Internet Competition & Regulation Of Online Platforms CPI

Edited by: Aitor Ortiz

Internet

Competition & Regulation

Of Online Platforms

CPI EDITORIAL BOARD

Team

CEO & Founder – David S. Evans

Editor in Chief – Elisa V. Mariscal

Managing Director – Aitor Ortiz

Managing Editor – Nancy Hoch

Latin America Editor – Jan Roth

109

119

134

143

155

TABLE OF CONTENTS

Preface – Daniel O’Connor, Vice-President of Public Policy, CCIA

Summary

Understanding Online Platform Competition: Common Misunderstandings

By Daniel O’Connor

The Move to Smart Mobile and its Implications for Antitrust Analysis of Online Markets

By David S. Evans, Hermant K. Bhargava & Deepa Mani

Failed Analogies: Net Neutrality vs. “Search” and “Platform” Neutrality

By Marvin Ammori

Antitrust Regulation and the Neutrality Trap: A plea for a Smart, Evidence-Based Internet Policy

By Andrea Renda

Multisided Platforms, Dynamic Competition, and the Assessment of Market Power for Inter-

net-Based Firms

By David S. Evans

The Double Duality of Two-Sided Markets

By Alfonso Lamadrid de Pablo

Should Uber be Allowed to Compete in Europe? And if so, How?

By Damien Geradin (Juan M. Delgado & Anna Tzanakis, ed.)

Online Intermediation Platforms and Free Trade Principles – Some Reections on the Uber Prelimi-

nary Ruling Case

By Damien Geradin

Competition Policy in Consumer Financial Services: The Disparate Regulation of Online Market-

place Lenders and Banks

By Thomas P. Brown and Molly E. Swartz

Legal Boundaries of Competition in the Era of the Internet: Challenges and Judicial Responses

By Zhu Li

Can Big Data Protect a Firm from Competition?

By Anja Lambrecht & Catherine E. Tucker

PREFACE

Daniel O’Connor,

Vice-President of Public Policy - CCIA

Dear Readers,

Currently, there is a robust debate on how to

police a relatively new economic actor: the

online platform. As competition regulators and

policymakers have this debate, it has become

increasingly clear that the basics of online

competition and how companies compete is

not widely understood. In this ebook, authors

explore Internet business models and the eco-

nomic phenomena associated with them. Fur-

thermore, contributors examine competition

law concepts and regulatory approaches to

these new actors.

In traditional brick-and-mortar markets, econ-

omists and regulators have become relatively

comfortable with understanding how competi-

tion works. Not to say controversies do not exist,

or antitrust lawyers cannot quibble over the ex-

act borders of a relevant market, but mattress

suppliers, car manufacturers, and airlines do not

present a paradigm shift in understanding who a

company’s likely competitors are. Furthermore,

large brick-and-mortar market shares tend to

be relatively durable. For example, eight of the

ten top global pharmaceutical companies by

revenue trace their roots back to the 19th cen-

tury. Eight of the top ten car companies were

founded before World War II. Although innova-

tion has been a hallmark of both these indus-

tries, it is largely incremental and competition

is relatively predictable. Since the companies

are durable, market shares are fairly accurate

gauges of market power. The toolkits of com-

petition enforcers and the general outlook of

regulators has been shaped by understandings

developed in the study of these markets.

However, understanding Internet competition,

where new product categories are created ev-

ery month, revenue models are complicated

(and in some cases, non-existent), multi-sided

business models are common and disruption

is the norm, proves to be a more complicated

task.

Presently, there are two major conceptual

themes in the robust discussion over competition

policy approaches to online platforms: whether

traditional competition enforcement is sufcient

to police Internet markets, or whether the par-

ticular characteristics of the Internet call for new

ex ante regulation. Also, within competition en-

forcement circles, there is discussion of how to

rene traditional economic tools to reect new

understandings of how markets, particularly on-

line platforms, operate.

Unsurprisingly, proponents of new sectorial reg-

ulation argue that competition enforcement is

not well tailored to these markets and that the

underlying characteristics of the Internet invari-

ably lead to global information monopolies,

which they have likened to public utilities, and

called on regulators and competition enforcers

to treat them as “essential facilities.” They tend

to place signicant weight on the role of net-

work effects in online markets, where the value

of the platform increases each time a new user

is added. When a network gets large enough,

it hits a tipping point at which point competition

against the market leader is futile and market

failure is endemic.

In this simplied view of Internet competition,

numerous, powerful monopolists are virtually

untouchable and therefore free to act in an

anticompetitive manner without the usual dis-

ciplinary threats posed by competition. As Tim

Wu, a U.S. competition enforcer and legal pro-

fessor noted in a Wall Street Journal editorial,

“it’s hard to avoid the conclusion that we are

living in an age of large information monopo-

lies.” As a result, those with this worldview often

call for an overhaul of competition law and spe-

cialized sectorial regulation. However, as sev-

eral authors in this book argue, this stylized view

of Internet competition is misleading and Inter-

net markets are incredibly competitive. Prob-

lematic market denitions, a lack of nuance in

the understanding of network externalities, the

failure to account for the interconnectedness of

markets connected via a multi-sided platform,

and an overemphasis on the competitive signi-

cance of data are some common mistakes that

contributors in this book highlight.

And, to be fair, policymakers and regulators are

not monolithic in their thinking. Some see no

need for new regulatory intervention, pointing

to the exibility of modern competition law to

ensure competitive online markets. As the bi-

partisan U.S. Antitrust Modernization Commis-

sion concluded, there is “no need to revise the

antitrust laws to apply different rules to industries

in which innovation, intellectual property, and

technological change are central features.”

Furthermore, the Commission expressed skep-

ticism about claims that antitrust enforcement

was not well suited to new sectors, saying legisla-

tors “should be skeptical of claims that econom-

ic regulation can achieve an important socie-

tal interest that competition cannot achieve.”

Similar thoughts were voiced by the head of

the UK’s Competition and Markets Authority,

Alex Chisholm, who with a co-author noted in

another CPI publication that the “signicant

risks associated with premature, broad-brush

ex-ante legislation or rulemaking point towards

a need to shift away from sector-specic regu-

lation to ex-post antitrust enforcement, which is

better adapted to the period we are in, with its

fast-changing technology and evolving market

reactions.”

This is not to say that traditional competition en-

forcement alone is enough. Robust competition

enforcement, divorced from an understanding

of the economic realities of the Internet, can do

more harm than good. Increasingly, legacy in-

dustries and displaced competitors have turned

to regulators as a means of hobbling more ef-

cient Internet-based competitors. And, given

that online platforms are often multi-sided mar-

kets that by denition involve multiple constitu-

encies that often share the costs and benets

of the platform asymmetrically, the temptation

is always present for constituencies on one side

to organize politically in an attempt to shift cost

burdens to different constituencies on the plat-

form. It is up to competition enforcers and regu-

lators to sort out when behavior is truly anticom-

petitive versus merely disruptive to established

companies. As the essence of competition law

is the protection of consumers, an emphasis

should be placed on identifying clear consumer

harm when a rm’s behavior is called into ques-

tion. Furthermore, as online platforms need to

balance the needs of multiple constituencies,

weighing the pros and cons of the behavior in

question and determining the net effect of that

behavior is crucial.

Compared to traditional markets, the Internet

has unleashed a boom of entrepreneurship and

competition, which is arguably unrivaled in his-

tory. When it comes to the Internet, enforcers

should consider employing what some have

called “regulatory humility” when faced with

conduct that is alleged to be anticompetitive

but that also has plausible benets for dynamic

competition or consumers, even if these bene-

ts cannot be quantied with specicity. Chilling

the incredible dynamism in these markets is the

last thing a competition enforcer wants to do.

ARTICLE

SUMMARIES

Understanding Online Platform Competition:

Common Misunderstandings

By Daniel O’Connor

The Internet is arguably the most powerful

force in the global economy. Similar to the

printing press, electricity and the steam en-

gine, the Internet has become, in the words

of the OECD, a “general purpose technology

enabler,” a once-in-a-generation technolo-

gy that reorganizes world economic activity

and spurs productivity. Although the positive

effects of this reorganization are apparent,

the rapid pace of innovation and the sub-

sequent restructuring of economic activity

have placed pressure on regulators to under-

stand this vital but rapidly changing new eco-

nomic sector.

Can Big Data Protect a Firm from Competi-

tion?

By Anja Lambrecht & Catherine E. Tucker

There is plenty of hype around big data, but

does it simply offer operational advantages,

or can it provide rms with sustainable com-

petitive advantage? To answer this question,

we look at big data using a classic framework

called the ‘resource-based view of the rm,’

which states that, for big data to provide

competitive advantage, it has to be inimi-

table, rare, valuable, and non-substitutable.

Our analysis suggests that big data is not in-

imitable or rare, that substitutes exist, and that

by itself big data is unlikely to be valuable.

There are many alternative sources of data

available to rms, reecting the extent to

which customers leave multiple digital foot-

prints on the internet.

The Move to Smart Mobile and its Implications

for Antitrust Analysis of Online Markets

By David S. Evans, Hermant K. Bhargava &

Deepa Mani

Online markets have changed as a result of

people shifting massively from using person-

al computers and browsers to using techno-

logically powerful mobile devices and apps.

These changes cover leading online players,

consumer behavior, and products. The use

of smartphones and mobile apps, and the

speed of change, vary between countries

and in particular between countries based

on their stage of development. Mobile app

use is lower in fast-growing countries, such

as India, than in developed ones, such as

the United States. However, as smart mobile

phones with mobile broadband connections

become ubiquitous among consumers in de-

veloping countries, mobile app use in these

countries is likely to leapfrog the use of per-

sonal computers and browsers.

Failed Analogies: Net Neutrality vs. “Search”

and “Platform” Neutrality

By Marvin Ammori

While many have lamented that the term

“network neutrality” is boring and unclear,

that concept has inspired millions around the

world to le comments with national regula-

tors,

and led those regulators to take action

in much of the Americas, Europe, and Asia.

