the

platform. When a young firm has a chance to compete for the entire market on the merits of its

innovation because there is no exclusionary conduct holding it back, success could lead to up to

100% of market profits, rather than a smaller acquisition payoff.

114

The availability of this option

would allow entrepreneurs to develop improvements, complements, or replacements for a

platform according to what has the highest returns. The inability to innovate in pursuit of the

whole market leads to lower entry in tech sectors that are already dominated by a single large

company.

iv. Current Trends in Innovation

The incipient but growing technical research supports a concern for the impact of big tech on

innovation. Much of this research uses startup formation and its fuel, venture capital, as a proxy

for innovation. In 2018, Facebook commissioned the consulting firm Oliver Wyman to write a

report on this topic.

115

Using Facebook, Google, and Amazon as a proxy for big tech platforms,

the authors drew four conclusions: i) Facebook, Google, and Amazon contribute a very small

portion of the total venture capital in tech; ii) Facebook, Google, and Amazon M&A activity has

no impact on aggregate investment levels; iii) Facebook, Google, and Amazon R&D has had no

impact on venture capital investment levels; and iv) the presence of Facebook, Google, and

Amazon does not dampen venture capital investment in technology relative to other mature

sectors.

The first three conclusions are factual and unsurprising. But the fourth conclusion, that the

presence of Facebook, Google, and Amazon does not dampen VC activity related to other

114

As previously noted, many of these markets will tend toward concentration naturally, so the size of profits is not

affected by effective competition enforcement, but the number of years they can be earned will be affected when an

incumbent cannot block entrants.

115

Assessing the Impact of Big Tech on Venture Investment, OLIVER WYMAN (July 11, 2018),

https://www.oliverwyman.com/content/dam/oliver-wyman/v2/publications/2018/july/assessing-impact.pdf.

sectors, is surprising. Ian Hathaway has summarized a straightforward rebuttal.

116

The Facebook-

commissioned report looked at industry-wide effects. However, as Hathaway points out, viewing

the industry in the aggregate masks the effects of dominating firms. By looking at the sub-

industries associated with each firm—social platforms (Facebook), internet software (Google),

and internet retail (Amazon)—a different trend emerges. Since 2009, change in startup investing

in these sub-industries has fared poorly compared to the rest of software for Google and

Facebook, the rest of retail for Amazon, and the rest of all VC for each of Google, Facebook, and

Amazon. This suggests the existence of so-called “kill-zones,” that is, areas where venture

capitalists are reluctant to enter due to small prospects of future profits.

117

In a study of the mobile app market, Wen Wen and Feng Zhu come to a similar conclusion: Big

tech platforms do dampen innovation at the margin.

118

Their study analyzed how Android app

developers adjust their innovation strategies in response to entry (or threat of entry) by Google:

[A]fter Google’s entry threat increases, affected developers reduce innovation and raise

the prices for the affected apps. Once Google enters, the developers reduce innovation

and increase prices further. However, app developers’ incentives to innovate are not

completely suppressed; rather, they shift innovation to unaffected and new apps. Given

many apps already offering similar features, Google’s entry may reduce social

inefficiency.

119

Ultimately, these studies are suggestive but not determinative. Much more research needs to be

done to property identify the existence and extent of “kill zones” for market entry and

innovation. Nonetheless, the evidence thus far does suggest that current digital platforms face

very little threat of entry and are negatively impacting investment in key digital areas. This is

reinforced by the fact that the key players in this industry remained the same over the last two

technology waves, staying dominant through the shift to mobile and the rise of AI.

120

In the past,

116

Ian Hathaway, Platform Giants and Venture-Backed Startups, IAN HATHAWAY BLOG (Oct. 12, 2018),

http://www.ianhathaway.org/blog/2018/10/12/platform-giants-and-venture-backed-startups.

117

Id.

118

Wen & Zhu, supra note 4.

119

Id.

120

This dynamic affects where we see innovation, and by whom. While distributed technology seems to be a

relatively open playing field with the potential for many entrants, the bulk of innovation in other spaces is dominated

by tech giants. This is the case, for example, in the AI and voice-assistant space. In the AI field, innovation is driven

by startups leveraging and customizing the existing technology and libraries developed by Microsoft, Amazon,

Facebook, and Google. These companies have been quick to acquire promising ventures. Since 2011, Facebook has

acquired seven AI startups: Face, Protogeo, Wit.ai, Bloomsbury AI, Dreambit, Grokstyle, and Ozlo. Since 2010,

Amazon has also acquired seven startups in the AI space: Yap, ORbeu, GoButler, harvest.ai, Graphiq, Body Labs,

and Dispatch. Finally, Google has acquired 20 AI startups since 2011, the most notable being Kraggle and

DeepMind. See List of Mergers and Acquisitions by Facebook, supra note 109; List of Mergers and Acquisitions by

Alphabet, supra note 109. The AI space may not see a successful IPO because of these dynamics. Similarly,

innovation in voice-assistant technology has been dominated by tech giants. Microsoft’s Cortana, Apple’s Siri,

Google’s Google Voice, and Amazon’s Alexa are examples of dominant firms leveraging their previous offerings to

retain a first-mover advantage in the market.

dominant business found it difficult to navigate innovation or disruption waves. By contrast,

Facebook, Google, Amazon, Apple, and even Microsoft were able to ride these waves without

significant impact on market share or profit margins. This indirect evidence corroborates the

argument that these companies are facing few competitive threats.

There is an informal sense from practitioners that in mobile, cloud computing, and AI the result

is the same: Innovation at the core has slowed and is centered around the tech giants.

121

As a

result, new innovation waves have emerged, centering around blockchain, distributed ledgers,

and decentralized models (the new Silicon Valley growth paradigm); and with specialized AI,

niche retail applications, or, in the enterprise sector, b2b models. It should be noted that, apart

from blockchain, whose decentralization eliminates network effects (in theory), the other areas of

innovation do not necessarily herald a change in future market structure. Some have also argued

that, in markets that are characterized by significant market power, innovation will likely shift

from focusing on consumer benefit (to entice a rise in market share), to consumer exploitation.

122

In summary, unlike last-century’s businesses, digital businesses will often lack competition in

the market. This absence may be redressed by intensive competition for the market, as firms

compete to be the winner that takes all. However, if dominant firms are able to maintain their

dominance, even as the external environment and tastes change over time, consumers may be

denied the benefits of competition for the market as well.

f. Old Wine in a New Bottle, or a New Reality?

The market characteristics and problems discussed above in Part I and Part II raise an important

question: Is there something new or different about the digital economy, which justifies a revised

approach to intervention? Is this simply old wine in a new bottle, the same old problems of

network effects and concentration, or do we face a new reality? As we outlined above, we

believe the digital platform may be a unique combination of economic forces that requires both

new analysis and new public policy.

We expect some to argue that there is no need for action. It is in the interest of the powerful

incumbent to deny there is any problem to solve in order to delay while the market remains in its

hands. The waiting game allows incumbents to collect profit from the status quo and use those

profits to raise doubts or marginalize concerns in order to delay or prevent effective

intervention.

123

121

See, e.g., Griffith, note 112, supra; Tech Monopolies Are Stifling U.S. Innovation. Antitrust Enforcement May Help,

RINK NEWS (Dec. 11, 2018), http://www.brinknews.com/is-antitrust-enforcement-the-wests-best-weapon-against-

chinese-tech; Andreeson Horowitz Makes Big Bet on Blockchain, R

EALDAILY (Feb. 13, 2018),

https://realdaily.com/andreessen-horowitz-makes-big-bet-on-blockchain.

122

See Ezrachi & Stucke, supra note 14.

123

See, e.g., Brody Mullins & Jack Nicas, Pay Professors: Inside Google’s Academic Influence Campaign, WALL

ST. J. (July 14, 2017), https://www.wsj.com/articles/paying-professors-inside-googles-academic-influence-

Of course, the rapid developments of our digital world are not yet fully understood and merit

ongoing study. However, the material above makes clear that when market power becomes

entrenched, it is difficult to return to a competitive market. Inaction could create decades of

market power leading to weakened innovation, increased rent extraction, and social domination

by gatekeeper tech firms. Making no policy change and standing back to let platforms carry out

the strategies of their choice carries unusual risk. The new reality of the digital landscape

presents new challenges that will not easily be resolved through organic competition or current

US antitrust enforcement norms. The United States now risks missing the moment to confront a

serious challenge to competition. That risk spurs our search for possible solutions, which we

outline in Part IV.

campaign-1499785286 (relating Google’s funding of professors for, among other things, research opposing antitrust

intervention in the search engine market).

IV. SOLUTIONS

In this part, we outline a range of solutions to address the issues raised above. Needless to say,

there is no magic pill. The legislature, judiciary, and regulators will each need to play a role.

Each of the proposed solutions brings with it benefits and costs and should be considered as part

of a balanced policy. Still, with these limitations in mind, these solutions have the power to

address changing market dynamics and enable society to move toward a more efficient, open,

and dynamic market environment.

Competition policy is only one facet of society’s wider goals. Alongside the promotion of

competition policy, policymakers may wish to implement policies to promote privacy,

democratic accountability, and fairness. While these goals may be advanced indirectly through

the measures proposed below, they can also be addressed more directly through laws and

regulations that we do not consider here. Other committees provide valuable insights on possible

advancements on that front.

Having established the impact of high entry barriers, we open with a fundamental question about

the ability of a market to self-correct.

a. Will Markets Self-Correct When Competition Problems Arise?

The question whether the market will self-correct is central to antitrust policy. It affects the scope

of illegality, as well as the zeal with which competition agencies approach intervention. It also

affects the gain from implementing both ex ante and ex post solutions to insufficient

competition. When there is a reasonable chance that technological progress and innovation will

speedily prevent or remove bottlenecks and maintain an active competitive process, intervention

should be limited. For example, the stock trading app Robinhood Markets, which slashed per-

trade fees to zero, has forced competitors to scramble—even JP Morgan is now offering a free

stock trading app for consumers.

124

Likewise, Amazon and Wal-Mart are currently vigorously

competing for fast and cheap delivery services, forcing others to follow.

125

Acolytes of the Chicago School have persuaded many federal officials and judges that markets

will generally quickly self-correct, and that antitrust scrutiny should therefore be limited. This

124

See Hugh Son, JP Morgan to Unveil New Investing App With an Eye-Catching, Disruptive price: Free, CNBC

(Aug. 21, 2018), https://www.cnbc.com/2018/08/21/jp-morgan-to-unveil-new-investing-app-with-an-eye-catching-

disruptive-price-free.html.

125

See Suman Bhattacharyya, To Compete with Amazon, Walmart and eBay are Spending on Logistics, DIGIDAY

(Jan. 15, 2019), https://digiday.com/retail/compete-amazon-walmart-ebay-investing-logistics;

Andrea Cheng, Amazon, Facing Growing Competition, Offers Free Shipping on All Orders for the First Time,

ORBES (Nov. 5, 2018), https://www.forbes.com/sites/andriacheng/2018/11/05/amazon-free-shipping-holiday-

minimum-order-walmart-target/#6c459dc555e1; Nick Statt, Amazon and Walmart’s Rivalry is Reshaping How We’ll

Buy Everything in the Future, V

ERGE (Aug. 3, 2018), https://www.theverge.com/2018/8/3/17630604/amazon-

walmart-competition-tech-partnerships-grocery-delivery.

conviction has extended into digital markets. Opponents of government intervention point to the

dynamic nature of technology, the role of disruptive innovation, and firms’ strong investment in

research and development. They defend the adequacy of a free-market, noninterventionist

approach, and argue that the cost of over-intervention outweighs the benefits of intervention.

While some markets may self-correct, the findings of this report suggest that rapid self-

correction in markets dominated by large digital platforms is unlikely, and that harms to

economic welfare from the exercise of market power in such markets are substantial. As

discussed above, entrants find it difficult to overcome the high barriers to take on digital platform

incumbents. Economies of scale, economies of scope, network effects, and negligible marginal

cost all work together to make entry difficult in existing markets. Moreover, while monopoly

profits are a lure to competitors, incumbents can use those very profits to entrench themselves

and protect their position. No matter how dynamic the technology, an entrant will not unseat a

monopolist if the monopolist is permitted to buy the dynamic entrant for a share of monopoly

profits. Both parties gain from such a transaction—and the public loses.

The result is less entry than a more competitive environment would create. Less entry into digital

markets means fewer choices for consumers, stunted development of alternative paths of

innovation, higher prices, and lower quality. Self-correction is not a realistic expectation in this

environment—indeed, the available evidence suggests it has not happened—and public policy

should not rely exclusively on it. Effective antitrust enforcement and regulation must take

account of this reality. If there is a force toward self-correction, it may require active promotion

to succeed, and in this way public intervention can be complementary rather than antagonistic to

market forces. Indeed, the other reports that have addressed this problem around the world have

accepted that policy changes are necessary in order to avoid stagnant and harmful digital

markets. We now turn to what policy options exist, and which of these the government might

adopt.

b. US Antitrust

i. Basic Principles

Antitrust law is intended to prohibit private conduct by firms that reduces economic welfare.

Although application of antitrust principles can be complex and the specific doctrinal

embodiments of those principles can seem arcane and arbitrary, at its core US antitrust law is

simple.

126

It can be summarized in a single sentence: Private conduct that creates or increases

market power, other than by efficiency-based competition on the merits, is illegal.

There are two fundamental components of any antitrust violation. The first is bad conduct,

commonly called “anticompetitive conduct.” The second is a resulting harm to economic welfare

126

See A. Douglas Melamed, Response: Antitrust Law is Not That Complicated, 130 HARV. L. REV. F. 163 (2019).

from the creation of more market power than would otherwise exist. A firm gains market power

when the competitive discipline imposed upon that firm by either actual or potential future rivals

is reduced. One way to achieve market power is through price-fixing and other forms of

collusion, but we will not address such behaviors here. Rather, this report will focus on antitrust

law as it applies to mergers and non-merger exclusionary conduct.

Importantly, a firm does not violate the antitrust laws if it gains market power by competing on

the merits rather than as a result of anticompetitive conduct. For example, if a firm enters the

market with a fantastic new product that has strong network effects, competes solely on the

merits of that product with no anticompetitive conduct, and finds itself with 95% market share, it

would not have violated the antitrust laws. Similarly, a firm does not violate the antitrust laws if

it engages in anticompetitive conduct but that conduct does not harm the market as a whole and

fails to result in additional market power.

127

Certain types of conduct, notably agreements among competitors fixing prices or allocating customers or service

areas, are regarded as unlawful per se, which means that they are unlawful without proof of harm to competition in

the market as a whole. Per se illegality in these circumstances is not a rejection of the principle described in this text.

Rather, per se illegality reflects a pragmatic judgment that those types of conduct are so likely to harm competition

and so unlikely to provide any welfare benefits that it would be needlessly costly and burdensome to require proof of

harm to the market as a whole on a case-by-case basis.

Box IV

More About US Antitrust Law

There are four substantive elements to any US antitrust violation: Private, as opposed to

government, conduct; anticompetitive conduct; creation or increase in market power; and a causal

connection between the conduct and the market power. There are three basic types of conduct that

can be anticompetitive. They are (i) mergers that lessen competition, (ii) conduct that excludes or

weakens actual or potential rivals, and (iii) conduct that constitutes or facilitates collusion (e.g.

cartels) among firms that would otherwise compete more vigorously. Possible antitrust problems

involving digital platforms are most likely to arise from the first two types of conduct.

Anticompetitive conduct is conduct that is likely to lead to the creation or maintenance of market

power for reasons other than an increase in allocative efficiency (i.e., by increasing output or

decreasing price where price is not less than marginal cost) or productive efficiency (i.e., by

reducing cost or increasing product quality). The range of potential anticompetitive conduct is

nearly boundless and can include burning down a rival’s factory, designing products to be

incompatible with rivals’ products, acquiring smaller rivals to shut them down or deny them to

other suitors, exclusive dealing and Most Favored Nation clauses (MFNs) in some circumstances,

and some forms of tying or bundling multiple products or services. Digital markets are prone to

new and innovative violations of the antitrust laws because marginal costs are often close to zero,

the business models themselves are often new and innovative, and firms are able to be creative

with new product bundles, contracts, and transactions.

More About US Antitrust Law (Cont.)

Current popular debate often seems to imply that high market share in and of itself is a violation

of the antitrust laws. It is important to understand that a firm can violate the antitrust laws only

if it engages in anticompetitive conduct, even if its conduct causes the firm to gain monopoly

power. Thus, for example, a firm does not violate the antitrust laws by gaining market power

solely because it has a better product that consumers choose to buy for that reason or because it

develops a better distribution channels that consumers find convenient.

Since at least the middle of the twentieth century, US antitrust law has embraced the view that

those who do not engage in anticompetitive conduct are entitled to the fruits of their labor,

including any market power or dominance that they might have gained. This view rests on the

concern that breaking up or restricting firms that obtain their success by competition on the

merits would deter the very kind of aggressive but procompetitive conduct that the antitrust

laws are intended to encourage. Such no-fault intervention would deter such conduct, both

because firms would fear that too much success would end up hurting them and because they

would be uncertain about the antitrust implications of their conduct. While firms that achieve

durable market power, especially in industries that are regarded as “natural monopolies,” are

sometimes subject to sectoral regulation, durable market power is not itself a sufficient basis for

antitrust intervention. Society has other legal mechanisms to regulate an industry that is not

delivering on social goals.

US antitrust law has also long required that antitrust plaintiffs prove, not just anticompetitive

conduct, but also that the conduct harmed competition in the market as a whole.

This

requirement limits antitrust enforcement to those matters that are important enough to warrant

costly and burdensome antitrust proceedings and ensures that antitrust law remains focused on

making markets work.

Both the anticompetitive requirement and the market power requirement protect the competitive

process, another concern that arises in the popular debate. The conduct requirement does this by

focusing antitrust enforcement on conduct that is not efficiency-based and can thus create

market power only by distorting competition on the merits. The market power requirement does

this by focusing antitrust enforcement on conduct that impairs the competitive discipline of

rivals and, in that way, harms the competitive process. Many of the current proposals for more

aggressive antitrust enforcement, although couched in antitrust language, would dispense with

either the bad conduct requirement or the market power requirement.

Conduct that harms competition without creating any efficiencies can readily be characterized

as anticompetitive. Examples of such conduct include mergers among competitors that do not

create efficiencies by combining complementary assets, conduct whose only purpose is to harm

or exclude competitors, and agreements among competitors about their prices or other terms of

trade.

1 See Brunswick Corp. v. Pueblo Bowl-O-Mat, Inc., 429 U.S. 477, 488 (1977) (“The antitrust laws . . . were

enacted for the protection of competition, not competitors.”) (internal citation and quote marks omitted).

While US antitrust law has long been flexible in combatting anti-competitive conduct, there is

increasing concern that it has been underenforced in recent years. First, there is increasing

evidence that the enforcement agencies and courts have permitted too many mergers between

competing firms that have led to post-merger price increases and other indications of increased

More About US Antitrust Law (Cont.)

When conduct both generates efficiencies and impairs the competitive efficacy of rivals, the

question whether it is anticompetitive is less clear and will depend on specific facts. The

conduct is more likely to be unlawful if substantially all of the efficiencies could be achieved

by an alternative that is less harmful to rivals or if the efficiencies are insubstantial and the

harm to rivals is great. Courts rarely deem conduct anticompetitive if it generates substantial

efficiencies that cannot be realized by an available alternative, regardless of the impact of that

conduct on rivals.

As explained in greater detail below, antitrust enforcement decisions are often made with

great uncertainty about the efficiencies and market impact of the conduct at issue. For the

past 30 years or so, antitrust law has reflected the view that mistakes in the direction of over-

enforcement are more problematic than those in the direction of under-enforcement. One

premise of this view is that under-enforcement results in market problems that are likely to be

corrected by innovation or other market developments. As noted above, this report assesses

the state of the evidence on entry and innovation and concludes that such self-correction is

unlikely to be prompt and effective, in part because digital platforms that have achieved

market power possess formidable barriers that inhibit entry and innovation. The assumption

that over-enforcement is costly is reflected in aspects of antitrust doctrine applicable to

exclusionary conduct, vertical contracts, and merger review as well as in courts holding

antitrust plaintiffs to demanding standards of proof.

The relative concerns about the risks and cost of over-enforcement and under-enforcement

embodied in current antitrust law has been subject to increasing criticism. This report

concludes that the tolerance of antitrust law for under-enforcement should be reassessed and

recalibrated. Careful recalibration of the benefits and costs of different mistakes is likely

essential to the development of more aggressive antitrust law that remains consistent with the

principles described above.

market power.

128

Vertical mergers are rarely challenged by the enforcement agencies,

129

and

claimed or expected merger-related efficiencies are often not realized. Second, monopsony—

market power of buyers or employers—appears to be a growing problem.

130

Recent studies

suggest that labor markets are less competitive than previously thought and employers have

exercised market power against workers in those markets.

131

Third, there have been few antitrust

challenges to exclusionary conduct since the government’s 1998 case against Microsoft, and

courts have in several instances been hostile to such cases and have imposed daunting proof

requirements on plaintiffs. Apparent under-enforcement is in part due to courts’ reliance on so-

called Chicago School assumptions that do not have a sound theoretical or empirical basis.

Regardless of whether or not antitrust enforcement has failed to keep up with conduct in the

economy more generally, the challenge of enforcing in the area of digital platforms presents new

issues. The platforms create new competitive environments; they provide opportunities for new

types of anticompetitive conduct; and they create new economic and conceptual challenges for

antitrust enforcement. This section is focused on analysis and recommendations designed to help

such future enforcement, though that enforcement will often be addressed to conduct that

occurred in the past.

The challenges facing future antitrust enforcement are more than just analytical and intellectual.

Antitrust law and its application by the courts over the past several decades have reflected the

128

See JOHN KWOKA, MERGERS, MERGER CONTROL, AND REMEDIES: A RETROSPECTIVE ANALYSIS OF U.S. POLICY

156, 159 (2014); Jonathan B. Baker, Presentation, Has the US Economy Become More Concentrated and Less

Competitive: A Review of the Data (Sept. 13, 2018),

https://www.ftc.gov/system/files/documents/public_events/1398386/georgetown-deck-913.pdf, at 39 (listing six

reasons to believe that industry concentration cannot be explained by scale economics, including the fact that

various forms of anticompetitive behavior are “insufficiently deterred” and that “market power is durable”); John E.

Kwoka, The Structural Presumption and the Safe Harbor in Merger Review: False Positives or Unwarranted

Concerns?, 81 A

NTITRUST L.J. 837 (2017); John Kwoka, The Changing Nature of Efficiencies in Mergers and in

Merger Analysis, 60(3) A

NTITRUST BULL. 231 (2015); Leemore Dafny et al., Paying a Premium on Your Premium?

Consolidation in the US Health Insurance Industry 102 A

MER. ECON. REV. 1161, 1163 (2012) (“[T]he mean

increase in local market [firm concentration] between 1998 and 2006 (inclusive) raised premiums by roughly 7

percent from their 1998 baseline”); Orley Ashenfelter & Daniel Hosken, The Effect of Mergers on Consumer Prices:

Evidence from Five Mergers on the Enforcement Margin, 53 J.L.

& ECON. 417 (2010). But see Michael Vita & F.

David Osinski, John Kwoka's Mergers, Merger Control, and Remedies: A Critical Review, 82 A

NTITRUST L.J. 361

(2018). See also Robert B. Kulick, Ready-to-Mix: Horizontal Mergers, Prices, and Productivity (U.S. Census

Bureau, Ctr. for Econ. Studies, Working Paper No. 17-38, 2017), https://papers.ssrn.com/abstract=2637961; Bruce

A. Blonigen & Justin R. Pierce, Evidence for the Effects of Mergers on Market Power and Efficiency (NBER,

Working Paper No. 22750, 2016), https://www.nber.org/papers/w22750, at 24.

129

See D. Bruce Hoffman, Acting Director, Bureau of Competition, Fed. Trade Comm’n, Vertical Merger

Enforcement at the FTC, Address at the Credit Suisse 2018 Washington Perspectives Conference (Jan. 10, 2018),

https://www.ftc.gov/system/files/documents/public_statements/1304213/hoffman_vertical_merger_speech_final.pdf

(admitting that “vertical merger enforcement is . . . a small part of [the FTC’s] merger workload”).

130

See, e.g., Michael Kades, In Conversation with Herbert Hovenkamp and Ioana Marinescu, EQUITABLE GROWTH

(Oct. 11, 2018), https://equitablegrowth.org/in-conversation-with-herbert-hovenkamp-and-ioana-marinescu.

131

Elena Preger & Matt Schmitt, Employer Consolidation and Wages: Evidence from Hospitals (Washington Center

for Equitable Growth Working Paper, 2019), https://equitablegrowth.org/working-papers/employer-consolidation-

and-wages-evidence-from-hospitals.

now outdated learning of an earlier era of economic thought, and they appear in some respects

inhospitable to new learning. Antitrust enforcement better suited to the challenges of the Digital

Age may therefore require new legislation.

Box V

EU Competition Law

EU competition law, like US antitrust law and the competition laws of most nations, is

intended in large part to promote economic welfare by prohibiting private conduct that

injures or is likely to injure competition. Not surprisingly in light of this shared objective, US

and EU law are very similar.

With respect to market power, similar to US antitrust law, a firm does not violate the EU

Competition laws if it gains power by competing on the merits rather than as a result of

anticompetitive conduct. EU Competition law only condemns the abuse of a dominant position.

An abuse, under EU law, may however include a wider range of prohibited exclusionary and

exploitative practices than under US antitrust law. As a result, some practices which may not

trigger enforcement action under US antitrust law, may nonetheless be regarded as infringing

EU competition law.

In this context, it is also worth noting the (wider) goals of EU competition law. According to

the European Commission, competition on the market is protected ‘as a means of enhancing

consumer welfare and of ensuring an efficient allocation of resources.’

This notwithstanding,

EU competition law has also consistently been held to protect ‘not only the interests of

competitors or of consumers, but also the structure of the market and, in so doing, competition

as such.’

Moreover, a genuinely indigenous objective is worthy of note, namely that of

promoting European market integration.

The multitude of competition goals, and their position within the wider normative EU values

may sometimes contribute to possible inconsistencies between the EU and US analysis.

EU Competition law is enforced both at public and private levels; public enforcement is carried

out by the European Commission and by the Competition Authorities of the member states.

Private enforcement takes place in the courts of the member states, where private parties may

bring follow-on or stand-alone damage claims.

1 European Commission, ‘Guidelines on the Application of Article 81(3) of the Treaty’ [2004] OJ C101/97,

para.13 (hereinafter ‘the General Guidelines’).

2 Case C-501/06 P GlaxoSmithKline Services Unlimited v Commission and Others [2009] ECR I-9291, para 63.

See also Case C-8/08 T-Mobile Netherlands and Others [2009] ECR I-4529, paras 31, 36, 38-39; Council

Regulation (EC) 1/2003 on the Implementation of the Rules on Competition Laid Down in Articles 81 and 82 of

the Treaty [2003] OJ L1/1, Recital 9; European Commission, ‘Green Paper on Vertical Restraints in EC

Competition Policy’ COM(96) 721 final, para 180.

3 Information Service High Authority of the European Community for Coal and Steel Luxembourg, ‘The Brussels

Report on the General Common Market’ (June 1956) (Spaak Report); David J Gerber, ‘The Transformation of

European Community Competition Law?’ [1994] Harvard Intl LJ 97, 102.

4 Ezrachi, Ariel, EU Competition Law Goals and the Digital Economy (June 6, 2018). Oxford Legal Studies

Research Paper No. 17/2018. Available at SSRN: https://ssrn.com/abstract=3191766

c. Special Challenges Presented by Technology Platforms

Technology platforms present particular challenges for antitrust enforcement. Markets tip and

resulting market power is durable, so even effective antitrust enforcement is unlikely to generate

fragmented markets. Nonetheless, enforcement that protects competition on the merits in the first

stage and prevents exclusionary conduct in the second stage will help ensure that market-

participants make unfettered choices among competing platforms and that entry and innovation

are not inhibited by private rent-seeking. Additionally, these markets move very quickly in areas

such as new product introduction, foreclosure, and tipping. Antitrust litigation does not move

quickly. Effective antitrust enforcement should move as rapidly as is practicable. However,

enforcers will be most effective when they choose enforcement priorities and remedies to

generate optimal deterrence of anticompetitive conduct.