Perhaps as a sign of net neutrality’s success

in public debate, some thinkers have started

borrowing the word “neutrality” for concepts

that are supposedly analogous to net neu-

trality, but really have very little in common

with it. The two best-known expressions are

“search neutrality” and “platform neutrality”

(which apparently also encompasses “app

store neutrality), all of which have prompted

discussion before regulators.

Antitrust Regulation and the Neutrality Trap:

A plea for a Smart, Evidence-Based Internet

Policy

By Andrea Renda

When they look at Internet policy, EU policy-

makers seem mesmerized, if not bewitched,

by the word ‘neutrality’. Originally conned

to the infrastructure layer, today the neutral-

ity rhetoric is being expanded to multi-sided

platforms such as search engines and more

generally online intermediaries. Policies for

search neutrality and platform neutrality are

invoked to pursue a variety of policy objec-

tives, encompassing competition, consumer

protection, privacy and media pluralism. This

paper analyses this emerging debate and

comes to a number of conclusions.

Multisided Platforms, Dynamic Competition,

and the Assessment of Market Power for In-

ternet-Based Firms

By David S. Evans

Market power on each side of a multisided

platform, whether in the form of increasing

prices or decreasing quality, is constrained by

the risk of losing sales on the other sides. That

tends to weaken market power on each side

and encourages platforms to keep prices low-

er and quality higher than they would absent

these feedback effects. In some cases the

nature of the business model, and competi-

tion, result in the platform allowing one type

of customers to participate in the platform for

free or even to subsidize their participation.

Non-price methods of attracting customers

are especially important in this case, partic-

ularly when the business model adopted by

the industry makes it difcult for platforms to

move from free participation.

The Double Duality of Two-Sided Markets

By Alfonso Lamadrid de Pablo

The increasing relevance of multi-sided mar-

kets

and business models in the economy has

over the past few years been mirrored in aca-

demic writings, mostly in economic literature,

and increasingly in competition law enforce-

ment. The intention of this brief intervention

is not to incorporate novel theories into the

discussion of multi-sided platforms nor to sum-

marize the main ndings of the literature that

is currently available. As an avid reader of ac-

ademic works on the subject, and although I

much appreciate their lessons, when I read

them I realize that the vast majority of papers

have been authored by economists, mostly

academics, and only in very rare cases by

lawyers in private practice.

Should Uber be Allowed to Compete in Eu-

rope? And if so, How?

By Damien Geradin (Juan M. Delgado &

Anna Tzanakis, ed.)

Uber’s arrival in Europe has generated mas-

sive demonstrations by taxi drivers and a num-

ber of court judgments banning or restricting

Uber’s services on the ground that the com-

pany engaged in “unfair competition”. Uber

and other online-enabled car transportation

services to connect passengers with drivers

offer an attractive alternative to regular taxi

services. The difculty is that these services

are protected by regulatory measures that

create signicant barriers to entry. Uber’s

business model presents many efciencies

and there is little doubt that it will prevail over

time. Regulatory authorities thus face two

options. One option is to resist the market en-

try of Uber and other similar companies. This

approach would deprive users of attractive

services and trigger many years of litigation.

The other option is to embrace technological

change and allow Uber to compete on a lev-

el playing eld with taxi companies.

Online Intermediation Platforms and Free

Trade Principles – Some Reections on the

Uber Preliminary Ruling Case

By Damien Geradin

Commercial Court No 3 of Barcelona sent a

request for a preliminary ruling to the CJEU

regarding the extent to which Uber, which

operated its uberPOP service in Spain without

an authorization from the Spanish authorities,

should be protected by EU law provisions

designed to ensure the free movement of

services in the European Union. The paper

demonstrates that uberPOP is not a “transport

service” falling under Title VI TFEU, but an “in-

formation society service” within the mean-

ing of the E-commerce Directive. Therefore,

uberPop benets from the protection against

undue trade restrictions provided by this di-

rective, as well as by Article 56 TFEU.

Competition Policy in Consumer Financial

Services: The Disparate Regulation of Online

Marketplace Lenders and Banks

By Thomas P. Brown and Molly E. Swartz

The tension between regulated entities and

new entrants is particularly acute in the con-

text of online marketplace lending. While

bank lenders enjoy regulatory privileges that

enable them to lend immediately to con-

sumers in all 50 states, non-bank lenders are

forced to engage in resource-intensive anal-

yses to satisfy state-specic compliance re-

quirements. As non-bank lenders expand ac-

cess to credit to those currently underserved

by banks—providing new underwriting meth-

odologies, real-time data transmission and

new nancing mechanisms—disparate reg-

ulation of banks and non-bank lenders ap-

pears problematic.

Platform Competition in Two-Sided Markets

By Jean-Charles Rochet & Jean Tirole

Many if not most markets with network exter-

nalities are two-sided. To succeed, platforms in

industries such as software, portals and media,

payment systems and the Internet, must “get

both sides of the market on board ”. Accord-

ingly, platforms devote much attention to their

business model, that is to how they court each

side while making money overall. The paper

builds a model of platform competition with

two-sided markets. It unveils the determinants

of price allocation and end- user surplus for

different governance structures (prot-maxi-

mizing platforms and not-for-prot joint under-

takings), and compares the outcomes with

those under an integrated monopolist and a

Ramsey planner.

Legal Boundaries of Competition in the Era

of the Internet: Challenges and Judicial Re-

sponses

By Zhu Li

Some new characteristics of competition in

the Internet industry, e.g., competition for at-

tention, innovation competition, cross-market

competition etc., have brought about new

challenges and difculties for the legal reg-

ulation of competition. In virtue of the theo-

retical innovation and the innovation of law

applicability, Chinese courts gave creative

judicial responses in the scopes of Anti-Unfair

Competition Law and Anti-Monopoly Law,

claried the legal boundaries of competition

and effectively regulated competition in the

online environment. Certain trends and rules

implicit in this kind of judicial responses are

worth noting.

Understanding

Online Platform

Competition:

Common

Misunderstandings

By: Daniel O’Connor

I. Introduction

The Internet is arguably the most powerful

force in the global economy. Similar to the print-

ing press, electricity and the steam engine, the

Internet has become, in the words of the OECD,

a “general purpose technology enabler,”

once-in-a-generation technology that reorga-

nizes world economic activity and spurs pro-

ductivity. Although the positive effects of this

reorganization are apparent, the rapid pace

of innovation and the subsequent restructuring

of economic activity have placed pressure on

regulators to understand this vital but rapidly

changing new economic sector.

Currently, there is a robust debate on how to

police a relatively new economic actor: the on-

line platform. While there is often a lack of clari-

ty involving exactly what is considered an online

platform, the term generally refers to any online

Daniel O’Connor is the Vice President of Public Policy at

the Computer & Communications Industry Association (CCIA), a tech-

nology trade association with a focus on global competition and Internet

policy. Views expressed in the paper are those of the author and should

not be attributed to any individual members of CCIA.

1 Organisation for Economic Co-Operation and Develop-

ment, Directorate for Science, Technology and Industry Committee for

Information Computer and Communications Policy, Broadband for the

Economy, Ministerial Background Report, (2007) (“OECD Report”),

https://www.oecd.org/sti/ieconomy/40781696.pdf.

service that can function as an intermediary

between two or more clearly identied groups.

Examples include Amazon, eBay, Facebook,

Google and Uber. Both the number and size of

Internet platforms, many of which displace parts

of the brick-and-mortar economy, have grown

rapidly in recent years.

One area of debate involves questions about

how existing laws should be applied to regulate

the behavior of online platform operators. An-

other area of debate is more fundamental and

controversial. It questions the adequacy of ex-

isting bodies of law to deal with the challenges

presented by online platforms, and raises the

prospect of new ex ante regulation of online

platforms.

Proponents of new online regulations argue

that the fundamental characteristics of the In-

ternet invariably lead to global information mo-

nopolies,

which they have likened to public util-

ities, and called on regulators and competition

enforcers to treat them as “essential facilities.”

They tend to place signicant weight on the

role of network effects in online markets, where

the value of the platform increases each time a

new user is added. When a network gets large

enough, it hits a tipping point at which point

competition against the market leader is futile.

These networks effects are then compounded

by the extensive user data possessed by large

2 MarthaIvanovas,TowardsaDenitionofOnlinePlatforms

in the European Digital Single Market (2015), http://crninet.com/2015/

paper/A2a.pdf.

3 Alex Chisholm & Nelson Jung, Platform Regulation—

Ex-Ante versus Ex-Post intervention: evolving our antitrust tools and

practices to meet the challenges of a digital economy. Competition Pol-

icy Int’l. Vol. 11, No. 1, (Spring-Autumn 2015), 7-21 (“Chisholm &

Jung”).

4 Tim Wu, In the Grip of the New Monopolists, Wall St. J.

(Nov. 13, 2010), http://www.wsj.com/articles/SB10001424052748704

635704575604993311538482.

5 Jakob Kucharczyk, Essential vs Useful: Can Online Ser-

vices be ‘Essential’ or Are They Simply Very Useful?, Disruptive

Competition Project (June 16, 2015) (“Kucharczyk”), http://www.proj-

ect-disco.org/competition/030415-essential-vs-useful-can-online-ser-

vices-essential-simply-useful; see also French Senate Report (March

20, 2013), http://www.senat.fr/rap/r12-443/r12-4431.pdf.

online networks. Since Internet companies uti-

lize data to rene and improve their offerings,

having access to more data reinforces and ac-

celerates this unbreakable positive feedback

loop.

In this oversimplied view of Internet com-

petition, numerous, powerful monopolists are

virtually untouchable and therefore free to act

in an anticompetitive manner without the usual

disciplinary threats posed by competition. As a

result, these markets require a reworking of an-

titrust law and specialized sectoral regulation.

As Professor Tim Wu, formerly at the U.S. Fed-

eral Trade Commission (FTC) and now in the

New York Attorney General’s ofce, put it in an

oft-cited opinion piece in the Wall Street Jour-

nal, “it’s hard to avoid the conclusion that we

are living in an age of large information mo-

nopolies.”

An ofcial 2014 presentation by the

French government likewise argued that “com-

petition law is not efcient enough” to deal with

Internet platforms. As such, it recommended

both “new tools” for competition law, includ-

ing “interim precautionary measures” and the

implementation of a specic “legal regime ap-

plicable to such platforms” with no fewer than

ten different policy prongs.

As with the French,

6 See Atlantic Council, Building a Transatlantic Digital

Marketplace: Twenty Steps Toward 2020, (2016) at 22, http://www.

atlanticcouncil.org/images/publications/Building_a_Transatlantic_Dig-

ital_Marketplace_web_0406.pdf (“Some observers believe that online

platformservicesbenetfromself-perpetuatingnetworkeffects.Great-

er usage leads to better data and increased condence in the service,

which reinforces—by design—the propensity for future use. This can

lead to questions about whether data monopolies exist and—to the ex-

tent that they do—can lead to anticompetitive behavior.”)

7 See, e.g., Adam Thierer, Perils of Classifying Social Media

Platforms as Public Utilities, 7-19, available at http://mercatus.org/pub-

lication/perils-classifying-social-media-platforms-public-utilities.

8 See Wu, supra n. 4. A longer version of Professor Wu’s

opinion piece presented a less pessimistic view of Internet competi-

tion. See Mathew Ingram, Should We be Afraid of Apple, Google and

Facebook?, Gigaom (Nov. 25, 2010), https://gigaom.com/2010/11/25/

tim-wu-google-facebook/.

9 Direction Générale des Entreprises (DGE), French views

regarding digital platforms of structural importance for the economy

(April 2015); see also Sigmar Gabriel and Emmanuel Macron, Letter to

the German government has also called on

the European Commission to consider ex ante

regulation of online platforms.

Calls for plat-

form neutrality obligations,

platform regulatory

agencies,

and a complete overhaul of com-

petition enforcement emanate from this line of

thinking.

Other public bodies have not made such

bold pronouncements but are studying online

markets with an eye toward possible future

regulation. The European Commission, in its

sweeping reconsideration of Internet regula-

tion as part of its Digital Single Market initiative,

conducted a public consultation on online plat-

forms.

Brieng memos indicate it is considering

a range of options, including an array of new

laws that would govern the behavior of online

platforms, rework competition law, and expand

telecommunications regulation to encapsulate

online platforms.

Although the Commission is

European Commission Vice President Andrus Ansip (April 28, 2015),

http://www.economie.gouv.fr/les/les/PDF/28042015_-_Lettre_A._

ANSIP.PDF.

10 Jeevan Vasager et al., Europe’s Demands on Google Mount,

Fin. Times (Nov. 26 2014), http://www.ft.com/intl/cms/s/0/66b5149e-

758a-11e4-b082-00144feabdc0.html.

11 See Conseil National du Numérique, Platform Neutrality:

Building and open and sustainable digital environment, (May 2014),

http://www.cnnumerique.fr/wp-content/uploads/2014/06/PlatformNeu-

trality_VA.pdf (“Platform Neutrality Report”); see also Jake Levine, It’s

Time for Social Network Neutrality, Business Insider (July 18, 2011) ,

http://www.businessinsider.com/its-time-for-a-social-network-neutrali-

ty-2011-7 (discussing calls for calls for “Social Network Neutrality”).

12 See Frank A. Pasquale III & Oren Bracha, Federal Search

Commission? Access, Fairness and Accountability in the Law of Search,

Cornell Law Review, Sept. 2008; U of Tex. Law, Public Research Paper

No. 123; Seton Hall Public Law Research Paper No. 1002453; Greg

Sterling, France Trying To Force Disclosure Of Google Algorithm,

Wants To Regulate the SERP (Apr. 20, 2015), http://searchengineland.

com/france-trying-to-force-disclosure-of-google-algorithm-wants-to-

regulate-serp-219243 (discussing French bill to allow telecom regulator

to monitor and regulate the algorithms of search engines).

13 European Comm’n, A Digital Single Market Strate-

gy for Europe, http://eur-lex.europa.eu/legal-content/EN/TXT/PD-

F/?uri=CELEX:52015DC0192&from=EN.

14 Leo Mirani, These documents reveal the EU’s thoughts on

regulating Google, Facebook, and other platforms, Quartz (Apr. 23,

2015), http://qz.com/389905/these-documents-reveal-the-eus-thoughts-

on-regulating-google-facebook-and-other-platforms.

still considering how to proceed, it seems that

some of the more expansive ideas, such as a

new category of regulation for online platforms,

have been tempered.

Given the prominence

of the European Union among international reg-

ulators, it is not surprising that this conversation

on online platform regulation has spread to oth-

er parts of the world.

Other policymakers see no need for new

regulatory intervention, pointing to the exibility

of modern antitrust law to ensure competitive

online markets. As the bipartisan U.S. Antitrust

Modernization Commission concluded, there

is “no need to revise the antitrust laws to ap-

ply different rules to industries in which innova-

tion, intellectual property, and technological

change are central features.”

This is not to say,

as they noted, that new economic understand-

ings should not be incorporated when applying

time-tested competition law concepts. Work

certainly needs to be done there, but rening

the application of the law is different than scrap-

ping it or signicantly altering it. The Commission

also noted the relative merits of competition law

versus economic regulation:

Public policy should favor free-mar-

ket competition over industry-specic

regulation of prices, costs, and entry.

Such economic regulation should be

reserved for the relatively rare cases of

market failure, such as the existence of

natural monopoly characteristics in cer-

tain segments of an industry, or where

economic regulation can address an

important societal interest that compe-

tition cannot address. In general, Con-

15 Id.

16 See, e.g., Yuji Nakamura, Airbnb Faces Major Threat in

Japan, Its Fastest-Growing Market, Bloomberg (Feb. 18, 2016), http://

www.bloomberg.com/news/articles/2016-02-18/fastest-growing-airb-

nb-market-under-threat-as-japan-cracks-down;

17 See Report and Recommendations, Antitrust Modernization

Comm’n (Apr. 2007) at 9, available at http://govinfo.library.unt.edu/

amc/report_recommendation/amc_nal_report.pdf (“AMC Recom-

mendations”).

gress should be skeptical of claims that

economic regulation can achieve an

important societal interest that compe-

tition cannot achieve.

In this paper, I examine common mistakes

that policymakers, regulators, and competition

enforcers make when evaluating Internet com-

petition. Dening markets based on a static

view of competition, applying one-sided anal-

ysis to multi-sided markets, not recognizing the

limits of network effects, and an overempha-

sis on the role of big data are common pitfalls

that often lead observers to conclude that the

Internet is far less competitive than it actually

is. These mistakes lead to the misapplication of

competition law itself, and fuel the perception

that competition law is not sufcient and addi-

tional regulation is needed to police online plat-

forms.

II. Dynamic Markets Living in Static

Worldview

Competition in online markets, including for

online platforms, differs from that of tradition-

al brick-and-mortar markets in important ways

with implications for antitrust enforcement. De-

termining which rms compete against each

other can be more challenging because rms

with varied origins and different business mod-

els may nevertheless provide a similar service

to consumers. Market shares for online services

say little about durable market power: the last

decade is riddled with examples of regulators

or policymakers believing an online rm to be

entrenched, only to become competitively irrel-

evant a few years later. The next market leader

can not only come from a startup entrepreneur,

but also repositioning by existing online rms into

adjacent markets. The current wave of online

platforms has helped lower entry costs even fur-

ther for new entrants, sometimes even challeng-

ing the platforms on which they run.

18 AMC Recommendations at 22.

Online platforms differ from traditional mar-

kets in several respects. Product cycles are short,

borders between “markets” are blurry, barriers

to entry are low, and competition is often for the

market: new companies with better ideas can

quickly come to market and dethrone industry

leaders.

When one looks at the technological foun-

dations on which the Internet is built, these

economic phenomena make sense. First, the

Internet is an open network with a common se-

ries of communications protocols anyone can

utilize. TCP/IP, the Internet protocol suite, gives

entrepreneurs a toolset they can use to trans-

mit data to anyone connected to the Internet.

The creators consciously decided not to pat-

ent the technology to ensure that it was avail-

able to anybody to use.

This decision paved

the way for its rapid deployment and universal

popularity. Second, Internet connectivity is rap-

idly expanding and the hardware that powers

the Internet has been commoditized and is be-

coming cheaper, smaller and more powerful.

Whereas 35 million people were connected to

the Internet in 1995,

that number is nearly 3 bil-

lion today.

The Internet is no longer a tech-

nology available only to the afuent; it is rapid-

ly becoming a worldwide market. In 2014, for

example, India added 63 million Internet users

making it the third largest Internet market in the

world with 232 million users. It is expanding at

a 37 percent year-over-year rate.

Third, soft-

ware platforms and programming languages

that allow people to build new platforms, apps

and websites have evolved and proliferated.

Innovators have a better, and constantly ex-

19 Ryan Singel, Vint Cerf: We Knew What We Were Unleash-

ing on the World, Wired Business (Apr. 23, 2012), http://www.wired.

com/2012/04/epicenter-isoc-famers-qa-cerf.

20 Mary Meeker, Internet Trends 2015 Code Conference,

KleinerPerkinsCaueldByers(May27,2015)at4(“MeekerReport”).

21 Victor Luckerson, Internet Users Surge to Almost 3 Billion

Worldwide, TIME (Nov. 25, 2014), http://time.com/3604911/3-bil-

lion-internet-users.

22 Meeker Report, supra n. 20, at 166.

panding, toolkit to build with. Combine these

“turbochargers,” as David Evans and Richard

Schmalensee refer to them,

and you have the

ingredients for a supercharged market.

A. Blurred Market Borders: Focus on Who

Companies Compete with, Not How They

Do It

Because of these dynamics, it is perhaps not

surprising that competition agencies and other

regulators have struggled to dene online mar-

kets accurately. Lines between markets are of-

ten drawn based on how companies compete,

rather than on whom customers could turn to

for alternatives. For example, a person need-

ing transportation from Washington, DC to New

York can drive, y, take a train or bus, or use an

online carpooling app. Consumers decide by

evaluating the price, quality and speed of those

offerings and will substitute between these op-

tions accordingly. A ridesharing app’s closest

competitor in this context may be a bus, train, or

airplane—none of which looks or operates any-

thing like a ridesharing app. In fact, in France,

the national state-owned railroad provider,

SNCF, sees the online carpooling platform Blab-

lacar as its main competitor.