As explained above, digital platforms are able to exploit behavioral biases to their advantage.

Economists describe these strategies as falling into two conceptually distinct categories of

conduct: exclusive (strategies that foreclose competition) and collusive (strategies that cause

higher prices).

132

The former occurs when platforms exploit behavioral biases to keep consumers

attached to their platforms and make switching to alternatives more difficult. These tactics

generally make consumers less receptive to competitive alternatives—they lower

contestability—and thus raise entry barriers. Platforms also exploit behavioral biases—such as

hyperbolic discounting and limited self-control—to extract surplus from both consumers and

content providers. Strategies such as offering addictive content at moments when consumers lack

self-control increase time spent on the platform and profitable ad sales even as the platform

lowers the quality of content. These tactics increase the welfare costs of market power.

Many technology platforms are distinctive because they provide valued services to consumers

without charging a monetary price. Instead, consumers barter their attention and data to the

platforms in exchange for these services. The platforms use that attention and data to generate

monetary payments from advertisers. While a barter transaction is, in principle, subject to

antitrust scrutiny just like any other transaction, antitrust enforcement has had vastly more

experience with transactions based on monetary prices, and that experience has prompted the

development of sophisticated tools to analyze money prices.

133

Where monetary prices are fixed

at zero while quality changes over time—in response to changes to the nature of the services,

privacy protections, content offerings and the like—the quality-adjusted prices change. Because

economists, antitrust agencies, and courts have less experience with quality, they lack equally

132

See Steven C. Salop, Practices that (Credibly) Facilitate Oligopoly Co-ordination, in NEW DEVELOPMENTS

IN THE ANALYSIS OF MARKET STRUCTURE 265 (Joseph E. Stiglitz & G. Frank Mathewson eds., 1986

133

For example, the US agencies’ traditional definition of an antitrust market is one in which a monopolist could

profitably raise price by a non-insignificant amount for a significant period of time. See U.S.

DEP’T OF JUST. & FED.

TRADE COMM’N, 2010 HORIZONTAL MERGER GUIDELINES (2010),

https://www.ftc.gov/sites/default/files/attachments/merger-review/100819hmg.pdf, at 9, § 4.1 (defining the

hypothetical monopolist test).

sophisticated tools for analyzing changes in quality-based prices. This is partially because quality

naturally presents itself in a less quantified form than price, but also because in regular markets

price often adjusts to quality, not the other way around, so economists have been able to rely on

price as one measure of quality. When enforcers are able to quantitatively link quality to price,

they will be better equipped to fit digital cases into older, price-centered jurisprudence.

Technology platforms also pose unusual challenges for antitrust merger enforcement. To the

extent that platforms are in winner-take-all or winner-take-most markets, mergers among

significant, existing competitors, which are the mergers most subject to antitrust challenge, are

likely to be rare. Instead, competition in such markets is largely for the future, often in evolving

and very different markets. This competition is sometimes called “competition for the market” or

“leapfrog” competition. In this context, acquisition by a dominant platform of a much smaller

and possibly nascent firm could be very damaging to competition if, absent the acquisition, the

smaller firm would develop into a major competitive threat or would lead to significant change

in the nature of the market. In a concentrated market structure, this potential competition from

very small entrants may be the most important source of competition faced by the incumbent

firm.

134

The problem is that it is very difficult to know at the time of an acquisition whether the acquired

firm is likely to develop into a competitor or whether, to the contrary, acquisition by the platform

offers the most promising path to the commercial development and use of the acquired firm’s

new technology or an essential exit strategy for investors in the acquired firm. Antitrust agencies

and antitrust law need to develop a better understanding of the circumstances under which

acquisitions of nascent competitors might be anticompetitive.

135

Enforcers may have to learn to

think more as venture capitalists do and understand their analytics in order to better identify

134

The D.C. Circuit in Microsoft endorsed this principle. See United States v. Microsoft Corp., 253 F.3d 34, 79

(D.C. Cir. 2001) (en banc) (per curiam) (“[I]t would be inimical to the purpose of the Sherman Act to allow

monopolists free reign to squash nascent, albeit unproven, competitors at will.”).

135

In addition, because small acquisitions are not subject to pre-merger review under the Hart-Scott-Rodino Act,

agencies are often unaware of the acquisitions until after the are consummated. The threshold for pre-merger

notification was raised in December 2000 from $10 million to $50 million. While transactions under the threshold

are still subject to Clayton Act enforcement, agencies do not see them in advance, may not find them at all, and can

only sue to unwind them after the fact. After the filing threshold increase, there was a sharp uptick of newly non-

notified mergers (between $10 and $50 million) between direct competitors—the type of mergers that likely would

have been blocked during HSR review, had it occurred. Small technology mergers fall in this category also and are

rarely reviewed or challenged. See Thomas Wollman, Stealth Consolidation: Evidence from an Amendment to the

Hart-Scott-Rodino Act, A

MER. ECON. REV. INSIGHTS (forthcoming 2019),

https://www.aeaweb.org/articles?id=10.1257/aeri.20180137. An example of a problematic merger of this type that

was challenged by DOJ after consummation is Bazaarvoice-PowerReviews. United States v. Bazaarvoice, 3:13-cv-

00133-WHO, 2014 WL 203966 (N.D. Cal., Jan. 8, 2014). While the government prevailed in that litigation, because

the two software firms had gone some way down the integration path, the result was arguably not as favorable for

consumers as if the transaction had been notified and blocked ex ante which would have preserved two robust

competitors. In general, the evidence demonstrates that firms act quickly to obtain market power in the absence of

government enforcement, and that the requirement to notify the government serves as an important deterrent to

anticompetitive mergers.

harms. A pattern of repeated purchases of small potential competitors that could have developed

into substitutes, or an acquisition price that reflects a sharing of monopoly rents, might be useful

indicators of possible competitive risks.

Antitrust law might also have difficulty policing conduct by established technology firms that

tends to exclude or marginalize smaller rivals. Recall that at launch a platform invites and

encourages complementors to provide content and functionality on its platform. It does this when

it is competing for consumers’ attention against other platforms because it wants its platform to

be attractive. A successful platform creates an ecosystem that is valuable to consumers.

However, one or both of the following issues may arise. In the first, a complement to the

platform seeks to become its horizontal competitor. In the second, the platform seeks to become

a horizontal competitor of one of its complements. Let us take these one at a time.

First, a complement can develop the ability to form a relationship with the end user that is

sufficiently free-standing and valuable to take the user off the platform and into a separate

relationship with the complement. The platform has an incentive to foreclose the complement to

prevent this loss of market power and profit. Because the complement is transitioning into direct

horizontal rivalry with the platform, US law does not have any trouble recognizing antitrust

violations of this form. For example, content providers like Yelp are Google’s complements—

people want to be able to find Yelp reviews via Google—but Yelp is also a search engine that

could grow to rival Google search. Google has entered the restaurant review market with its map

product, and Yelp now alleges that Google is engaging in foreclosure.

136

Second, if the platform observes that a complement (say, complement Z) is earning strong

profits, the platform may seek to enter that complementary market. Because the platform and

rival complement Z providers are now horizontal competitors, the platform has the incentive—

and, often, the ability—to foreclose those competitors. This could take the form of banning rival

complements from the platform, reducing their ability to interoperate, raising their costs, steering

customers elsewhere, and so forth. If the platform’s new product Z is a wonderful innovation,

there will be an efficiency (in the form of higher quality) to weigh against any harms from less

competition.

137

At root, there is a fundamental question about whether all the rents of the

platform are part of the competitive return to the creator of the platform, particularly if the

complementors’ added-value can be driven to zero once the network effects are operational. In a

setting with single-homing users, those complements may have no bargaining power ex post and

the platform may have the incentive and ability to give them a share of zero. This is a

136

See Duhigg, supra note 23.

137

Lina Khan, Amazon’s Antitrust Paradox, 126 Yale L.J. 710, 780-83 (2017). See also Feng Zhu & Qihong Liu,

Competing with Complementors: An Empirical Look at Amazon.com, 39 S

TRAT. MGMT. J. 2618, 2618 (2018)

(finding that “[w]hile Amazon's entry discourages affected third‐party sellers from subsequently pursuing growth on

the platform, it increases product demand and reduces shipping costs for consumers”).

particularly interesting problem when those complementors added value to the platform at a

critical competitive moment before network effects and entry barriers protected it.

Addressing this issue, the U.K. Furman Report and the EC Vestager Report both devote

significant discussion to platforms being “fair” to complementors. First, a sectoral regulator is

likely to be better than the antitrust laws at enforcing fairness norms. Antitrust law is focused,

not on fairness itself, but on anticompetitive conduct that creates market power and might

therefore permit various types of conduct that some might regard as “unfair” because, for

example, a platform extracts the rents of partners that offer a good product but have no

bargaining power. Because the complementor is not threatening to replace the platform, the

simple exclusion theory of harm explained above does not apply directly. The situation is

informally akin to the familiar “open early, closed late” strategy.

138

In the “open early, closed

late,” strategy a firm at first encourages others to become dependent on connecting to it and

relying on it, and later uses their dependence to shut out competitors or extract monopoly

rents.

139

This type of behavior might be regarded as exploitation or excessive pricing and thus

deemed to be an unlawful abuse of dominance under EU law. It would violate US law only if,

among other things, it enabled the platform to gain or preserve market power it otherwise would

not have in either the platform market or the market in which the excluded firm did business. In

the latter case under US antitrust law, this conduct might be captured under the “duty to deal”

framework.

Because large technology platforms have huge scale and benefit from network effects, they are

often able to engage in aggressive conduct targeted at rivals without violating existing antitrust

standards. The platforms might be able to copy rivals’ innovations or otherwise increase the

value of their services to consumers without pricing below cost, and they might be able to insist

that rivals using their platforms enter into agreements, such as agreements regarding access to

consumer data, that enhance the platform’s ability to compete. A platform might be able to

require exclusive contracts or loyalty-based contracts that cause single-homing by one side (e.g.

drivers or consumers). Such single-homing might cause the market to tip in the platform’s favor

and exclude a competitor. This kind of conduct often has efficiency benefits that make it difficult

to challenge the conduct under the antitrust laws. Certain aspects of antitrust law might be

adjusted, for example antitrust rules intended to prohibit dominant firms from engaging in

conduct that would exclude an equally efficient competitor could be revised to better protect

138

See Scott A. Sher & Bradley T. Tennis, Exploiting Others’ Investments in Open Standards, COMPETITION POL.

INT. (Sept. 15, 2016), https://www.competitionpolicyinternational.com/exploiting-others-investments-in-open-

standards/; Carl Shapiro, Exclusionary Conduct: Testimony Before the Antitrust Modernization Commission (Sept.

29, 2005), http://faculty.haas.berkeley.edu/Shapiro/amcexclusion.pdf, at 15 (“[I]n a network industry, a firm might

obtain a dominant position based in part on certain ‘open’ policies that induce reliance by complementary firms, and

then later exploit that position by offering less favorable interconnection terms or by refusing to interconnect with

them altogether.”).

139

For an example, see the discussion of Cisco-Arista at note 16, supra, and accompanying text.

smaller competitors. Nonetheless, it is unlikely that such adjustments would entirely eliminate

the competitive advantages inherent in large firms with substantial scale and scope economies.

Pinpointing the locus of competition and therefore the relevant market in which technology

platforms compete can also be challenging because the markets are multisided and are often ones

with which economists and lawyers have little experience. This complexity can make market

definition another hurdle to effective enforcement. For example, two platforms might compete in

general search, while also each offering social media and mapping functionalities, among other

services. Advertisers that buy ads on searches may be a common set of customers. While courts

and agencies have substantial experience analyzing advertising markets, for example, they are

less knowledgeable about markets for attention or barter transactions involving data made

available to providers as an unintended byproduct of using a digital platform. The problems are

compounded by the facts that technologies surrounding the products’ functions in digital markets

are continually changing and changes in quality-adjusted prices are difficult to observe.

Only one litigated US case has explicitly addressed these issues. That case culminated in a 2018

decision by the US Supreme Court.

140

Although the decision has been praised by some

conservative commentators,

141

it has been widely criticized by others.

142

The case itself involved

the credit card business and what the Court called a “transaction platform,” in which the platform

(American Express) facilitated simultaneous transactions between consumers and merchants.

143

It should not, therefore, be legal precedent applicable to other kinds of platforms like Google and

Facebook. But the case does suggest that the five-Justice majority on the Court is hostile to

antitrust enforcement (at least in vertical and exclusion cases), does not understand multi-sided

markets very well, and might be more influenced by ideological preconceptions than by evidence

in the case or fact-finding by district court judges.

144

140

Ohio v. American Express Co., 138 S. Ct. 2274 (2018).

141

See, e.g., FED. TRADE COMM’N, Transcript: FTC Hearing #3: Competition and Consumer Protection in the 21st

Century (Session 4) (Oct. 15, 2018),

https://www.ftc.gov/system/files/documents/public_events/1413712/ftc_hearings_session_3_transcript_day_1_10-

15-18.pdf, at 228-235 (Vinson & Elkins partner Darren Tucker praising the American Express decision).

142

See, e.g., Dennis W. Carlton, The Anticompetitive Effects of Vertical Most-Favored-Nation Restraints and the

Error of Amex, C

OLUM. BUS. L. REV. (forthcoming 2019),

https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3328628; Herbert Hovenkamp, Platforms and the Rule of

Reason: The American Express Case, C

OLUM. BUS. L. REV. (forthcoming 2019),

https://scholarship.law.upenn.edu/cgi/viewcontent.cgi?article=3060&context=faculty_scholarship; A. Douglas

Melamed & Nicolas Petit, The Misguided Assault on the Consumer Welfare Standard in the Age of Platform

Markets, 54 R

EV. INDUSTRIAL ORG. 1 (2019); Michael Katz & Jonathan Sallet, Multisided Platforms and Antitrust

Enforcement, 127 Y

ALE L J. 1724 (2018). See also FED. TRADE COMM’N, supra note 141, at 226 (Goldstein &

Russell partner Eric Citron describing the American Express decision as “economically illiterate”).