The FTC appears to have made precise-

ly this error in its Google “search bias” inves-

tigation, which it closed in 2013. In its closing

statement, the Commission suggested that it

considered “general purpose search” as the

relevant market.

This neglected to consid-

23 David S. Evans & Richard Schmalensee, Matchmakers: The

New Economics of Multisided Platforms 40-45 (2016) (“Matchmak-

ers”).

24 Matt Cowan, Blablacar Has Turned Ride-Sharing Into a

Multi-Million-Euro Business, Wired UK (Apr. 15, 2014), http://www.

wired.co.uk/magazine/archive/2015/05/features/blablacar/viewall.

Airbnb is also an online platform with a sharing economy business

model. It allows people to quickly and easily rent out short-term space

in their own houses and apartments. It completes principally against

hotels.

25 Fed. Trade Comm’n, Statement of the Federal Trade Com-

mission Regarding Google’s Search Practices, File No. 111-0163 (Jan.

er that online search engines are just one way

for consumers to get answers and nd informa-

tion.

A consumer can, for example, search for

Italian restaurants in New York City on Google

or Bing, or can use apps like Yelp, TripAdvisor or

Foursquare. The German Federal Cartel Ofce

appears to be repeating this mistake in investi-

gating Facebook for abuse of dominance in an

alleged “market for social networks.”

Consum-

ers can turn to a wide array of substitute services

such as blogs and microblogs, professional net-

works, online forums, photo and video sharing

services, news aggregators, messaging services,

product review sites, social gaming apps, and

virtual worlds.

B. Ephemeral Success: Static Analysis Miss-

es Internet Dynamism

Even in a properly dened relevant market,

market shares may tell us little about durable

market power in online markets. The same un-

derlying fundamentals that allow innovative

new rms to grow their user base quickly also

empower rapid movement away from a wide-

ly-used incumbent when a better alternative

comes along. Yahoo! was the unquestioned

market leader in search before Google came

along with a better idea for ranking websites,

analyzing hyperlink relationships instead of tex-

tual web search.

Facebook overtook both

3, 2013), https://www.ftc.gov/system/les/documents/public_state-

ments/295971/130103googlesearchstmtofcomm.pdf. The Commission

didnotformallydenearelevantmarketordeterminemarketshares,

but the statement noted that “[g]eneral purpose search engines are dis-

tinct from ‘vertical’ search engines.” Id.

26 Daniel O’Connor, The Fast Evolving Market for ‘Answers’,

Disruptive Competition Project (June 8, 2012), http://www.project-dis-

co.org/competition/the-fast-evolving-market-for-answers/.

27 GermanFed.CartelOfce(Bundeskartellamt),Bundeskar-

tellamt initiates proceeding against Facebook on suspicion of having

abused its marker power by infringing data protection rules, (Feb. 3,

2016), http://www.bundeskartellamt.de/SharedDocs/Meldung/EN/

Pressemitteilungen/2016/02_03_2016_Facebook.html?nn=3591568.

28 European UnionLaw, Denitionof relevantmarket, (last

updated Apr. 20, 2011), http://eur-lex.europa.eu/legal-content/EN/TX-

T/?uri=URISERV%3Al26073.

29 Tom Hormby, The Rise of Google: Beating Yahoo at Its Own

Friendster and Myspace because it organized

social networking in a way that provided bet-

ter value to users.

At the time, many believed

that Yahoo!

and Myspace

were unbeatable.

Competition regulators have at times over-

estimated the ability of online rms to maintain

their leading positions. As Adam Thierer has

noted, the temptation to xate on the current

“market” structure is all too commonplace and

“reects the short-term, static snapshot think-

ing we all too often see at work in debates over

media and technology policy. That is, many cy-

ber-worrywarts are prone to taking snapshots of

market activity and suggesting that temporary

patterns are permanent disasters requiring im-

mediate correction.”

In 2002, competition enforcers in the United

States worried about the market power a com-

bined AOL-Time Warner would have, including

AOL’s “dominant” instant messaging platform.

In fact, they were so worried that they imposed

conditions on instant messaging as part of the

merger approval. One FCC Commissioner not-

ed:

AOL’s dominant instant messaging

platform should be interoperable. In-

stant messaging is far more than a nar-

rowband text messaging service; it’s an

Game, Low End Mac (Aug. 15, 2013), http://lowendmac.com/2013/the-

rise-of-google-beating-yahoo-at-its-own-game.

30 Matchmakers, supra n. 23, at 145-48.

31 Tair-Rong Sheu & Kathleen Carley, Monopoly Power on

the Web -- A Preliminary Investigation of Search Engines (Oct. 27,

2001), available at http://arxiv.org/pdf/cs/0109054.pdf (“If there were

to be only a few search engine sites, who are they likely to be? This

analysis suggests that Yahoo would be a contender. It ranks high on

all indicators and it is an old site, offering a variety of services, with a

powerful network position and high audience reach.”).

32 Victor Keegan, Will Myspace Ever Lose its Monopoly?,

The Guardian (Feb. 8, 2007), available at http://www.theguardian.com/

technology/2007/feb/08/business.comment.

33 Adam Thierer, Twitter, the Monopolist? Is this Tim Wu’s

“Threat Regime” In Action?, The Technology Liberation Front (July 1,

2011), https://techliberation.com/2011/07/01/twitter-the-monopolist-is-

this-tim-wus-threat-regime-in-action.

essential platform for the development

of future high-speed Internet-based ser-

vices that rely on real-time delivery and

interaction.

Despite these predictions, AOL Instant Mes-

senger quickly faded into obscurity, replaced

by mobile messaging apps like WhatsApp, Viber

and WeChat. The merged mega-company did

not have durable market power in the Internet

communications space as predicted.

C. Repositioning: Competition from Estab-

lished Players

New competition frequently emerges not

only from unknown Internet startups with a bet-

ter idea, but also from established Internet com-

panies repositioning into new lines of business.

The same dynamics that make it easy for a start-

up to enter also mean that a successful Internet

company with skilled engineers, data centers,

cash reserves, and a good idea on how to do

something better can use its advantages to en-

ter new markets and inject new competition.

Amazon is a perfect example. Amazon be-

gan as an e-retailer. Fifteen years ago, few peo-

ple thought that Amazon would be competing

directly against Microsoft and Google. But the

expertise that Amazon acquired running the

world’s largest e-commerce website and the

enormous, exible infrastructure that powered it

gave Amazon the knowledge and skill to cre-

ate a cloud platform that it could offer to others

to power their own websites. As Amazon’s CFO

said in a 2006 earnings call:

34 Evan Hansen, AOL, Time Warner complete merger with

FCC blessing, CNET, (Jan. 11, 2002), http://www.cnet.com/news/aol-

time-warner-complete-merger-with-fcc-blessing.

35 In addition, the AOL-Time Warner combination was seen

as one of the worst mergers in recent history. See Tim Arango, How the

AOL-Time Warner Merger Went So Wrong, N.Y. Times (Jan. 10, 2010),

http://www.nytimes.com/2010/01/11/business/media/11merger.html.

The reason we’re doing the web ser-

vices that we are doing is because they

are things that we’ve gotten good at

over the last 11 years in terms of build-

ing out this web scale application called

Amazon.com. So as we go about ex-

posing the guts of Amazon, there are

other developers out there who require

those same sorts of web scale services.

So these are assets and skills that we

needed to build for ourselves anyway.

What we're doing here is exposing those

and, over time, build that into a mean-

ingful business.

What started off as a side project that some

market analysts dismissed as a “distraction”

quickly led to Amazon becoming the world’s

largest provider of cloud services, generating

nearly $8 billion in revenue in 2014.

Two of Am-

azon’s closest competitors in cloud hosting to-

day are Microsoft and Google, neither of which

began as cloud infrastructure companies ei-

ther.

Now, all three (and many others) ercely

compete for each other’s customers.

Ama-

36 Amazon.com Q3 2006 Earnings Call Transcript, (Oct.

24, 2006), available at http://seekingalpha.com/article/19142-ama-

zon-com-q3-2006-earnings-call-transcript?part=single.

37 Jeff Bezos’ Risky Bet, Bloomberg Businessweek Cov-

er Story (Nov. 13, 2006), http://www.bloomberg.com/news/arti-

cles/2006-11-12/jeff-bezos-risky-bet.

38 RobertHof,TenYearsLater,AmazonWebServicesDees

Skeptics, Forbes, (Mar. 22, 2016), http://www.forbes.com/sites/rober-

thof/2016/03/22/ten-years-later-amazon-web-services-defies-skep-

tics/#4fc19d7b4dcf. Amazon Web Services (AWS) is the fastest grow-

ing part of Amazon’s revenue stream. Despite only constituting seven

percent of Amazon’s revenue, AWS accounts for a third of Amazon’s

totalprots.

39 Eric Knorr, 2016: The year we see the real cloud leaders

emerge, InfoWorld (Jan. 4, 2016), http://www.infoworld.com/arti-

cle/3018046/cloud-computing/2016-the-year-we-see-the-real-cloud-

leaders-emerge.html.

40 Ari Levy, Can Google catch Amazon and Microsoft in

cloud?, CNBC (Mar. 23, 2016), http://www.cnbc.com/2016/03/23/goo-

gle-aims-to-catch-amazon-microsoft-in-cloud.html; Marco della Cava,

Cloud warriors led by Amazon, Microsoft battle for $300B in spoils,

USA TODAY, (Mar. 25, 2016), http://www.usatoday.com/story/tech/

news/2016/03/25/cloud-warriors-battle-2019s-300b-spoils/82134082.

zon’s disruptive march is not over. It is using its

knowledge and data from running the world’s

largest online store to offer an advertising plat-

form that is competing with Google, Facebook

and Yahoo!.

The Amazon example is far from unique. The

Internet allows companies to quickly reposition

assets and compete in new arenas, particular-

ly if their current business operations give them

unique insights into new lines of business. Apple

and Google were once so non-threatening to

each other that Google’s CEO, Eric Schmidt, sat

on Apple’s board. Now, they are direct com-

petitors in mobile platforms. Facebook, once

seen as a desktop-focused social network, is

now a leader in mobile advertising and mes-

saging, has developed search technology, and

is even challenging YouTube in video sharing.

It is also exploring ways to capitalize on its user

base in e-commerce.

However, these efforts do not always work,

as Google’s failed attempt to get a foothold in

social networking through Google+ (and Orkut

and Google Buzz before that)

and Amazon’s

unsuccessful foray into the mobile phone mar-

ket with its Fire Phone

demonstrate. Big com-

panies without a better idea or solid execution

41 Kelly Liyakasa, Behind Amazon’s Pitch To Advertisers,

AdExchanger, (Oct. 7, 2013), http://adexchanger.com/ecommerce-2/

behind-amazons-pitch-to-advertisers.

42 Julia Greenberg, We Thought Mobile Ads Were Doomed,

Now It’s a $32B Business, Wired Business (Aug. 14, 2015), http://

www.wired.com/2015/08/thought-mobile-ads-doomed-now-32b-busi-

ness. Lara O’Reilly, Facebook Video Is Driving YouTube Off Facebook,

Business Insider (Dec. 9, 2014), http://www.businessinsider.com/face-

book-video-v-youtube-market-share-data-2014-12.

43 Ryan Mac, Facebook Goes All In On E-Commerce By

Bringing Businesses Into Messenger, Forbes,(Mar. 25, 2015), http://

www.forbes.com/sites/ryanmac/2015/03/25/facebook-goes-all-in-on-e-

commerce-by-bringing-businesses-onto-messenger/#5dbb7d0d4747.

44 Garett Sloane, A Look Back at Google’s History of Social

Media Failures, AdWeek (July 1, 2014), http://www.adweek.com/news/

technology/look-back-googles-history-social-media-failures-158700.

45 J.P. Mangalindan, Why Amazon’s Fire phone failed, FOR-

TUNE (Sept. 29, 2014), http://fortune.com/2014/09/29/why-amazons-

re-phone-failed.

can still fail, even as startups succeed in the

same space (compare Pinterest’s success

Google+’s failure). This applies to new markets

too, as Uber was able to pioneer a ride-sharing

app several years after Google failed in a simi-

lar venture.

In 2004, eBay acquired the largest

Chinese online auction company and was the

unquestioned online auction leader in China,

but Taobao gave users a more useful e-com-

merce exchange and eBay shut down Chinese

operations in 2006.

D. New Tools: Platforms Build on Other

Platforms

The frequency of new online entry and re-

positioning owes much to the ever-declining

cost and difculty of developing and marketing

a new online service. Instead of the Internet

drifting towards a world where Internet giants

cement their market positions, each success-

ful wave of Internet platform development has

given would-be entrepreneurs new tools. Now,

Internet companies can build on top of an in-

creasing array of other platforms providing

global reach. Maria Aspan, Senior Editor at Inc.

Magazine, discussed this phenomenon:

Websites, billing, payment process-

ing, cloud computing, communications,

funding—all have been made simpler

by the likes of Squarespace, Slack,

Kickstarter, Dropbox, Amazon’s ubiqui-

tous Web services division, and PayP-

al…. In the past 10 years, these building

blocks have greatly reduced the time—

46 Katie Roof, Matthew Lynley, Leaked Pinterest Documents

Show Revenue, Growth Forecasts, TechCrunch (Oct. 16, 2015), http://

techcrunch.com/2015/10/16/leaked-pinterest-documents-show-reve-

nue-growth-forecasts/.

47 Drew Olanoff, Remember When Google Tried to Start

Uber In May 2005?, TechCrunch (Nov. 4, 2015), http://techcrunch.

com/2015/11/04/remember-when-google-tried-to-start-uber-in-

may-2005.

48 Hanna Halaburda & Felix Oberholzer-Gee, The Limits of

Scale, Harvard Bus. Rev. (April 2014) (“Limits of Scale”).

and cost—involved to start a business,

especially high-tech ones. Thanks to ‘the

emergence of the internet, open-source

software, cloud computing, and other

trends,’ some experts estimate tech-re-

liant ideas ‘that would have cost $5

million to set up a decade ago can be

done for under $50,000 today,’ accord-

ing to a 2014 paper from the National

Bureau of Economic Research.

Netix, for example, uses Amazon Web Ser-

vices (AWS) as its cloud infrastructure.

It also

uses Hadoop’s open source data platform to

handle its highly complex data crunching op-

eration.

Shopify, a fast growing Canadian

e-commerce company listed on the New York

Stock Exchange, has made integration with oth-

er web platforms a key to its growth strategy.

Instead of designing all the functionality its cus-

tomers need, it fosters its own app developer

platform to encourage others to make its offer-

ing more useful to Shopify customers.

By both

integrating with other online platforms, such as

cloud hosting providers, CRM platforms, and on-

line payment processers, and encouraging app

developers to build more functionality for Shopi-

fy users, Shopify has harnessed the robust online

platform ecosystem to build a better product

more quickly than it could have a decade ago.

Furthermore, Shopify has grown quickly despite

competing with e-commerce leader Amazon.

In fact, Amazon’s own platform for small and

medium-sized enterprises (SMEs), Amazon Web-

store, is shutting down

on account of losing the

49 Maria Aspan, Inside the $5 Billion Company That Will

Launch Your Business, Inc., http://www.inc.com/magazine/201604/

maria-aspan/stripe-online-payments-patrick-john-collison.html.

50 EvaTse,NetixCaseStudy,https://aws.amazon.com/solu-

tions/case-studies/netix/.

51 Sriram Krishnan & Eva Tse, Hadoop Platform as a Service

intheCloud,TheNextixTechBlog(Jan.10,2013),http://techblog.

netix.com/2013/01/hadoop-platform-as-service-in-cloud.html.

52 See Shopify Developers, https://developers.shopify.com/.

53 Amy Gesenhues, Amazon To Close Its E-Commerce Plat-

form Website in 2016, Marketing Land (Mar. 19, 2015) http://mar-

ketingland.com/amazon-to-close-its-e-commerce-platform-webstore-

in-2016-122170

competitive battle with Shopify and other SME

e-commerce platforms.

Uber is another example of a successful com-

pany enabled by other platforms. When Uber

started, it used Amazon’s AWS

and Google

Maps

to power its product. Needless to say, if

Uber had to painstakingly assemble its own com-

prehensive mapping infrastructure and build its

own data centers, it is highly unlikely Uber would

have achieved viability quickly enough to be-

come the $62 billion company that it is today.

And in another twist that illustrates the dynamism

of online markets, not only did Uber later begin

acquiring mapping companies

and rolling out

its own eet of mapping vehicles (purchased

from Microsoft)

to develop its own mapping

solution to rival Google’s and Apple’s, but Uber

and Google are set to become erce compet-

itors in the self-driving car, ride-hailing market of

the future.

In sum, successful online rms are anything

54 Rich Duprey, How Did Shopify and Bigcommerce Beat

Amazon At Its Own Game?, The Motley Fool (Mar. 25, 2015), http://

www.fool.com/investing/general/2015/03/25/how-did-shopify-and-

bigcommerce-beat-amazon-at-its.aspx.

55 Although Uber is now migrating off of Amazon’s cloud,

it used Amazon’s Cloud during its initial phase of operations. See

Cade Metz, Why Some StartUps Say The Cloud Is A Waste of Mon-

ey, Wired Business (Aug. 15, 2013) http://www.wired.com/2013/08/

memsql-and-amazon.

56 JillianD’Onfro,AninuentialGoogleMapsexecjustgot

poached to build products at Uber, Business Insider (June 16, 2015),

http://www.businessinsider.com/google-exec-brian-mcclendon-goes-

to-uber-to-work-on-mapping-technology-2015-6.

57 Eric Newcomer, Uber Raises Funding at $62.5 Billion Valu-

ation, Bloomberg Technology (Dec. 3,. 2015, 1;21 PM, updated Dec. 3,

2015), http://www.bloomberg.com/news/articles/2015-12-03/uber-rais-

es-funding-at-62-5-valuation.

58 Joel Zand, Why Uber is buying map companies, The Nex-

tWeb (2015), http://thenextweb.com/insider/2015/07/15/why-uber-is-

buying-map-companies/#gref.

59 Edgar Alvarez, Uber starts rolling out its own mapping cars,

Engadget (Oct. 19, 2015), http://www.engadget.com/2015/10/19/uber-

maps-car.

60 Andrew J. Hawkins, Google vs Uber and the race to

self-driving taxis, The Verge (Dec. 16, 2015), http://www.theverge.

com/2015/12/16/10309960/google-vs-uber-competition-self-driving-

cars.

but insulated from competition. New rivals can

emerge from unexpected places—including

startups, rms in complementary verticals, and

even former partners—and new entrants can

quickly gain access to hundreds of millions of po-

tential customers without having to set up phys-

ical stores, build large manufacturing plants, or

establish complex global supply chains.

III. New Thinking on Multi-sided

Platforms

Around the time the number of worldwide

Internet users passed 600 million, two French

economists made an important breakthrough.

After studying telecommunications networks,

credit cards and operating systems, Jean Tirole

and Jean-Charles Rochet began to see key sim-

ilarities in what they once viewed as very differ-

ent markets.

Instead of being traditional mar-

kets subject to a standard supply and demand

function, many markets actually involved plat-

forms that brought together different groups of

interdependent customers. Pricing and rules on

one side of the platform affected the demand

on other sides of the platform.

Other econo-

mists, recognizing that traditional economics

did not explain platform markets, quickly built

upon Tirole and Rochet’s work. As Evans and

Schmalensee note:

The old formulas—including the ones

we have taught generations of under-

graduates in Econ 1—do not give the

right answers for multi-sided platforms.

61 Platform Economics even has a Nobel Prize as Jean Tirole

received the honor largely on account of his work on the economics of

platforms. See Matthew Yglesias, One paper by Nobel Prize winner

Jean Tirole that every internet user should know, Vox Technology (Oct.