143

American Express, 138 S. Ct. at 2277.

144

Anthony Kennedy, one of the members of the five-justice majority in American Express, has since retired, but his

replacement, Brett Kavanaugh, is likely even more ideological on antitrust matters. See Stephen Calkins, How Might

a Justice Kavanaugh Impact Antitrust Jurisprudence?, P

ROMARKET BLOG (July 20, 2018),

https://promarket.org/might-justice-kavanaugh-impact-antitrust-jurisprudence.

The harm from lack of competition in digital markets will manifest itself in quality and

innovation, as well as from higher prices to advertisers. As detailed by Giulio Federico and

colleagues, the impact on consumer welfare of a decline in innovation due to lack of competition

is likely to be large, especially in the case of fast-moving technologies that affect many

consumers and related businesses.

145

Very often the uncertainty involved in evaluating harms to

innovation will be high, especially in contrast to the analysis of price forecasts. It is possible to

measure pipeline projects and current R&D to obtain a sense of competitive overlap or

trajectory, but the tools do not yet exist to accurately forecast the speed and direction of

innovation in the longer run. Likewise, obtaining quantitative evidence about the innovations or

products that would have been offered to consumers in the absence of the conduct is often not

possible.

Perhaps in part as a result of these challenges, US antitrust has not been active in policing

allegedly anticompetitive conduct by technology platforms. The government’s last

monopolization case involving issues raised by platforms was the Microsoft case in 1998. Today,

the European Commission and the European National Competition Authorities effectively act as

the global enforcers for allegedly exclusionary conduct that operates at a global level in

industries such as software, chips, and digital platforms.

146

d. Could a Reformed Antitrust Law be Effective in Dealing with

Digital Platforms?

There are many ways that US antitrust law could be revised to make it more aggressive in

addressing competition problems while maintaining the objective of prohibiting private conduct

that reduces economic welfare. Such changes could improve competition enforcement in digital

markets as well as others.

i. Reform by What Means?

With few exceptions, antitrust law has in the past evolved in a common-law-like process by

which it has reflected new learning and judicial and market experience. This process is

continuing, at least to some extent, as antitrust law and enforcement have recognized, for

example, previously unnoticed competition problems in labor markets and doctrine has evolved

to incorporate new learning about competitive problems that can be created by most favored

145

Giulio Federico, Fiona Scott Morton, & Carl Shapiro, Antitrust and Innovation: Welcoming and Protecting

Disruption (NBER Innovation Policy and the Economy, Working Paper, 2019) (on file with authors).

146

See, e.g., Press Release, Euro. Comm’n, Antitrust: Commission Fines Google €4.34 Billion for Illegal Practices

Regarding Android Mobile Devices to Strengthen Dominance of Google’s Search Engine (July 18, 2018),

http://europa.eu/rapid/press-release_IP-18-4581_en.htm; Press Release, Euro. Comm’n, Antitrust: Commission

Closes Formal Proceedings Against Qualcomm (Nov. 24, 2009), http://europa.eu/rapid/press-release_MEMO-09-

516_en.htm; Report on the Monitoring Exercise Carried Out in the Online Hotel Booking Sector by EU Competition

Authorities in 2016, E

URO. COMM’N (2017), http://ec.europa.eu/competition/ecn/hotel_monitoring_report_en.pdf.

nation (MFN) and other vertical agreements. The challenges posed by the big technology

platforms and the current populist political climate have, however, put the issue of antitrust

reform before Congress in various legislative proposals. There are advantages and disadvantages

to both common law evolution and new legislation.

Evolution by a common law-like process takes time. It took the Chicago School roughly 20 years

to refocus antitrust law, from the early expressions of its perspective in 1950s and 1960s

147

until

the Supreme Court’s seminal decision in GTE Sylvania in 1977

148

—and the obstacles are

probably greater now. While there had been ebbs and flows in antitrust enforcement in the mid-

twentieth century, there was not a well-formed conceptual framework calling for aggressive

enforcement that the Chicago School proponents had to overcome. The structure-conduct-

performance paradigm was widely accepted as an economic proposition, but it was not

embedded in a rigorous normative and conceptual antitrust framework. By contrast,

oversimplified Chicago School thinking has provided a widely accepted framework for antitrust

analysis for more than thirty years. Perhaps more importantly, many federal judges, appointed by

an increasingly ideological vetting process, are trained in and adherents of that framework. Many

seem unaware of new economic research that calls into question many of the tenets of that

framework and continue to cite outdated Chicago School publications of the 1970s and 1980s.

And, while there has been a great deal of economic research and literature on which a new

antitrust paradigm could be constructed, there is not a widely accepted, alternative paradigm that

is comprehensible to and administrable by lawyers and judges. Even if such a paradigm were

written tomorrow and rapidly became widely accepted, it would likely take years for that

paradigm to be manifest in doctrinal changes and market outcomes.

New legislation could in principle be adopted and take effect much more rapidly. New

legislation would not need to depart from the dual requirements of bad conduct and harm to

competition in the market as a whole. Such legislation might, instead, implement a recalibration

of the relative tolerance of antitrust law for the risk of over-enforcement and under-enforcement

by prescribing rebuttable presumptions that would ease the high proof requirements currently

imposed on antitrust plaintiffs and place on defendants a more rigorous burden of proving

efficiencies. Some possible new presumptions and similar reforms are outlined below.

The risk, of course, is that new legislation will not be enacted by experts committed to sound,

economically-focused antitrust. It will be designed by Congress in a politically charged

environment subject to pressure from the very companies who stand to lose their market power if

subject to increased antitrust oversight, or who benefit if their trading partners are subjected to

excessive oversight.

147

See, e.g., Ward S. Bowman Jr., Tying Arrangements and the Leverage Problem, 67 YALE L.J. (1957); Robert H.

Bork, Legislative Intent and the Policy of the Sherman Act, 9 J.L.

& ECON. 7 (1966); Robert H. Bork & Ward S.

Bowman, Jr., The Crisis in Antitrust, F

ORTUNE, Dec. 1963, at 138, reprinted in 65 COLUM. L. REV. 363 (1965).

148

Continental T.V., Inc. v. GTE Sylvania Inc., 433 U.S. 36 (1977).

There is more at stake than the risk of flawed legislation. Antitrust law has maintained legitimacy

and widespread support for nearly 130 years in part because it applies to all forms of commercial

activity and is not perceived as special interest legislation. In our view it is very important that

antitrust law not have different rules aimed at different sectors—such as technology

149

agriculture

150

—that would differentiate industries and undermine political support for antitrust

law in general. For this reason, the report outlines a number of useful digital platform

interventions that can be undertaken by a sectoral regulator rather than falling to the task of

antitrust enforcement.

Equally important, antitrust law has benefited immensely from the brevity and vagueness of the

key statutory provisions because they have enabled antitrust law to evolve in response to new

learning. The challenge with new legislation is to embrace enduring normative principles without

codifying current economic learning in a way that will prevent the law from evolving to take

account of newer economic findings.

ii. Reform of Antitrust Law Through a General Tightening

As noted, most antitrust cases require uncertain decisions about unknowable future events like

innovation or entry or about unobservable economic elements like demand curves, marginal cost,

and product quality. Much US antitrust law reflects judgments about how to deal with such

uncertainty. Those judgments are embodied in rules regarding burdens of proof, evidentiary

presumptions, and decision-theoretic approaches to fact finding. They are also embodied in

substantive legal doctrine, such as rules regarding predatory pricing and unilateral refusals to

deal that are intended to reduce the likelihood of false positives (erroneously finding a business

has violated the law) even at the risk of false negatives (erroneously deciding that a business has

not violated the law). Digital markets typically have high levels of uncertainty and move quickly.

Given uncertainty, courts must determine how much weight to put on the risk of enforcement

mistakes: both the likelihood of a mistake and its cost. Much US antitrust law is driven by a

judgment, embraced by the Chicago School, that avoiding false positives (good conduct judged

to be bad) is more beneficial to society than avoiding false negatives (anticompetitive conduct

judged to be good).

151

This judgment rests on the beliefs that false positives are difficult to

correct but that false negatives will be quickly corrected by market forces. These beliefs seemed

149

As suggested, by, e.g., Elizabeth Warren, Here’s How We Can Break Up Big Tech, MEDIUM (Mar. 8, 2019),

https://medium.com/@teamwarren/heres-how-we-can-break-up-big-tech-9ad9e0da324c.

150

As suggested by, e.g., Elizabeth Warren, Leveling the Playing Field for America’s Family Farers, MEDIUM (Mar.

27, 2019), https://medium.com/@teamwarren/leveling-the-playing-field-for-americas-family-farmers-

823d1994f067.

151

For a classic statement of this proposition, see Frank H. Easterbrook, Vertical Arrangements and the Rule of

Reason, 53 A

NTITRUST L.J. 134, 135-36 (1984) (“There are limits on the ability of courts to sort the beneficial from

the deleterious manifestations of [restrictive dealing] practices, and most of the time it is better not to try than to try

and fail.”). For a criticism of this view, see Jonathan B. Baker, Taking the Error Out of ‘Error Cost’ Analysis:

What’s Wrong with Antitrust’s Right, 80 A

NTITRUST L.J. 1 (2015).

plausible in 1975 in a Chicago School framework, but they have never been empirically

demonstrated and have fallen into disrepute. Moreover, their logic is incomplete because the cost

to society of a false negative, for example, depends not only on its likelihood, but also on both

the magnitude and the duration of the resulting harm.

It is time for antitrust law to recalibrate the balance it strikes between the risks of false positives

and false negatives. Underenforcement is likely to be costlier than previously thought because,

among other things, market power of large technology platforms is more enduring. False

negatives are almost certainly more common than previously thought because certain types of

conduct that were previously thought to be benign are now understood to be anticompetitive.

Especially in technology markets, the most important competitive threats to incumbent firms are

likely to come from new entrants that might be vulnerable to exclusionary conduct or

anticompetitive acquisitions when their competitive prospects are uncertain. In addition, false

positives might be less common than previously thought because of the development in the past

few decades of more sophisticated and reliable econometric and simulation tools for assessing

conduct and market power effects. New work by economists studying multi-sided markets,

network effects, the economics of nominally “free” goods and services, and restrictive vertical

agreements suggests that antitrust agencies and courts will continue to improve in their ability to

make sound enforcement decisions. And false positives might be less costly than previously

thought because firms are finding new and different ways to realize efficiencies.

A recalibration of this type should influence antitrust law in two basic ways. First, it should

provide a basis for revising certain aspects of antitrust doctrine that were adopted explicitly in

order to minimize the risk of over-enforcement. Second, it could more broadly provide a basis

for courts to impose less demanding proof requirements on antitrust plaintiffs, especially where

facts are difficult to observe or prove directly and indirect or circumstantial proof is available.

Again, we pursue our theme of harm to entry through the next section. Exclusion of existing or

potential entrants is well established in both the economics literature and the antitrust

jurisprudence as a harm to competition, but the law is not well calibrated to recognize this

familiar tactic in its new setting.

Some of the specific ways in which the law might be revised are set forth below. They could be

achieved by common law-like evolution of antitrust law or by new legislation.

iii. Specific Areas of Possible Antitrust Reform [designed for the

specialist reader]

To address the issues raised by technology platforms, antitrust enforcement agencies and courts

will need to understand the unusual factual context that those platforms often present and be

receptive to recent and future economic learning about the implications of that context and how

to apply antitrust principles to it. Economists and other experts will need to develop new

understanding and new tools to aid agencies and courts in addressing these matters, among

others:

• How to assess the quality-adjusted price paid for a good or service sold in a barter

transaction with zero or close to zero monetary price, and how to define and analyze

markets in which a substantial portion of the sale take the form of barter transactions.

• How learning from behavioral economics and related disciplines about addictive or

exploitative content should be considered in addressing issues regarding economic

welfare.

• How technology platforms are able to take advantage of consumer biases (such as

salience, status quo bias, or impatience) to bind consumers to their platforms and make

switching to alternatives more difficult than imagined by lay intuition (“competition is

one click away”).

152

• How market circumstances affect the likelihood and nature of innovation and how to

evaluate innovation, whether using qualitative or quantitative tools, in both its magnitude

and direction.

• How to assess potential competition from new or small firms or not-yet-identified future

innovators and entrants. This is especially important in markets that depend on

technological change and in which competition in the market is less important than leap-

frog competition for the market.

• How to assess consumer welfare in a two-sided market. Learnings from platform

economics suggests that users on different sides of a platform generally have divergent

interests and thus that defining a single two‐sided market—rather than two, closely

interrelated ones—obscures the analysis. Platform economics also demonstrates that

neither the change in the two sided price nor the change in the transaction volume is a

sufficient statistic for a how a firm’s conduct affects consumer welfare.

153

152

David Wismer, Google’s Larry Page: ‘Competition is One Click Away,’ (and Other Quotes of the Week),

ORBES (Oct. 14, 2012), https://www.forbes.com/sites/davidwismer/2012/10/14/googles-larry-page-competition-is-

one-click-away-and-other-quotes-of-the-week.

153

One of the features of such platforms is that, because of the feedback effects between the two sides of a platform

and the importance of the relative prices charged on the two sides, increased output by the platform does not

necessarily imply increased economic welfare. Whether increased output by the platform will increase economic

welfare depends on how the benefits and costs of that output are allocated across the two sides. See Michael L. Katz,

Platform Economics and Antitrust Enforcement: A Little Knowledge is a Dangerous Thing, 28 J.

ECON. & MGMT.

STRAT. 138 (2019); Jean-Charles Rochet & Jean Tirole, Platform Competition in Two-Sided Markets, 1 J. EURO.

ECON. ASSOC. 990 (2003). There is another sense in which increased output of digital platforms does not necessarily

increase economic welfare. Because of some of the unique attributes of digital platforms discussed above, consumer

demand for digital services does not necessarily reflect consumer welfare from those services. See Section II.1,

Other antitrust reforms, including those below, would require changes to doctrine:

• Antitrust law prohibits unilateral refusals to deal only under very unusual circumstances.