13, 2014), http://www.vox.com/2014/10/13/6968423/jean-tirole-plat-

form-competition.

62 Take OpenTable as an example. OpenTable, an Internet

platform that facilitates restaurant reservations, needed to get restau-

rants to participate in order to attract users, but the platform also needed

users to get restaurants to participate. OpenTable solved this problem

by subsidizing users (free use and rewards) and providing participating

restaurants with table management software. Matchmakers, supra n.

23, at 13-14.

The math is simply wrong. Traditional

economics holds, for example, that it’s

never protable to sell products at less

than cost. The new multi-sided eco-

nomics shows that even paying some

customers rather than charging them

anything can be protable in theory.

This realization has important implications for

competition enforcers and regulators. Business

decisions that could traditionally be viewed as

anticompetitive and predatory—such as below

cost pricing and rules that limit user exibility—

could in fact be welfare-enhancing behavior in

a competitive environment. As a result, looking

at prices on only one side of the market can

lead regulators astray and indicate competi-

tion problems when there are none. Economist

Julian Wright notes that, far from representing

predatory pricing, below-cost pricing is often

used to attract users that provide the greatest

benet to other users of the platform.

Evans

and Schmalensee also note that, for the same

reasons, pricing above marginal cost on one

side of the platform is not indicative of market

power, thus examining pricing on just that side

of the market alone “would result in a false pos-

itive test result for market power.”

Problems

with one-sided analysis can extend beyond just

the platform’s price structure. As discussed in

Part IV, a platform’s restrictive terms for one side

of the market can often be explained not as an

abusive practice but instead by the need to at-

tract participants or maintain the platform’s val-

ue on the other sides of the platform.

Multisided platform analysis has become

63 Id. at 15.

64 Julian Wright, One-Sided Logic in Two-Sided Markets

(Sept. 2003) at 5, available at http://papers.ssrn.com/sol3/papers.

cfm?abstract_id=459362(“Wright”). Wrightidentiedeight assump-

tions from traditional economics that are problematic when applied to

multi-sided platforms.

65 David S. Evans & Richard Schmalensee, The Antitrust

Analysis of Multi-sided Platform Businesses at 20 (Feb. 2013), http://

www.nber.org/papers/w18783.pdf.

“well within the economic mainstream”

with

“no serious controversy among economists on

the topic.”

Yet, competition law and regulato-

ry policy still have not incorporated fundamen-

tal insights from this line of work.

Recent reg-

ulatory investigations and intervention in credit

and debit card platforms, for example, were

based on “conventional wisdom rather than the

logic of two-sided markets.”

Others have raised concerns about over-en-

forcement and the resulting harm to consumers

due to a failure to understand multi-sided mar-

ket dynamics. After examining both regulatory

guidelines and recent competition cases, Alfon-

so Lamadrid observes that the “lack of reliable

measures to quantify the [positive cross-network

externalities]” has led courts to overemphasize

the negatives of supposed anticompetitive be-

havior in platform markets and discount the pos-

itive effects of the behavior on the entire system.

As a result, the practices of multi-sided platforms

that tend to be targeted are “precisely the same

ones [that] may yield benets for consumers.”

And, given that multi-sided markets by denition

involve multiple constituencies that often share

the costs and benets of the platform asymmetri-

cally, the temptation is always present for constit-

uencies on one side to organize politically in an

attempt to shift cost burdens to different constit-

uencies on the platform.

The growth of the Internet and its many

66 David S. Evans, The Consensus Among Economists on

Multisided Platforms and its Implications for Excluding Evidence that

Ignores It at 3 (Apr. 13, 2013), https://www.competitionpolicyinterna-

tional.com/assets/Uploads/EvansJune-3.pdf.

67 Id.

68 See generally Alfonso Lamadrid de Pablo, The double

duality of two-sided markets (“Lamadrid”), available at https://anti-

trustlair.les.wordpress.com/2015/05/the-double-duality-of-two-sid-

ed-markets_clj_lamadrid.pdf. To be sure, some efforts have been made

to incorporate two-sided market analysis in international competition

enforcement, such as the 2009 OECD roundtable on two-sided mar-

kets, but even that discussion illustrated signicant differences and

confusion on how to proceed. OECD Policy Roundtables, Two-Sided

Markets (2009) at 11-15, available at https://www.oecd.org/daf/compe-

tition/44445730.pdf (“OECD Two-Sided Markets”).

69 Wright, supra n. 64, at 1.

70 Lamadrid, supra n. 68, at 8.

71 OECD Two-Sided Markets, supra n. 68 at 226.

multi-sided platforms make incorporation of

two-sided market analysis into competition pol-

icy increasingly important. If not, competition

enforcement and regulation will increasingly

penalize behavior that is benecial, ultimate-

ly leading to consumer harm.

Although pos-

itive externalities and cross-group benets can

sometimes be difcult to quantify using tradi-

tional economic tools, this should not serve as

an excuse to ignore what is now accepted eco-

nomic doctrine.

IV: Networks Effects, Nuance,

and Rules

Historically, economists and competition agen-

cies viewed network effects as a signicant

barrier to entry and protective of strong mar-

ket positions. More modern economic thought

has recognized the various limits of network ef-

fects and negative consequences of platform

growth. There is also increasing recognition that

a successful platform must balance the inter-

ests of the various platform adopters, which

may involve imposing restrictions on platform

participants to minimize negative externalities.

Companies such as Friendster and Myspace

that failed to set and enforce rules to maximize

the value of their platform to all users failed de-

spite early rapid growth, while companies such

as Google (with respect to Android), Facebook,

Yelp, and Uber that enforced rules to maintain

the value of their networks continue to succeed

and grow. Regulatory efforts that would pre-

vent platform managers from setting and en-

forcing restrictions on platform participants risk

the competitive viability of these platforms.

Early economic literature, often discussing

the telephone network, focused on the positive

72 Take Interchange fees for credit card platforms as an exam-

ple. Australia’s decision to limit interchange fees charged to merchants

resulted in higher fees for banks and card holders on the other sides of

the platform and relatively modest cost savings for merchants. See

Howard Change, David S. Evans & Daniel D. Garcia Swartz, The Effect

of Regulatory Intervention in Two-Sided Markets: An Assessment of

InterChange-Fee Capping In Australia, LECG, LLC (Sept. 26, 2005),

http://www.marketplatforms.com/wp-content/uploads/Downloads/

The-Effect-of-Regulatory-Intervention-in-Two-Sided-Markets-An-As-

sessment-of-Interchange-Fee-Capping-in-Australia.pdf.

externalities conferred by adding new users (or

“nodes”) on a network.

This literature exam-

ined direct network effects, which, in the case

of a telephone network, is the idea that every

telephone user gets a direct benet from ad-

ditional telephone users. The widespread view

was that once a telephone network achieved

scale, it made little sense to build a competing

telephone network, and, as a result, the tele-

phone network operator would have signicant

market power. This understanding of network

externalities permeated competition law and

regulatory economics for many years. Markets

with high upfront costs, signicant economies of

scale and network effects lend themselves to

ex ante regulation, as disciplinary competition is

thought to be unrealistic. Furthermore, competi-

tion enforcers tend to view the presence of net-

work effects as an indication of market power:

Indeed, most attention paid to net-

work effects by antitrust enforcers and

scholars—later consolidated in prec-

edents and guidelines—eminently re-

lates to their characteristic as a barrier

to entry. As a result, network effects

have proved to be, in practice, a most

effective basis for legal arguments chal-

lenging allegedly anti-competitive con-

duct.

Modern economic literature on network ef-

fects has added several new wrinkles to this

analysis, with network effects scholarship par-

alleling the advances in the understanding of

multi-sided platforms. As Tirole and Rochet not-

ed, network effects are a common component

73 See, e.g., Jeffrey Rohlf, A Theory of Interdependent De-

mand for a Communication’s Service, Bell Labs. (1974), http://www.

stern.nyu.edu/networks/phdcourse/Rohlfs_A_theory_of_interdepen-

dent_demand.pdf.

74 European Comm’n, Guidance on the Commission's en-

forcement priorities in applying Article 82 of the EC Treaty to abusive

exclusionary conduct by dominant undertakings, ¶¶ 17, 20, http://eur-

lex.europa.eu/legal-content/EN/TXT/HTML/?uri=CELEX:52009X-

C0224(01)&from=EN.

75 Lamadrid, supra n. 68, at 9.

of multi-sided markets.

More recent economic

literature has also examined both direct and in-

direct network effects. Matthew Clements dis-

cussed the difference between the two types of

network effects:

A telephone becomes more valuable

to an individual as the total number of

telephone users increases. This is a direct

network effect. A DVD player becomes

more valuable as the variety of avail-

able DVDs increases, and this variety

increases as the total number of DVD us-

ers increases. This is an indirect network

effect. Network effects have generally

been modeled in a direct sense—indi-

vidual utility increases with the total num-

ber of users—even when the effect is

thought to operate in an indirect sense,

through a complementary good.

Online platforms often exhibit both direct and

indirect network effects. Xbox users benet from

the existence of additional Xbox users because

of the ability to play against additional people

online—a direct network effect. The Xbox plat-

form also exhibits indirect network effects: more

Xbox users attract more video game develop-

ers and accessory makers to the platform.

A. The Perils of Growth: Multi-directional

Network Externalities

Economists have recognized that the exis-

tence of network effects does not necessarily

ensure the success of the rst mover. Networks

can have negative externalities as much as

they have positive externalities. Network ef-

fects may not be uniform (or even in the same

76 Jean-Charles Rochet & Jean Tirole, Platform Competition

in Two-Sided Markets, 1 J. Eur. Econ. Ass’n, 990 (2003

(“[M]any if not most markets with network externalities are character-

izedbythepresenceoftwodistinctsideswhoseultimatebenetstems

from interacting through a common platform.”).

77 Matthew T. Clements, Direct and Indirect Network Effects:

Are They Equivalent? (Jan. 30, 2014) at 2, available at http://sites.sted-

wards.edu/mattclements/les/2011/08/indirect.pdf.

direction) for different groups utilizing the same

platform. For a search engine, more users make

the network more attractive for advertisers, but

more advertisers (or more users) have little to no

effect on user demand.

In other multi-sided

platforms, like newspapers, the effect can be

negative. More readers create more value to

advertisers, but more advertising actually dimin-

ishes the value of the platform to readers.

For

social networks, more people mean more po-

tential connections, which is positive, but also

network congestion, increased search costs,

and more users trying to use the platform for

disruptive or illegal activities. In these instances,

growth can invite competition, not foreclose it.

Several other aspects of Internet services

counter the effects of positive network exter-

nalities. Use of online platforms is inexpensive

or free, which prevents consumer “lock in.” Fur-

thermore, consumers can participate in sever-

al online networks at once, a practice called

“multi-homing.”

For example, many Facebook

users also use Twitter, Instagram, LinkedIn or You-

Tube. Online dating platform users frequently

keep proles on multiple services. Given low

switching costs and user multi-homing, the pres-

ence of network effects do not represent a sig-

nicant barrier to competition in many cases.

As networks grow, incentives for different

groups on a multi-sided platform can diverge.

Negative externalities, such as congestion, in-

creased search costs, and platform fragmenta-

tion, proliferate and can lead to the demise of

78 Giacomo Luchetta, Is Google Platform a Two-Sided Mar-

ket?(Apr. 30, 2012), http://papers.ssrn.com/sol3/papers.cfm?abstract_

id=2048683.id=2048683.

79 E. Glen Weyl, A Price Theory of Multi-sided Platforms,

Am. Econ. Rev., Vol. 100, Issue 4 (2010).

80 Andrew Hagiu & Simon Rothman, Network Effects Aren’t

Enough, Harvard Bus. Rev. (Apr. 2014), https://hbr.org/2016/04/net-

work-effects-arent-enough.

81 Rochet & Tirole, supra n. 76, at 992.

a company or product.

As Hanna Halaburda

and Felix Oberholzer-Gee found, operators of

expanding platforms frequently neglect “to take

into account differences among [their] custom-

ers,” which often leads to the quick collapse of

a platform.

Operating a successful multi-sided

market requires platform operators to capture

as many positive externalities as possible while

mitigating the negative effects of growth.

B. Balancing Incentives: Guarding Against

Fragmentation

Android, Google’s open source mobile oper-

ating system, is a case study in managing this

dynamic. Android was initially very attractive to

phone handset manufacturers who either had

to develop their own mobile operating systems

(something that is expensive, difcult, and not

necessarily a competitive strength of hardware

companies), or license expensive proprietary

operating systems from other companies. In

addition, hardware manufacturers would not

need to cultivate their own networks of app

developers. Instead, the Android platform was

intended to allow developers to write apps

across devices of multiple mobile phone man-

ufacturers. However, as the platform increased

in popularity, the incentives for individual phone

manufacturers to differentiate their products

from other Android companies increased.

But,

if different implementations of Android are not

compatible with one another, Android’s value

to app developers decreases, as it becomes

harder to design apps that function correctly on

different phones. This dynamic is known as op-

erating system fragmentation. If an operating

system such as Android fragments, the value—

and ultimately, the adoption—of the network

will disintegrate. Governance rules can help

82 Ingram, supra n. 8.

83 Limits of Scale, supra n. 48.

84 Gordon Kelly, Samsung is hurting Android, Trusted Re-

views (May 10, 2013), http://www.trustedreviews.com/opinions/sam-

sung-is-hurting-android.

limit fragmentation and maintain the value of

the platform to software developers and end

users, which in turn enhances inter-brand com-

petition and allows open source platforms “to

compete against rivals that have a proprietary

model in which the platform owner has com-

plete control.”

Tech journalists fretted that the Android open

source operating system was doomed to fail on

account of fragmentation.

There was histor-

ical precedent for this fear. Symbian was the

leading smartphone platform in the mid-2000s.

It started as a joint venture between several

handset manufacturers who wanted an alter-

native to Microsoft. The jointly developed oper-

ating system originally achieved great success:

60 percent of smartphones sold between 2004

and 2008 ran Symbian OS.

However, several

manufacturers eventually developed their own

versions of Symbian, creating a hopelessly frag-

mented operating system.

By 2015, Symbian

had gone from the market leader to having

one-tenth of one percent share of the mobile

platform installed base.

UNIX, a popular desk-

top and server operating system originally de-

veloped in 1969, had a similar fate.

85 David S. Evans, The Antitrust Analysis of Rules and Stan-

dards for Software Platforms (2014) at 32, http://chicagounbound.uchi-

cago.edu/cgi/viewcontent.cgi?article=2389&context=law_and_eco-

nomics

86 Steven J. Vaughan, Five reasons Android can fail, ZDnet

(June 22, 2011), http://www.zdnet.com/article/ve-reasons-android-

can-fail/; Adrian Kingsley-Hughes, Android fragmentation is real, ZD-

Net (June 4, 2010), http://www.zdnet.com/article/android-fragmenta-

tion-is-real.

87 Matchmakers, supra n. 23, at 111.

88 For an extended discussion of Symbian, see id. at 110-113.

89 comScore Reports July 2015 U.S. Smartphone Subscriber

Market Share, comScore, Inc. (Sept. 3, 2015), https://www.comscore.

com/Insights/Rankings/comScore-Reports-July-2015-US-Smart-

phone-Subscriber.

90 Tineke M. Egyedi & Ruben van Wendel de Joode,

Standards and Coordination in Open Source Software, at 2 avail-

able at http://ieeexplore.ieee.org/xpl/login.jsp?tp=&arnum-

ber=1251198&url=http%3A%2F%2Fieeexplore.ieee.org%2Fiel5%2F

8864%2F28013%2F01251198.pdf%3Farnumber%3D1251198 (“The

classic example of fragmentation is UNIX, a multi-user operating sys-

Google recognized the risk of fragmenta-

tion from the start. If the platform were to frag-

ment, the negative externalities associated with

growth could quickly destroy the viability of

the platform. As a result, when it launched the

Android project, it required companies partici-

pating in the Open Handset Alliance

to agree

to platform governance rules in the form of

anti-fragmentation agreements. These agree-

ments allow participants to modify and custom-

ize Android but not so much as to cause incom-

patibilities. In addition, Google offered app

distribution agreements, allowing phone manu-

facturers to provide their customers a baseline

suite of Google apps and the Play applications

store.

The fact that Google has offered an-

ti-fragmentation and app distribution agree-

ments since the launch of the Android platform

is compelling evidence that these agreements

are intended to ensure the viability of Android.

C. Follow the Rules: Managing Chaos on

Social Networks

Social networks face similar problems in con-

trolling the negative externalities associated with

growth. Instead of the chaos of fragmentation,

social network operators must guard against a

variety of anti-social and illegal behavior, which

tem and an open source initiative avant la lettre. UNIX was a de facto

standard in the late 1970s. However, different UNIX variants devel-

oped, which fragmented the market.”).

91 According to its website, “The Open Handset Alliance is a

group of 84 technology and mobile companies who have come together

to accelerate innovation in mobile and offer consumers a richer, less

expensive, and better mobile experience. Together we have developed

Android™,therstcomplete,open,andfreemobileplatform.Weare

committed to commercially deploy handsets and services using the An-

droid Platform.” See http://www.openhandsetalliance.com/. As part of

itsefforts to promotea uniedAndroid platform, OHA members are

contractually forbidden from producing devices that are based on in-

compatible forks of Android. Edward Moyer, Alibaba: Google just

plain wrong about our OS (Sept. 15, 2012), http://www.cnet.com/news/

alibaba-google-just-plain-wrong-about-our-os/.

92 Daniel O’Connor, Time To Stick a Fork In These Android

Competition Complaints, Disruptive Competition Project (June 16,

2015), http://www.project-disco.org/competition/061615-time-to-stick-

a-fork-in-these-android-competition-complaints.

reduce the value of the networks for most users.

The demise of Friendster and Myspace are vivid

examples of this phenomenon. Friendster and

Myspace beat Facebook to market with similar

social networking products. However, their fail-

ure to enforce rules that maximized the value of

their networks led to their demise.

Friendster was the rst popular modern on-

line social network. It was launched in 2002

and had over three millions users in a matter of

months.

Friendster wanted to make the online

world more “authentic” and “accountable,”

where people would post real pictures and real

proles. Although the idea worked in the be-

ginning, “fakesters” started to proliferate on its

network. Creative users created fake proles of

celebrities, objects and even concepts.

Not

surprisingly, the funniest and most clever fake

proles quickly became the most popular. Be-

cause Friendster highlighted the most popular

proles, even more users gravitated to them.

Soon the website’s servers were overrun, and

the network slowed.

Fearing that its network

was spiraling out of control and losing value to

the average user, Friendster took steps to rid its

network of fakes proles and limit users responsi-

ble for signicant network activity. This created

a backlash among some users, particularly the

owners of fake proles who had built large fol-

lowings.

Friendster never recovered.

During the peak of Friendster’s turmoil in 2003,

Myspace launched. Thinking that Friendster

was making a mistake by banning fake proles,

93 GaryRivlin,WallowerattheWebParty,N.Y.Times(Oct.

15, 2006), http://www.nytimes.com/2006/10/15/business/yourmoney/

15friend.html?pagewanted=1&_r=2.

94 See, e.g., danah boyd, None of this is Real, at 22, http://

www.danah.org/papers/NoneOfThisIsReal.pdf. One particularly cre-

ative usercreated a “War” prole, which listed its profession as “re-

solving disputes.” Lessley Anderson, Attack of the Smartasses, SF

Weekly (Aug. 13, 2003), http://www.sfweekly.com/sanfrancisco/at-

tack-of-the-smartasses/Content?oid=2149018&showFullText=true.

95 boyd, supra n. 94.

96 Id.

Myspace’s founders welcomed them. Sever-

al high prole “fakesters” publicly migrated to

Myspace and the social network grew rapidly.

The website grew so rapidly, in fact, that market

analysts labeled it a “natural monopoly,”

and

newspaper columnists discussed whether Mys-

pace “would ever lose its monopoly.”