Current law reflects, among other things, concerns about the difficulty of determining the

required terms of trade and incentive effects of required dealing on both the dominant

firm and its rivals.

154

The law thus gives platforms substantial freedom both to refuse to

deal with actual or potential rivals, including complements, and to deal with them only on

onerous terms. Those terms might include access to customer data, interoperability, and

other terms that raise the costs of rivals and enable the platforms to reinforce their

dominant positions. This doctrine should be reconsidered in light of the substantial

importance of large technology platforms and, in particular, their central role as

distribution channels, both of which suggest that the benefits of antitrust intervention

might be greater than previously appreciated.

• Predatory pricing law has been shaped in large part to avoid over-enforcement and with

explicit acknowledgement that the law permits some forms of anticompetitive pricing

conduct.

155

Courts have adopted a narrow and rigid notion of recoupment

156

and have

made it almost impossible to prove that prices are below cost, even where it seems

likely.

157

Digital goods often have a marginal cost close to zero, which makes tests that

require prices to be below incremental or variable cost almost impossible for a plaintiff to

meet. The law has also been construed to protect only rivals that are equally efficient at

the time of the conduct at issue and thus to disadvantage smaller rivals that have not yet

reached efficient scale. Predatory pricing law should be modified so that it will be better

able to combat anticompetitive pricing by digital platforms and other firms.

• The paradigm of predatory pricing law has also been relied upon to assess more complex

pricing strategies, such as loyalty discounts. Loyalty discounts and similar contracts can

be used to drive one side of a platform to single-home, which can cause a market to tip

and enhance market power. The efficiency benefits of loyalty discounts are very different

from those of low prices in general; for example, unlike low prices themselves, even

above-cost loyalty discounts do not necessarily increase static welfare. Antitrust law

supra. This latter concern applies to other products as well, such as mortgages and prescription drugs; and it is very

relevant to possible regulation of digital platforms, as discussed below. By contrast, while antitrust enforcers and

courts need to understand those attributes of digital platforms in order to understand how the relevant markets work,

antitrust law is for several reasons based on the assumption that consumer preferences reflect consumer welfare.

154

See, e.g., Verizon Commc’ns Inc. v. Law Offices of Curtis V. Trinko, 540 U.S. 398 (2004); see also A. Douglas

Melamed, Exclusionary Conduct Under the Antitrust Laws: Balancing, Sacrifice and Refusals to Deal, 20

ERKELEY TECH. J. 1247 (2006).

155

See Barry Wright Corp. v. ITT Grinnell Corp., 724 F.2d 227, 231, 234 (1st Cir. 1983) (Breyer, J.).

156

See C. Scott Hemphill & Philip J. Weiser, Beyond Brooke Group: Bringing Reality to the Law of Predatory

Pricing, 127 Y

ALE L.J. 2048 (2018).

157

See, e.g,, United States v. AMR Corp., 335 F.3d 1109 (10th Cir. 2003).

should not rely exclusively upon predatory pricing standards to assess loyalty

discounts.

158

• The Supreme Court held in American Express that a plaintiff in a case involving a

vertical restraint must define and prove a relevant market and may not rely on direct

proof of harm to competition.

159

This holding was based on the notion that vertical

restraints almost always enhance efficiency and almost never harm competition. Scholars

over the past 30 years have demonstrated that that notion is false and therefore, that

vertical restraints must be evaluated individually on the specific facts.

160

Where there is

direct evidence of harm to competition, antitrust law should not require circumstantial

evidence via a defined relevant market.

• Courts are generally very reluctant to second guess a defendant’s product design

decisions, even where the design harms competitors.

161

Yet product design decisions

involving, among other things, app stores, mobile device screen layouts, data storage and

analysis, and interface design are often key elements in digital market competition.

Antitrust courts should be more willing to assess product design decisions where

appropriate.

• There should be no safe harbor based on the short-term of exclusive dealing agreements

and other restrictive vertical agreements when they are used by dominant platforms or

firms. Such firms can use their market power to induce desired behavior in trading

partners without relying on long-term contracts.

162

Perhaps most importantly, antitrust law might be revised to relax the proof requirements imposed

upon antitrust plaintiffs in appropriate cases or to reverse burdens of proof. Burdens of proof

might be switched by adopting rules that will presume anticompetitive harm on the basis of

preliminary showings by antitrust plaintiffs and shift a burden of exculpation to the defendant or

by ensuring that plaintiffs are not required to prove matters to which the defendants have greater

knowledge and better access to relevant information. These proof requirements include the

following, which are likely to be important in the application of antitrust standards to technology

platforms:

158

See Hemphill & Weiser, supra note 156.

159

Ohio v. American Express Co., 138 S. Ct. 2274 (2018).

160

See Steven C. Salop, Revising the Vertical Merger Guidelines: Presentation at the FTC Hearings on Competition

and Consumer Protection in the 21st Century (Nov. 1, 2018),

https://www.ftc.gov/system/files/documents/public_events/1415284/ftc_hearings_5_georgetown_slides.pdf; Steven

C. Salop, Invigorating Vertical Merger Enforcement, 127 Y

ALE L.J. 1742 (2018).

161

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

162

Compare United States v. Dentsply Int’l, Inc., 399 F.3d 181 (3d Cir. 2005), with Omega Envtl., Inc. v. Gilbarco,

Inc., 127 F.3d 1157 (9th Cir. 1997).

• Mergers between dominant firms and substantial competitors or uniquely likely future

competitors should be presumed to be unlawful, subject to rebuttal by defendants. This

presumption would be valuable, not because it would identify anticompetitive mergers

with precision, but because it would shift the burden to the party with the best access to

relevant information on issues of competitive effects and efficiencies from the merger.

• Courts should not presume efficiencies from vertical transactions. Crediting of

efficiencies should require strong supporting evidence showing merger-specificity and

verifiability.

163

• Courts should be more willing to permit plaintiffs to prove harm to competition by

circumstantial evidence, especially where the propositions in question are not observable

and there thus cannot be direct evidence.

164

iv. A Competition Court

Revisions to the law may have little effect to the extent that judges see antitrust cases only rarely

and have difficulty understanding the economic underpinning of antitrust law. One way to

ameliorate this problem would be to establish a specialized antitrust court on which a certain

number of Article III judges would sit for a specified term of several years. These judges could

be expected to hear multiple antitrust cases and to develop substantial antitrust expertise. And,

because they would be chosen from the general federal bench, they would bring the broader

perspective of generalist judges. This model could be used at the trial court level, the appellate

level, or both. If judges are selected from the larger pool of Article III judges and rotated on and

off the specialized court, judges on the specialized court will be less likely to develop an overly

narrow intellectual interpretation of antitrust law.

v. A Regulatory Partner Could Enhance Effective Antitrust

Enforcement

Digital markets move quickly. Using the internet as a distribution channel to reach an installed

base of billions around the globe means that new products can be rolled out quickly—as can

anticompetitive conduct. The pace of antitrust enforcement is far slower—judging by the

evidence, too slow—to protect small entrants except by creating deterrence. A sectoral regulator,

by contrast, could be endowed with the authority to move quickly. By taking steps to preserve

competition before markets have tipped or entrants have been purchased, a regulator could prove

a valuable complement to antitrust enforcement.

163

For an example of the Court presuming efficiencies from vertical transactions with minimal evidence, see

American Express, 138 S. Ct. at 2289.

164

See, e.g., id. (injury to competition in a two-sided market); United States v. AMR Corp., 335 F.3d 1109 (10th Cir.

2003) (predatory pricing).

Antitrust enforcement proceeds on a case-by-case basis, and, apart from merger enforcement, it

is largely backward looking—it looks at conduct already undertaken that is alleged to harm

competition. A regulator, by contrast, can look forward and establish rules to constrain future

conduct before there has been harm to competition, or before investments by the defendant and

third parties have been shaped by the conduct. This is particularly useful in cases where ex post

conduct remedies would be costly or ineffective.

Effective antitrust enforcement requires effective remedies. Treble damages and financial

penalties can compensate for past harms and deter future bad conduct, but they do not restore

competition to markets in which competition has been harmed. Even an injunction to forbear

from the same or similar anticompetitive conduct going forward will not restore the lost

competition if entry barriers are high. For example, if the market has tipped and network

externalities are very strong, the firm that became a monopolist through violations of the antitrust

laws could stop the conduct at issue and yet retain its monopoly position and the associated

stream of profits. An antitrust authority that wants to restore lost competition must induce entry

and/or impose structural remedies. Such remedies are likely to require detailed and often

technical monitoring and years of effort. Antitrust enforcers are not suited to that type of

oversight whereas a sectoral regulator is. One possible solution is to permit antitrust authorities

and courts to design antitrust remedies and rely on an expert sectoral regulator to oversee their

implementation. A similar approach was approved by the Supreme Court in the Otter Tail

case.

165

e. Regulation

For the reasons above, we believe the establishment of a sectoral regulator should be seriously

considered. Given the tasks detailed by the other committees in this project as well as this one,

there would be much for a regulator to do that would improve the impact of digital platforms on

society. A digital regulator, should Congress choose to create one, could consider regulations and

actions along the following lines. We call this potential regulator the Digital Authority (DA).

The general harm identified above is insufficient entry, and therefore insufficient competition, in

digital platforms. Many of the regulations below are designed to lower barriers to entry directly.

Others are responsive to the difficulty of effectively prosecuting antitrust cases today in digital

markets. Regulations that mimic the antitrust laws but lower the burden of proof for the regulator

and allow it to move faster are a way to gain effective enforcement in this sector, if not others.

Regulation offers a valuable addition to antitrust enforcement. It can help design the digital

landscape and align the interests and incentives of platforms and key providers with those of

consumers and society. When carefully designed, a regulatory regime can limit or even preempt

the harmful effects detailed above, while minimizing its impact on the dynamic nature of digital

165

Otter Tail Power Co. v. United States, 410 U.S. 366 (1973).

markets. It can offer clarity and legal certainty as to the boundaries of acceptable competition.

Moreover, some of the problems discussed above may have only one structural solution: breakup

of the platform. An enforcer might not want to choose that option because it is very disruptive.

But less disruptive ex post remedies require ongoing monitoring, which antitrust enforcers are

not well-positioned to do. Handing that job off to a regulator might better serve consumers.

Therefore, the committee suggests considering regulatory steps in conjunction with improved

antitrust enforcement. It is important that regulation support, and indeed enhance, competition.

The report focuses here on regulations targeted at improving competition, not other policy

challenges related to digital platforms. And it is critical that any new regulatory framework avoid

the well-documented pitfalls that regulation often invites (e.g., agency capture, revolving door

syndrome, or incumbent protection). By having a pro-competition mandate, our hope is that the

DA will be able to use the tools at its disposal in ways that resist the natural impediments to

entry identified in Part I, rather than entrench powerful incumbents.

This combination of enforcement tools applied to one industry is nothing new in the US

economy. Virtually every sector of our economy has required both antitrust and specific

regulatory oversight, from banking to agriculture to communications, in order to promote

competition and other public interest goals. The regulations we propose below are focused on,

and limited to, those that will enhance competition.

The communications sector may offer the best guidance for how to approach public

accountability for digital platforms. Telephone, cable, and wireless communications networks

have many of the same attributes as digital platforms. Large capital expenditures to build

networks, declining costs as consumers use the networks and buy more services on the networks,

and the difficulties for new players to enter these markets and compete have led to a world of one

or few players in each market. And just like social networking, online search, and web-delivered

content, communications networks have been the lifeblood of how we communicate and practice

our democracy by delivering TV, radio, text, and conversation, making competitive outcomes in

the sector critical. The FCC has served as the sector-specific regulator for telecommunications.

In complex industries, a sector-specific regulator can have a wider remit than an antitrust

authority as is detailed in the box below.

Box VI

The FCC Model

In the Telecommunications Act of 1996, Congress sought to affirmatively promote, not just

protect, competition.

Congress’ vision included service availability that was universal and

affordable to all and a commitment to local and diverse ownership of news sources to support

a robust marketplace of ideas.

It wanted to prevent undue discrimination, limit ownership to

prevent excess market power (both horizontally and vertically in some instances), and police

against abusive contract provisions that distort fair market practices.

Some of these goals

were achieved with more success than others. And in general, the regulatory process

inherently restricts behavior and therefore efficiencies. The net benefits of regulation should

factor in these lost efficiencies.

The phone number portability rule was a clearly pro-competitive regulation. The FCC

published the Wireless Local Number Portability rule in 2003.

The rule allows cell phone

users to keep their phone number when they switch between wireless carriers. Prior to the

rule, a consumer who wanted to take advantage of a low price or better quality with a

competing provider would have had to change her phone number. This significant switching

cost dampened competition between carriers. The number portability rule made switching

easier and thereby strengthened competition between carriers.

Similarly, when portability of

1-800 numbers (e.g., 1-800-flowers) was introduced, prices of those services fell.

After Congress found the cable companies to be local monopolies in 1992, Congress

identified a dearth in cable competition partially caused by vertically-integrated cable

operators’ refusal to sell their programming to potential competitors. It therefore temporarily

prohibited exclusive programming contracts, banned a variety of abusive contracting

practices, and required that vertically-integrated firms sell their content to competitors under

reasonable prices, terms, and conditions.

1 H.R. Rep. No. 104-458 (1996) (Conf. Rep.). Even before the 1996 Act, the FCC had already begun the work

of promoting competition through the broad authority it had been granted in earlier statutes.

2 47 U.S.C. § 151 (2018).

3 Id.; 47 U.S.C. § 160 (2018).

4 See generally Wireless Local Number Portability (WLNP), F

ED. TRADE COMM’N,

https://www.fcc.gov/general/wireless-local-number-portability-wlnp.

5 Press Release, Federal Communications Commission, FCC Observes First Anniversary of Wireless Local

Number Portability (Nov. 24, 2004), https://www.fcc.gov/document/fcc-observes-first-anniversary-wireless-

local-number-portability (“Wireless local number portability (LNP) eliminated a barrier to full competition in

mobile telephone services and between landline and wireless services”).

6 V. Brian Viard, Do Switching Costs Make Markets More or Less Competitive? The Case of 800-Number

Portability (Cheung Kong Graduate School of Business Research Paper No. 1773R2, 2004),

https://ssrn.com/abstract=371921

7 47 U.S.C. § 251 (2018).

The FCC Model (Cont.)

The resulting increased competition in the cable industry enabled the Direct Broadcasting

Satellite industry to grow,

opened the door for telephone companies and smaller cable

providers to compete in the provision of distribution,

and ultimately drove the digital video

market to develop today’s cable-modem-powered broadband internet services.

The 1996 Telecommunications Act included interconnection requirements between

competing carriers to expand competition. The Act outlined a regulatory regime of duties to

connect, of parity in quality between connections offered to the incumbent’s own affiliates

and competitors, and of rates and contract terms that were just, reasonable, and

nondiscriminatory.

The regulation was designed to protect all of the businesses that needed

to connect to the long-distance wire for their business but competed with a vertically-

integrated local wire. Not only was raising rivals’ costs prohibited, but complete foreclosure

was also prohibited. This duty to deal in a non-discriminatory way is an example of a policy

that arguably promoted entry of cable, local telephone, and long-distance competitors.

Similarly, the FCC developed Customer Proprietary Network Information (CPNI) rules,

designed to protect sensitive business data transmitted through public telecommunications

networks. The FCC recognized that a dominant platform, such as the telephone company, on

which businesses rely in order to reach their customers, could easily gather and take

advantage of sensitive business user data to promote the phone company’s business and harm

competition.

Phone service competitors, as well as data-driven businesses (such as home-

security monitoring firms, hotels, and airlines), rely upon these rules to grow their businesses

without interference from telecom network owners. The rule was so effective at promoting

competition that Congress codified it in the 1996 Telecommunications Act. The harvesting of

data flowing through a platform—generated by and belonging to others—is a standard

business practice among today’s digital platforms.

Of course, these past regulatory efforts have also helped us learn what not to do. Regulations

could be and have been used to entrench incumbent firms’ market power, erecting regulatory

barriers to entry for new or innovative competitors.

1 47 U.S.C. § 548(a) (2018).

2 47 U.S.C. § 521(6) (2018); 47 U.S.C. § 548 (2018).

3 47 U.S.C. § 251 (2018).

4 47 U.S.C. § 222 (2018).

i. The Digital Authority

To be effective, a proposed regulatory regime requires an enforcement body capable of carefully

designing and enforcing the relevant regulations. We start therefore with a proposal for Congress

to pass legislation creating a Digital Authority with the mandate to develop targeted regulation to

achieve the goals described above and subsequently engage in monitoring and enforcement.

We anticipate that this regulator will also be tasked with non-competition digital goals, such as

those in the areas of privacy, media, data-use restrictions, and consumer protection. While the

antitrust agencies will employ structural interventions to protect competitive markets wherever

possible, the focus of this regulator will be on both carrying out remedies for the antitrust

authority that require ongoing oversight, and on developing regulations going forward that are a

combination of structural safeguards, such as unbundling or separation, with limited behavioral

interventions in areas where traditional antitrust tools are insufficient. Other jurisdictions that are

The FCC Model (Cont.)

Regulatory capture is a common problem.

For example, AT&T has historically had a

symbiotic relationship with the US government.

Perhaps the height of this relationship was

the explicit Kingsbury Commitment in 1913, which allowed AT&T, rather than the

government, to set the “solution” for managing competition.

The current Chairman of the

FCC, Ajit Pai, has called this “a cautionary tale about the dangers of regulatory capture.”

was likely this close relationship that allowed the AT&T monopoly to persist for so long,

eventually requiring an antitrust case to address the problem.

1 See George J. Stigler, The Theory of Economic Regulation, 2 BELL J. ECON. REG. 3 (1971) (arguing that, “as a

rule, regulation is acquired by the industry and is designed and operated primarily for its benefit.”); see also

Sheldon Whitehouse, Legal Responses to Regulatory Capture, ACS

BLOG (Nov. 9, 2018),

https://www.acslaw.org/acsblog/blog-post-for-the-american-constitution-society-legal-responses-to-regulatory-

capture (“We unfortunately live in a time of widespread regulatory capture. We should not pretend otherwise.”);

Elizabeth Warren, Corporate Capture of the Rulemaking Process, REG. REV

. (June 14, 2016),

https://www.theregreview.org/2016/06/14/warren-corporate-capture-of-the-rulemaking-process (“Under the law,

it is easy for business groups to challenge a rule for being too strong or too restrictive. But it is much harder for

public interest groups or ordinary citizens to challenge a rule for being too weak or riddled with loopholes.”);

Daniel Carpenter, Challenges in Measuring Regulatory Capture, REG. REV.

(June 22, 2016),

https://www.theregreview.org/2016/06/22/carpenter-challenges-measuring-regulatory-capture (

“[T]here are

limitations when it comes to preventing capture. Perhaps the most important is that capture is difficult to

measure”).

2 Chris Matthews, AT&T and the Government Have Been ‘Friends’ for a Really Long Time, FORTUNE (Aug. 18,

2015), http://fortune.com/2015/08/18/att-nsa.

3 Ajit Pai, Chairman, Fed. Commc’n Comm’n, Remarks on the 100

Anniversary of the Kingsburgy Commitment

(Dec. 19, 2013), https://www.fcc.gov/document/pai-remarks-100th-anniversary-kingsbury-commitment.

4 Id.

assessing competition in digital platforms all propose some form of regulation.

166

Having

forward-looking regulations in place will increase business certainty about what conduct is

permitted and how enforcement actions are likely to proceed. Ideally, this predictability and

clarity will encourage companies to comply with the law, thus requiring fewer government

resources for enforcement.

The DA legislation will require Congress to define the scope of regulatory power. The definition

must include digital businesses that facilitate transactions of any kind (including the sale of

advertising). It should have clear and broad authority over digital business models in order to

prevent firms subject to regulation from evading its oversight.

We offer a menu of potential regulation, starting with the least intrusive form and building to the

more severe interventions. Each type of regulation comes with costs and benefits which we

briefly detail. The size of the costs and benefits will be different across settings, meaning that the

choice of the best tool for any particular case may vary. The committee also suggests separating

out some types of regulation that will apply to virtually all market participants while others are

only appropriate tools to apply to companies with bottleneck power.

1) Bottleneck power

“Bottleneck power” describes a situation where consumers primarily single-home and rely upon

a single service provider (a “bottleneck”), which makes obtaining access to those consumers for

the relevant activity by other service providers prohibitively costly. As the U.K. Furman Report

put it,

[O]ne, or in some cases two firms in certain digital markets have a high degree of control

and influence over the relationship between buyers and sellers, or over access by

advertisers to potential buyers. As these markets are frequently important routes to

market, or gateways for other firms, such bottlenecks are then able to act as a gatekeeper

between businesses and their prospective customers.

167

The finding of bottleneck power will employ consideration of the forces that, as discussed above

in this Report, tend to impede entry and lead to foreclosure. The Furman Report similarly

explains that this single-homing foreclosure tends to happen when users experience high

switching costs, such as loss of valued personal data or reputational indicators at the point of

switching; contract terms that deter switching; technical barriers to switching, such as complex

switching processes or a lack of interoperability between the old service and the new or second

service; tying services, which can be by contract or technical; and the inertia of defaults.

168

166

See, e.g., AUSTRALIAN COMPETITION REPORT, supra note 1, at 13-14; EC COMPETITION REPORT, supra note 1, at

8-10; U.K. C

OMPETITION REPORT, supra note 1, at 60-61.

167

U.K. COMPETITION REPORT, supra note 1, at 41.

168

Id. at 36.

Digital businesses that have this incentive and ability to develop and preserve a single-homing

environment should be considered entities with bottleneck power. The DA, pursuant to

congressional guidance, should have the sole authority to define bottleneck power and should

update the definition either regularly or on an “as needed” basis. This is not a determination that

should be left for definition by generalist judges. If Congress establishes a specialized panel of

judges who review antitrust matters (as described above), it would be wise to grant the same

panel sole judicial review authority over DA regulatory actions.

2) Data

Should Congress pass any laws concerning consumer data, the authority could set forth pro-

competitive rules concerning the consequences of consumers’ control over their data and about

user choice in the sharing of data. Where users are simply and clearly informed and given the

opportunity to make viable choices about which companies get their data, this valuable

competitive information can be shared as the consumer prefers. The DA could design data

sharing rules with the general goal of reducing single-homing and promoting entry.

3) Partnership with the Antitrust Agencies

Antitrust enforcement agencies or courts could designate the DA as the administrator and/or

architect of remedies in antitrust cases. In cases where structural remedies are not appropriate to

restore the lost competition, the DA could be directed to either carry out a remedy designed by

the competition authority, or provide an effective behavioral monitoring remedy consistent with

the DA’s mandate. Since the difficulty of designing an appropriate and administrable antitrust

remedy that can be enforced by a court often limits antitrust enforcement, having the option of an

expert regulator to craft the remedy, or to simply be responsible for ongoing monitoring or other

execution of the remedy, could better ensure cost-effective enforcement of pro-competition

policies.

The DA can partner with the antitrust agency when it observes anticompetitive conduct. Due to

their frequent interaction with firms in the industry and their real-time data feed, regulators may

observe exclusionary conduct before the antitrust authority, and be able to give it relevant

information and data.

ii. Menu of Regulations

We lay out below a menu of regulations that could be used to solve the problems identified

above from least interventionist to most interventionist. For each, we describe the likely benefits

and the costs.

1) Broadly Applicable Regulations

Some of these regulatory tools should be applied broadly to all firms in the industry, not only to

bottleneck firms. It would be appropriate, however, to include a small business exception and

perhaps even a new business exception, to allow very small entrants, who may benefit

competition, time to ramp up against larger established companies.

a. Data collection

The Digital Authority could regularly collect data on market transactions, with an emphasis on

data from businesses with bottleneck power. The information would allow policy makers and

researchers to assess the performance of the sector and improve rule enforcement. For example,

the data may include, inter alia, a sample of searches at a set of websites, a sample of queries

followed by purchases at other sites, a sample of downloads of applications at an app store, or a

sample of activity and ads shown on a social media site. The DA may further regularly collect

information, including what types of data the business collects, how it uses that data, and who

else is bartered or sold access to which elements of the data. It may include experiments to

evaluate the true portability of data, the ads generated by certain user information, and other

information relevant to assessing how and if competition is taking hold in the market. This

program will require ongoing costs to run, but these should be reasonable compared to the

benefits.

Box VII

Real Time Regulation in Financial Services

Technological innovation and the data revolution are also disrupting the financial services

market, spurring regulators into action. Banks and other companies are undergoing an

innovation and digitization spree, partially in response to competition by non-financial

institutions and startups. This rapid change increases the challenges for regulators to follow

and analyze what regulated companies are doing.

The response by some financial regulators has been to embrace new technologies focused on

machine readable regulation (with the help of machine learning and AI) and digital reporting

(with the help of data sharing frameworks, APIs, cloud computing, and advanced analytics).

This is leading to an important shift from manual and template driven compliance and

regulatory reporting to a near real time or real time regulatory reporting and supervision.

Examples of authorities adopting this system range from the European Central Bank, various

national central banks in continental Europe, the Monetary Authority of Singapore, or more

notably, the Bank of England and the Financial Conduct Authority (FCA) in the UK.

These data may be requested by an antitrust agency to aid in an investigation or prosecution.

This ability to transfer existing data will speed up the enforcement of antitrust laws. After an

investigation is opened, rather than subpoenaing the relevant data from the targeted firms and

spending months arguing over definitions and formats, the antitrust authority could immediately

begin analysis with the DA dataset.

In addition, the DA should make as much of these data public as possible, subject to keeping

personally identifiable information and business secrets confidential. This will allow academics

and nonprofits to study particular markets and consumer behavior as they do in industries such as

airlines (where “Data Bank 1A” is publicly available)

169

to great public benefit. Leveraging the

nonprofit and educational sector to help the DA and elected officials understand these markets is

likely to be helpful and cost-effective.

b. Restrictions on Practices That Enhance Behavioral “Mistakes”

As described above, it is well known that behavioral “nudges” can lead consumers to make better

choices. Better choices promote competition because they generate increased market share for

firms that make better offers to consumers. The DA should have a mandate to create such “light

touch” rules when they will make markets more competitive. For example, automatic renewals

can discourage consumers from comparison shopping when a contract ends. A restriction on

automatic renewals in that setting could lower prices market-wide. Regulations that require firms

to make salient the most important terms of an offer can improve competition (e.g., a credit card

offer must show the APR in large font). The U.K. likely has the most advanced regulatory

169

See Bureau Transportation Stat., Sources of Aviation Data, U.S. DEP’T TRANSPORTATION (2017),

https://www.bts.gov/explore-topics-and-geography/topics/sources-aviation-data.

Real Time Regulation in Financial Services (Cont.)

The Bank of England and the FCA are arguably the most advanced when it comes to digital

regulatory reporting. Both were the first financial regulators to set up innovation labs and

programs to interact with startups and new technologies. Their aim is to assess which regulation

is suitable for machine reading, what data could be included in a standardized fashion within a

digital regulatory reporting framework and how these changes will impact data models currently

used for regulatory stress testing, leading to upgrades. This standardization then allows for near

real time data sharing, lowering compliance costs to small and big firms alike.

Closer to home, in the United States, the SEC is implementing rules requiring trading firms to

report intra-day trading data in real time. The SEC, the CFTC and the FDIC have also each set

up fintech/regtech innovation labs to assess fintech solutions in general and regulatory tech

solutions in particular. A Digital Authority could learn from, and expand on, these experiences

of near real time regulation when overseeing an industry that has data at its core.

regime in this regard, and the DA may be able to learn from it and other jurisdictions that have

already taken these steps.

170

c. Data Portability and Mobility

Congress may, at some point, pass a data law of some type that gives consumers control of their

data. For this control to translate into more competitive markets, it must be used to lower

switching costs and facilitate entry. The DA has a role to ensure that users can easily transfer

their data from one service to another in industries where there is a common business model

(e.g., social media, banking, or online grocery shopping). The DA will identify industries where

porting is likely to aid the competitive process. Being able to port one’s data directly lowers the

cost of moving from one service to another, which in turn causes businesses to compete harder to

keep those customers.

171

Consumers who control their data and have the right to receive it in a

standardized format from the business will be able to take advantage of a new entrant by porting

themselves to it, along with their own data. With information about a user’s past purchases, likes,

friends, and so forth, the entrant can provide a higher-quality service and grow more quickly.

The DA could propose a standard for exchanging the data, but remain open to options that

industry favors, provided the format is not itself an entry barrier. The data porting standard

should be updated frequently to accommodate new innovations in the industry. New innovators

can think broadly about what services users might like that rely on this data, or are compensated

through access to the data.

172

The Vestager Report divides data into personal and not personal;

and content into volunteered, observed, and inferred.

173

The DA could determine which of these

types of data must be included in the portability standard.

While a porting regulation lowers consumer switching costs greatly, they may still be high

enough that demand is not sufficiently contestable to induce entry.

174

The DA could also set up a

process by which a customer can choose to send her data to an entrant by authorizing it to be

transferred directly from her former service provider. The DA would need to authorize the

entrant to offer this facility to its consumers and establish regulations to require the incumbent to

transfer the consumer’s data upon the authorized request from the entrant. This may be

particularly useful as the Internet of Things becomes more important; a consumer may wish to

170

See Fletcher, supra note 47.

171

This is why incumbents may create or maintain systems that make data portability difficult. See Section I.1.B.1,

supra.

172

See, e.g., the RadicalxChange group. Mission and Values, RADICALXCHANGE (2019),

https://radicalxchange.org/about/. The group’s chair, Glen Weyl, has argued in a paper written with colleagues that

it’s a mistake to view internet services priced at $0 as simply free; instead, data can be viewed as a form of labor or

barter. See Imanol Arrieta-Ibarra et al., Should We Treat Data as Labor? Moving Beyond “Free”, 108 A

MER. ECON.

ASSOC. PAPERS & PROCEEDINGS 38 (2018), https://pubs.aeaweb.org/doi/pdfplus/10.1257/pandp.20181003

(exploring “whether and how treating the market for data like a labor market could serve as a radical market that is

practical in the near term”).

173

See EC COMPETITION REPORT, supra note 1, at 8; id. at 24.

174

See U.K. COMPETITION REPORT, supra note 1, at 129, § 5.11.

port the food supply service that was bundled with her new refrigerator from Amazon to an

entrant. Being able to authorize the entrant to obtain all her data from Amazon will lower the

cost of switching. Conceptually, automatic porting is no different from manual porting, but it is

mechanized in a way that is likely to raise contestability and therefore can make entry more

profitable. The Vestager Report notes that the GDPR Article 20 provides these data portability

rights to Europeans. If consumers have the right to quickly and easily patronize an entrant

without data lock-in, there will be more incentive to enter into these markets. However, other

entry barriers remain, which we discuss below.

d. Open Standards to Promote Competition

The DA should move preemptively to prevent the consolidation of control over users’ identities,

as this would create a large new source of market power. The DA could create an open standard

so that new entrants can easily offer their own digital identity product that allows users to access

goods and services online. One example of this type of product is Solid, by Tim Berners-Lee,

often named as the creator of the World Wide Web. Solid offers users a “POD” that safeguards

their digital identity that they can use to connect with different services.

175

Several government-

backed efforts at identity portability are underway across the world. These include Estonia’s e-

Estonia initiative to give citizens a unique digital identifier; India’s Aadhaar, a verifiable 12-digit

identity number issued for each citizen which serves as an identifier and authenticator for a

variety of offline and online services; Sweden’s and Norway’s BankId, which allows companies,

banks, and governmental agencies to identify and conclude agreements with individuals over the

internet; and even self-sovereign identity solutions studied by start-ups that would use

blockchain to allow individuals to own their identity credentials and control who can access their

data in online services. If an individual could then port their identity to the platforms and

providers they wish to use, this would again promote entry of new services and erode the

switching costs of established platforms.

The DA could consider creating an open standard that would facilitate micro-payments among

consumers and digital entities. The coordination needed among stakeholders to create a

successful micro-payment system is substantial, and it likely will require assistance and

oversight from a regulator.

e. Merger Review

The behavior that may be of greatest concern to the many policymakers studying powerful

digital businesses is their acquisition of potential competitors. These acquisitions often fall below

the value threshold under which the buyer would need to notify competition authorities in

advance of the deal. As a consequence, authorities have limited or no ability to assess whether a

175

See How It Works, SOLID (2019), https://solid.inrupt.com/how-it-works.

given deal is procompetitive or harmful to competition before it closes.

176

Markets move quickly

and a competitor’s window of opportunity to gain traction against the incumbent is short. For

these reasons, Congress could give the DA merger review authority. Similar to the FCC’s merger

review role, this would be conducted concurrently with the antitrust review done by the FTC or

DOJ, but with different standards and tools. It would not be prudent to alter the nation’s antitrust

laws to accommodate one difficult and fast-moving sector where false negatives are particularly

costly. Therefore, giving additional power over merger review to the sectoral regulator is a good

solution.

These specific merger regulations should require merging firms to demonstrate that the

combination will affirmatively promote competition. This shifting of the burden of proof from

the government (to prove harm) to the parties (to prove benefit) will assist the DA by placing the

job of demonstrating efficiencies on the parties, who have a greater ability to know what they

are. In some cases, the DA’s review may be the only merger review conducted, as it should not

be subject to the minimum size limitations on HSR filings. In particular, notification and pre-

clearance could be required for any acquisition by a business designated as having bottleneck

power.

177

In its merger review process, the DA could be explicitly tasked with evaluating a given merger’s

likely harm to existing as well as potential competition. Another example discussed above is the

case of content or complements that could expand from that position to compete with the digital

bottleneck business itself. As already noted, entry from elsewhere in the vertical (or

conglomerate) chain may be the most effective and promising entry point to challenge an

established bottleneck business. Mergers with either of these types of entrants have the effect of

neutralizing companies that might one day have posed a competitive challenge to the bottleneck

business. This view of potential competition should drive DA merger review.

The decision in Credit Suisse and the dicta, or language, in Trinko greatly expanded the

industries and conduct that have become, for practical purposes, exempt from antitrust

scrutiny.

178

The agencies have understandably been skittish to expend limited resources bringing

cases that risk being thrown out on Trinko grounds.

179

It is important that Trinko not be used to

176

See Section III.2.B, supra, for a further discussion of these difficulties.

177

When network effects are strong, a digital business with bottleneck power will likely only have very small

competitors. Therefore, even small transactions can neutralize an important potential competitor that is poised to

grow. See Section III.3.B, supra.

178

See Credit Suisse Securities (USA) LLC v. Billing, 551 U.S. 264 (2007); Verizon Communications v. Law

Offices of Curtis V. Trinko, LLP, 540 U.S. 398 (2004).

179

Howard Shelanski, at the time the FTC’s top antitrust economist, testified in Congress on behalf of the

Commission a few years after the Credit Suisse decision. He argued that a narrow interpretation of Trinko was

possible. See Is There Life After Trinko and Credit Suisse? The Role of Antitrust in Regulated Industries: Hearing

Before the Subcomm. on Courts and Competition Policy of the H. Comm. of the Judiciary, 111th Cong. (2010)

(statement of Howard Shelanski, Deputy Director for Antitrust in the Bureau of Economics at the FTC),

https://www.ftc.gov/sites/default/files/documents/public_statements/prepared-statement-federal-trade-commission-

courts-and-competition-policy-committee-judiciary-united/100615antitrusttestimony.pdf [hereinafter Shelanksi

create a no-man’s land where neither regulation nor antitrust are applied to harmful behavior. A

traditional antitrust savings clause can no longer be relied upon, as the clause in the 1996

Telecom Act was found insufficient to protect antitrust enforcement in Trinko. Legislation

creating regulation and antitrust enforcement for digital businesses should address this concern

head on. The statute must be extremely specific, explaining for each tool and goal whether it is

intended to supersede antitrust or not. Antitrust enforcers and other agencies can share dual

authority with different review standards and goals. They can account for each other's

determinations in a manner that will minimize inconsistencies without having one always take

priority over the other. Antitrust must remain in full force except where Congress explicitly says

otherwise.

The merger review process must move rapidly. The agency will need a simple and efficient

merger review process so that businesses can move forward without undue delay, and the agency

does not expend more resources than necessary. These concerns indicate that the burden of proof

must primarily be placed on the merging parties who have the incentive, data, and resources to

quickly deliver the right information to the authority. Decisions should be subject to judicial

review, like a rulemaking process under the Administrative Procedure Act, with reliance on the

specialist Competition Court discussed above. However, in light of the difficulty of winning

antitrust cases in current US courts, judicial review will be effective at supporting the mission of

the DA when combined with a more explicit and tougher enforcement statute or a Competition

Court, or both.

In order for the agency to be most efficient, it should use a combination of adjudication and

rulemaking for these merger transaction reviews. Rules help businesses know what behavior is

proscribed, and help the agency clarify its intentions without having to wait for a good test

case—or multiple test cases—to fully explain the issue. Clear rules will help companies avoid

proposing anticompetitive transactions to begin with.

The cost of such regulation is duplicative merger reviews. However, given the importance of

markets that tip and the uncertainty about whether an acquisition is a substitute or complement,

having two reviews is perhaps helpful in getting to the right answer. If the regulator can block an

incumbent platform’s acquisition of potential competitors, and those competitors have the ability

to quickly get their customers’ data, entrants will plan to enter and compete with the platform,

rather than enter and be bought. If the competitive environment is improved by data portability

Testimony]. The key facts in Trinko were that the legislation at issue, the 1996 Telecom Act, went farther than

antitrust law; an agency, the FCC, had issued rules directly regulating the conduct at issue; and the FCC actively

administered those rules. See Trinko, 540 U.S. 398. Shelanski argued that “[w]here a competent agency actively

administers a rule whose standard for the competitive conduct at issue in litigation is more demanding on the

defendant than antitrust law, the Court was right to find it relevant whether the marginal gains outweigh the potential

costs of antitrust enforcement against the same conduct.” Shelanski Testimony at 9. Yet he expressed concern that

courts may use much broader interpretations of the line of cases. The Court in Trinko expressed concern about

misuse of antitrust law by impudent plaintiffs, so some preemption could be limited to private plaintiffs, with expert

agencies being given greater leeway.

and the elimination of exclusionary conduct, the chance of success will rise and consumers will

benefit from increased competition.

f. Interoperability

There may be settings where the DA is charged by Congress to prevent the creation of market

power in the first place due to the importance of the market and the potential harms from the

market power. Such a market would be a candidate for open interoperability standards that could

be used by all competitors and would promote entry. The DA could oversee the creation of an

open standard and its adoption in the market of interest. For example, devices in the home might

be required to adhere to an open standard so that any platform could connect with any device.

Without such a standard, an Apple thermostat would not function in conjunction with a Google

home assistant. Under an open standard, if a consumer had purchased a number of Amazon

appliances and then switched her home assistant from Amazon to Apple, she would not need to

purchase a new refrigerator, thermostat, and security system because all of those devices could

connect to the new platform and stay functional. It is possible that such open standards can slow

down innovation that depends on the interface, but open standards will drastically reduce lock-in

and market power, leading to greater incentive to innovate on the service itself.

2) Regulations that apply to firms with bottleneck power

Some regulations should apply only to firms that meet the DA’s definition for bottleneck power.

In these cases, the bottleneck firm has the incentive and ability to harm competition and is an

important source of consumer welfare. These firms require extra monitoring to be sure they are

not violating antitrust, or other laws, because of the uncertainties in technology and demand, the

speed at which platforms tip, the irreversibility of tipping, and the need for expert evaluation of

the design of algorithms. For all of these reasons, the cost of false negatives is high and

therefore, under conditions of uncertainty, the public interest requires the DA to take a more

interventionist approach.

180

a. Mergers

The DA could be given merger review authority over all transactions involving digital businesses

with bottleneck power because new competition against these entities is the most valuable for

consumers. Businesses with bottleneck power would notify the DA and obtain pre-clearance for

an acquisition of any size. While this would potentially result in many reviews for the DA, their

number will be limited by the number of platforms with bottleneck power.

The DA may want to use its merger review authority over bottleneck firms to assess

consummated mergers, just as the antitrust authority can under section 7 of the Clayton Act. Past

180

If these regulations fail to create or maintain competitive digital markets, the DA should alert Congress that

neither antitrust nor existing regulatory tools are effective so that Congress may consider stronger steps.

mergers of potential competitors that created monopoly positions could be assessed in this way.

Similarly, past vertical mergers (e.g. a business that develops tools for brands to place ads and a

business that runs mechanisms to set ad prices) may be found to cause higher prices or otherwise

lessen competition. Any such anticompetitive mergers could be unwound by the DA.

b. Non-discrimination and foreclosure

Discrimination against current or future rivals is an important tool in a foreclosure strategy by a

digital business with bottleneck market power.

181

As explained above, there are broadly two

types of foreclosure of a complement: one that operates against a complement that is a potential

competitor of the platform itself, and one that operates only on the platform between rival

providers of content. Because large digital platforms today are conglomerates with large eco-

systems of complements, this is an important area in which to enforce competition. Non-

discrimination can be a helpful tool in creating a competitive environment in which entrants are

protected and can thrive, while allowing a platform to vertically integrate to some degree. Non-

discrimination requirements should be used only after careful study because they can also

prevent efficient forms of service that enhance competition.

Anticompetitive foreclosure of a firm that is a current or potential competitor can be addressed

by the antitrust authority if the antitrust law is strong enough. If not, there will be a role for the

DA to develop a stronger rule in order to successfully enforce against such foreclosure by firms

with bottleneck power. Furthermore, the DA could use similar rules in a forward-looking

regulatory context. Effective non-discrimination rules can foster entry and diversity, create

potential sources of disruptive innovation and protect start-ups and other entrants. The goal of

the forward-looking regulation is to prevent a digital business with bottleneck power from

exercising it in order to protect entrants on the platform so that they have the chance to become

competitors of the platform.

Platform strategies to prevent multi-homing are an important category for the DA to include in

its analysis of foreclosure. A platform contract to induce single-homing on one side of a multi-

sided market can be used to reduce competition (e.g., a loyalty payment to drivers of a ride-

sharing service). Likewise, a platform can make it costlier for users to multi-home among

applications on the platform by, for example, limiting data sharing. Such strategies can foreclose

entrants and harm competition; appropriate regulation could limit their use by firms with

bottleneck power.

The second reason for forward-looking regulation is to prevent digital businesses with bottleneck

power from inefficiently expropriating rents created by complements on their platform. As

described above, this harm is less well-protected by antitrust laws so the need for enhanced

181

For a deeper discussion of the power of discrimination in preventing entry by disintermediation, see Section

II.2.B.3, supra.

regulation will be greater when considering foreclosure on the platform. The DA may be able to

build on the concept of “business to platform” regulation that is developing in Europe to create

effective non-discrimination rules.

182

The motivation of this EC regulatory effort is to create a

“fair, predictable, sustainable, and trusted legal environment” in which complementors and

content providers can invest safely and contribute to social welfare.

183

The Vestager Report

states that platforms “[i]f dominant . . . have a responsibility to ensure that they regulate in a pro-

competitive way. Dominant platforms should be subject to a duty to ensure interoperability with

suppliers of complementary services.”

184

Likewise, the Furman Report recommends developing

a platform code of conduct to ensure fairness.

185

The DA could promulgate regulations prohibiting the foreclosure of a competing content

provider on a platform that is vertically integrated. The authority would need to develop rules to

identify foreclosure that might depend on finding certain anticompetitive conduct, market share,

or market power to make such determinations. The DA must also account for potential pro-

competitive innovations that a company with bottleneck power seeks to provide. The data

collected by the DA will allow it to examine such outcomes and weigh all factors to make pro-

competitive determinations. This balancing will be costly and, despite its best efforts, the

regulator may err. However, as the report emphasizes, non-intervention is also costly. Non-

discrimination rules in the past suffered from slow and expensive adjudication, which limited

their usefulness.

186

Strategies for speedy adjudication are addressed in detail below.

c. Bundling

A digital platform with bottleneck power may have a contract with complementors (e.g., retailers

on an ecommerce platform) that bundles together access to their transaction data along with

logistics services. This could have harmful anticompetitive effects.

187

As described above, the

business may also compete against those retailers on its ecommerce site. The business could use

182

For regulatory developments, see Fairness in Platform-to-Business Relations, EURO. COMM’N,

https://ec.europa.eu/info/law/better-regulation/initiatives/ares-2017-5222469_en.

183

Proposal for a Regulation of the European Parliament and of the Council on Promoting Fairness and

Transparency for Business Users of Online Intermediation Services, at 3, COM (2018) 238 final (Apr. 26, 2018),

https://eur-lex.europa.eu/legal-content/EN/TXT/?uri=CELEX%3A52018PC0238 (“The present proposal aims at

ensuring a fair, predictable, sustainable and trusted legal environment for business users, corporate website users,

providers of online intermediation services and online search engines alike, which will limit the occurrence and the

impact of harmful platform-to-business trading practices occurring in certain online activities, thereby safeguarding

trust in the online platform economy and preventing further legal fragmentation of the Digital Single Market.”).

184

EC COMPETITION REPORT, supra note 1, at 71.

185

See U.K. COMPETITION REPORT, supra note 1, at 5.

186

See Leased Commercial Access Modernization of Media Regulation Initiative, Further Notice of Proposed

Rulemaking, 33 FCC Rcd 5901 (2018), https://docs.fcc.gov/public/attachments/FCC-18-80A1.pdf, at 1, n.2 and

accompanying text (noting that “the [Federal Communication] Commission’s 2008 Leased Access Order . . . has

[been] stayed for a decade in conjunction with several judicial appeals”).

187

Retailers may accept what otherwise seems like a bad bargain due to the importance of being available to

customers through the bottleneck. See Section II.2.B.4., supra for a discussion of bottleneck firms’ ability to dictate

business terms.

the retailers’ data to learn which products are selling well and expropriate the ideas and strategies

of the retailer. That data advantage over rivals can enable a company to achieve and/or maintain

critical economies of scale, better predict consumer behavior, and form a powerful barrier to

entry for potential competitors.

188

Bundling may also discourage multi-homing.

Another example of possibly harmful bundling could occur when a platform owner requires

installation of a bundle of applications. Those apps might be chosen to block the growth of rival

apps that were extracting rents from the platform or threatening to be a future competitor of the

platform. The Internet of Things will create more settings in which bundling policy will be

critical. Will a consumer’s new fridge arrive with a supply contract from Amazon’s

WholeFoods? Will the consumer be able to change that contract in some period of time, or is the

consumer permitted to purchase the fridge without any contract?

An antitrust case in these settings may be ineffective in protecting entrants and competition in

digital bottleneck businesses due to the complexity of the problem and the slow pace of

litigation. However, the DA could establish regulations that prohibit anticompetitive bundling by

firms with bottleneck power. Such a firm would be required to demonstrate that its bundle was

on balance procompetitive if foreclosure was alleged. The DA could require unbundling and an

offer to business customers of a choice of contracts in the case of anticompetitive bundling. The

DA would need to enforce such contracts.

iii. DA-Enforced Remedies for Antitrust Violations

When a company has been found liable for violating the antitrust laws, the antitrust authority is

tasked with devising a remedy to restore the lost competition. A fine does not restore lost

competition. No longer engaging in the illegal conduct may help the next entrant or complement

that wishes to interoperate, but it will typically not restore the competition that has already been

lost, particularly in the face of durable barriers to entry that protect incumbent digital platforms.

Antitrust authorities are good at enforcing structural remedies that require no ongoing

monitoring, such as requiring a divestiture between a platform and its content, or the sharing of a

dataset or intellectual property with the entrant, royalty-free. However, reducing entry barriers

often requires a remedy that involves ongoing monitoring, as do behavioral remedies, such as

firewalls between platforms and content. As mentioned above, the DA could also enforce

remedies for antitrust violations identified and addressed by existing antitrust agencies.

189

Below

are some types of remedies for which the DA would be a more appropriate body to enforce.

190

188

See Section I.1.B.1, and Section I.2.A.3, supra.

189

See Section III.3.A., supra.

190

For example, ongoing monitoring. See Section III.3.A, supra.

1) Data Sharing

Anticompetitive conduct may result in a market that has tipped in favor of a single provider

which then benefits from unparalleled access to data. In those cases, a new entrant may find it

impossible to service users with new products as it lacks the scale needed for effective operation.

Data sharing could restore the lost competition. The relevant data to share may not be just

historical data, but present and future data also. Because data are non-rivalrous, an incumbent

can both share its data with a competitor and also keep it. Thus, access to data forms a very

important remedy in the toolkit of both the antitrust authority and the DA. The Furman Report

recommends that agencies mandate “data openness” (which leads to data sharing) to enhance

competition.

191

The Vestager Report likewise recognized that data sharing can help level the

playing field.

192

2) Full Protocol Interoperability

Another useful tool that could restore lost competition is an open protocol and interoperability

standard that would be available for entrants to use on a continuing basis and allow them to

overcome network effects. A bottleneck business whose anticompetitive conduct created a

monopoly position could be required to interoperate with its competitors. Entrants, previously

rendered uncompetitive by network effects, could use the APIs to bring information from the

incumbent bottleneck firm to its own users. In a social media context this would allow the users

of the new service to see not only all the content on their own service, but also content from

friends on an incumbent site that was subject to an interoperability requirement. The network

barrier to entry would no longer protect the incumbent firm, which would then encourage entry

into the industry. Interoperability would facilitate ongoing competition on the merits of the user

experience, rather than on the size of the installed base, and potentially stimulate robust

competition.

Interoperability managed by the DA would be necessary due to the ongoing monitoring needed,

the likelihood of technical change, and the incentive for non-cooperation by the incumbent firm.

The DA could mandate the standard protocols or APIs to be applied and tightly control the

process to avoid having competition undermined by actions of the dominant firm. The DA would

need a process to update protocols at the time of the launch of new functionality or innovation. It

would need rules to protect the privacy and choice of users on one service as some form of

access to them is granted to users of another service. With easy interoperability, users will be

free to make a real choice about which service they prefer. This will encourage new market entry

and vigorous competition between providers.

191

See U.K. COMPETITION REPORT, supra note 1, at 10, 74, § 2.79.

192

See EC COMPETITION REPORT, supra note 1, at 98-107.

3) Non-discrimination

The clear and simple remedy for a case when a bottleneck digital business favors its own content

or complement is divestiture of one of the businesses, either the bottleneck business or the

content/applications. This removes both the incentive and the ability for the conduct. However,

this structural remedy could be costly to consumers in various ways, leading to the conclusion

that a behavioral non-discrimination remedy might be more appropriate. Requiring a dominant

bottleneck to abide by a non-discrimination rule could induce competitive entry by allowing

complementary businesses to thrive and eventually become horizontal competitors to the

bottleneck.

193

The ongoing monitoring necessary to enforce this type of remedy in a specific

antitrust case is not an ideal role for an antitrust agency. However, if the antitrust agency

determines that such a remedy run by the DA would restore and protect competition, the law

would allow it the option of requesting the DA to carry out the remedy. A speedy mechanism to

adjudicate complaints would be of key.

4) Un-Bundling

As described above, the requirement to unbundle contracts could be an antitrust remedy that is

less onerous than divestiture. Such a remedy would require ongoing monitoring that would best

be performed by the DA.

iv. Aligning other policies with competition

In addition to the structural competition tools, the authority should be empowered to align

privacy protection, as well as AI and algorithmic oversight, with competition goals. Privacy

protections that shield consumers from misuse or over-collection of their data can be set up in a

way that raises or lowers entry barriers. This point is often forgotten inside a specialist agency

and therefore we strongly suggest that the DA’s mandate include evaluating and then directing

regulatory solutions in a pro-competitive direction.

194

Mandating that the DA have vigorous

competition as one of its goals will help to make sure that potentially complex analysis takes

place and that the needs of entrants are taken seriously when making policy decisions. The

authority may also need to examine the development and use of algorithms to capture consumer

attention, maximize advertising revenue, and drive consumer purchases or information selection,

and examine how structural or other tools can promote competition in that space as well.

v. Adjudication Process

Adjudication of disputes by this new authority must be quick. Due to the fast pace of change in

these industries, the short amount of time it takes to destabilize or eliminate an entrant, the

193

See Section III.2.B.3.E.

194

It is worth considering whether the competition function and staff of the DA will be more effective when

incorporated into other agency tasks or separated into a distinct division.

substantial discrepancy in bargaining power between digital bottlenecks and their business

customers, and the necessity to use government resources efficiently, a speedy process is crucial.

This could be achieved through mandatory deadlines for dispute resolution or other procedural

rules requiring the authority to produce a decision in a fixed number of days, as well as by

crafting clear standards that are simple to enforce where possible.

The regulatory framework outlined above would coexist with and complement antitrust

enforcement, only blocking transactions and preventing behavior that harm the public interest in

ways clearly defined by Congress and that strengthen the reach of antitrust. The authority could

help elucidate the common line-drawing problem of what behaviors are in or out of the reach of

antitrust through communication with the antitrust enforcement agencies and with the public. An

effective regulator will devise rules that promote competition and new avenues for innovation.

V. Conclusion

This report has discussed the unique combination of attributes of large digital platforms and their

tendency toward entrenched market power. The entry barriers that result are in part due to certain

characteristics of digital technology, but in part also due to behaviors of market participants.

Consumers create entry barriers with their behavioral biases, and incumbents create entry

barriers through strategic use of contracts and technologies, as well as by engaging in various

other activities. The resulting monopoly or concentrated market structures do not serve

consumers as well as would a market in which entry is a credible, or actual, reality.

Public policy has been slow to respond to economic harms resulting from these conditions. Such

harms include advertising prices that are higher than would be expected in a more competitive

environment. The markups distort decision-making and are subsequently passed through to

consumers in the goods and services they buy. There are also quality harms. Platforms have

incentives to provide low quality in order to keep users “engaged” and sell more ads. More

generally, a lack of competition lessens the pressure on any platform to deliver high quality to its

customers for fear that they will move to a rival platform. Perhaps most importantly, insufficient

competition among and for digital platform position distorts and reduces innovation in a sector

that has been—and, under the right conditions, will continue to be—the source of huge benefits

for consumers and society.

It is unlikely that these problems will self-correct, meaning new and revised rules and incentives

will be needed to prevent market power from entrenching a few dominant tech firms as

economic and social gatekeepers. The United States is very far behind the frontier in antitrust

enforcement, both because courts have taken a conservative view of what constitutes

anticompetitive conduct and because agencies have not yet developed expertise in digital

competition cases. Considerable work can be done by academics to help provide new relevant

knowledge and tools to both agencies and courts. It also may be necessary for Congress to pass

new legislation that revises the antitrust laws, establishes a specialist Competition Court, or

both. The committee believes that vigorously enforcing the antitrust laws under these conditions

would be likely to increase entry in digital platform industries, competition, and consumer

welfare. Moreover, such enforcement would result in remedies to restore competition that has

already been lost as well as serve as a deterrent to future anticompetitive conduct. Finally,

because the problems we identify may require action beyond antitrust, we also propose the

establishment of a new digital regulatory agency, or Digital Authority.

Such an agency could increase social welfare by establishing baseline market rules and

conditions that both promote competition and limit the dimensions of competition to those that

benefit consumers, rather than exploit them. In addition, the agency could assist in carrying out

modern competition enforcement, using burdens of proof that reflect new understandings of

market behaviors.

Whether the platforms that are dominant today have achieved that position entirely on the merits

bears further examination. If illegalities are determined, appropriate remedies will be needed to

restore the lost competition. However strong antitrust remedies likely will require coupling with

market-opening regulations to overcome market power that prevented competitive market entry.

Going forward, two important goals for public policy are ensuring entry and competition for

platforms, as well as creating competition on platforms. Forward-looking laws and regulations

should be geared to lowering entry barriers as much as possible and vigorously guarding against

exclusion, lock-in, and foreclosure in the platform context. Without a swift policy pivot to apply

these combined tools, digital markets may tip toward levels of entrenched power that undermine

the benefits of innovative digital businesses for American consumers.