Never-

theless, Myspace’s more open environment spi-

raled out of control. Partial nudity and obscene

content became commonplace. News reports

of minors who lied about their age and child sex

predators who preyed on them caused public

concern.

Although some people were at-

tracted to the Wild West of the Web, advertisers

were not and the site oundered.

100

Facebook launched in 2004, a year after

Myspace. It started off as a small network of

students at a limited number of colleges. Face-

book realized from the start that anonymous

online activity can lead to general misbehavior

and harassment. It enforced “real” proles by

requiring valid college email addresses. Face-

book gradually expanded to other colleges

and to wider networks, and zealously enforced

its real name policy. Furthermore, the company

rigorously enforced its terms of service and out-

lawed obscene and objectionable content and

nudity.

101

Needless to say, the strategy worked,

and the company grew exponentially. In 2015

it collected nearly $18 billion in revenue, almost

entirely from advertising.

102

97 John Barrett, MySpace Is a Natural Monopoly, TechNews-

World (Jan. 17, 2007), http://www.technewsworld.com/story/55185.

html.

98 Keegan, supra n. 32.

99 Sam Jones, MySpace removes 90,000 sex offenders,

The Guardian (Feb. 9, 2009), http://www.theguardian.com/technolo-

gy/2009/feb/04/myspace-social-networking-sex-offenders.

100 Matchmakers, supra n. 23, at 145-48.

101 Even famous people, like Lindsay Lohan, were kicked off

the platform for violating its policies. Duncan Riley, Lindsay Lohan

Banned By Facebook, Inquisitr (Dec. 2, 2008), http://www.inquisitr.

com/10537/lindsay-lohan-banned-by-facebook.

102  Yahoo! Finance, http://nance.yahoo.com/q/ks?s=F-

B+Key+Statistics.

D. Regulating Fairness: Platform Neutrality

Not the Answer

The rapid growth of online platforms has led

to antitrust investigations and litigation related

to platform governance mechanisms.

103

Anti-

trust complaints, such as those leveled against

Google Search, Android’s mobile platform, and

interchange fees targeting restrictions on mer-

chant pass throughs,

104

have targeted platform

operators for practices aimed at balancing

interests between different constituencies on

their networks. Websites that are demoted for

low-quality scores may object, but search op-

erators need to ensure that consumers receive

high-quality search results. Some mobile carriers

might want to force Android users to utilize their

proprietary apps, but consumers and app de-

velopers benet from a standard array of APIs

and services on different implementations of

Android. Merchants might want the freedom

to pass on interchange fees to consumers, but

consumers benet from the clarity of the stan-

dardized pricing of products.

There have also been increasing calls for

platform regulation premised on an amorphous

goal of platform neutrality.

105

In introducing its

public consultation on platforms, the European

Commission discussed platforms being powerful

intermediaries who have the “power to shape

the online experience of its customers on a per-

sonalized basis and to lter what the customer

sees.”

106

In testimony before the British House of

Lords, Charly Berthet, Rapporteur for the French

Government’s Digital Council (Conseil National

103 Mark Scott, Facebook Faces German Antitrust Investiga-

tion, N.Y. Times (Mar. 2, 2016), http://www.nytimes.com/2016/03/03/

business/international/facebook-faces-german-antitrust-investigation.

html.

104 AmEx Limits on Merchants Violate US antitrust Laws,

Judge Says, LA Times (Feb. 19, 2015), http://www.latimes.com/busi-

ness/la--american-express-antitrust-lawsuit-20150219-story.html

105 See, e.g., Platform Neutrality Report, supra n. 11.

106 DSM Staff Working Document, pg. 53

du Numérique), complained that the decisions

of online platforms to curate and present con-

tent on their networks were threats to “plural-

ism” and to “freedom of expression” and called

for a regulatory response.

107

The examples above are not isolated cir-

cumstances. All platform operators eventually

grapple with how to balance interests and how

to enforce rules that maintain the value of their

networks as they grow. As noted by Parker et

al. in their book Platform Revolution, “multi-sid-

ed platforms involve interests that don’t always

align. This makes it difcult for platform manag-

ers to ensure that various participants create

value for one another, and make it likely that

conicts will emerge that governance rules must

resolve….This is a juggling act that even giants

and geniuses get wrong.”

108

Yelp bans fake reviews that can make the

platform less useful to users, and it spends sig-

nicant resources removing questionable re-

views and punishing violators of its terms of

service.

109

eBay has strict terms and conditions

that, among other things, prevent transactions

107 He pointed to the controversy surrounding Facebook’s orig-

inal decision to ban posts of Gustave Courbet’s painting “The Origin of

the World” because it featured nudity as an example of this threat, stat-

ing:“[o]nlineplatformsdenetheirownrulesregardingwhatcontentis

authorised and what content can be withdrawn.” See Revised transcript

of evidence taken before The Select Committee on the European Union

Internal Market Sub-Committee Inquiry on Online Platforms and the

EU Digital Single Market (Nov. 9, 2015) at 13, available at http://data.

parliament.uk/writtenevidence/committeeevidence.svc/evidencedocu-

ment/eu-internal-market-subcommittee/online-platforms-and-the-eu-

digital-single-market/oral/24970.pdf (“Berthet Testimony”); see also

Facebook could be sued for censoring this 19th-century nude painting,

The Associated press (Feb. 12, 2016), http://mashable.com/2016/02/12/

french-court-facebook-nude-painting/. The fact that Courbet’s painting

could be seen in thousands of other places online, at the slightest touch

of a mouse or swipe of a smartphone, did nothing to prevent this argu-

ment.

108 Geoffrey Parker, Marshall W. Van Alstyne and Sangeet

Paul Chaudry, The Platform Revolution: How Networked Markets Are

Transforming the Economy And How to Make Them Work For You 159

(2016).

109 JudyWoodruff,Spottingthefakesamongtheve-starre-

views, PBS (Jan. 19, 2015), http://www.pbs.org/newshour/bb/spotting-

fakes-among-ve-star-reviews.

from migrating off its platform, ban sellers from

bidding on their own items to drive up the price,

and outlaw misrepresenting your identity. These

terms restrict behavior on individual users on var-

ious sides of its platform to ensure a “safe and

fair environment” for all. Sellers might want the

freedom not to disclose their identify or to nal-

ize the sale outside of eBay’s platform to avoid

commissions, but buyers would have no protec-

tion in case of fraud.

110

Airbnb mandates that

its hosts adopt one of three of its cancellation

policies

111

and meet minimum hospitality stan-

dards

112

to guarantee a baseline user experi-

ence across its offerings. As more drivers come

onto its platform, Uber aggressively screens

them through background checks and utilizes

user feedback to quickly drop low-rated drivers

from its platform. Despite complaints that the

ratings system is harsh and unfair,

113

all it takes is

a few negative experiences before riders ee to

competitive offerings like Lyft or taxis.

As these examples illustrate, platform opera-

tors must make delicate judgments, particular-

ly when different platform constituencies have

diverging interests, to keep their platforms valu-

able to all users. Attempting to mandate plat-

form neutrality or penalize behavior based on

its effects on one side of the platform can easily

throw off this balance to the detriment of con-

sumers. Furthermore, treating online platforms

as “essential facilities,” which would limit the

ability of online platform operators to choose

who accesses their platform and on what terms,

could have similar negative effects and harm

110 eBay Inc. Rules and Policies, http://pages.ebay.com/help/

policies/overview.html

111 Airbnb, Inc. Cancellation Policies, https://www.airbnb.com/

home/cancellation_policies

112 Airbnb, Inc. Hospitality Standards, https://www.airbnb.

com/hospitality

113 Polly Mosendz, Uber Drivers Are Fighting for Better Treat-

ment from the Company, The Wire (June 25, 2014), http://www.thewire.

com/business/2014/06/uber-drivers-protest-their-employer/373388.

competition and innovation.

114

As noted by

Alex Chisholm and Nelson Jung from the United

Kingdom’s Competition and Markets Authority

in a discussion of the essential facilities doctrine,

“concepts applying to ports cannot simply be

copy-pasted into the digital world.”

115

V. Can Controlling User Data Lock

Out Competitors?

Some enforcers and academics have iden-

tied control over data as a potential antitrust

concern in recent years.

116

While their state-

ments are not always clear as to what they

mean by “big data,” these advocates usually

focus on user data that is collected by online

rms as part of their business operations—for ex-

ample, a shopping site collecting data about

which products its users have looked at, which

ones they have purchased and how much they

have paid, or a search engine collecting data

about which results users clicked on after prior

searches.

The theory postulates that data is a necessary

input for online rms—many have analogized it

to the oil of the 21st Century

117

—as many Internet

businesses rely on user data to improve their ser-

vices. This creates what some have described

as a “feedback loop,” whereby a rm needs

a large body of user data to create a product

114 See Kucharczyk, supra n. 5.

115 Chisholm & Jung, supra n. 3, at 11.

116 See, e.g., Maurice E. Stucke & Allen P. Grunes, Debunking

the Myths over Big Data and Antitrust, CPI Antitrust Chronicle (May

2015); European Data Protection Supervisor, Privacy and competi-

tiveness in the age of big data: The interplay between data protection,

competition law and consumer protection in the Digital Economy (Mar.

2014), available at https://secure.edps.europa.eu/EDPSWEB/webdav/

shared/Documents/Consultation/Opinions/2014/14-03-26_competiti-

tion_law_big_data_EN.pdf; Nathan Newman, Search, Antitrust and the

Economics of the Control of User Data (Sept. 14, 2013), available at

http://papers.ssrn.com/sol3/papers.cfm?abstract_id=2309547.

117 Nathan Newman, Taking on Google’s Monopoly Means

Regulating Its Control of User Data (Nov. 24, 2013),

 http://www.hufngtonpost.com/nathan-newman/taking-on-goo-

gles-monopol_b_3980799.html.

that will attract new users, leading to signicant

economies of scale and causing the market to

permanently “tip” in favor of already dominant

platforms.

118

Of course, using more and better

inputs like user data to improve products and

services is generally the essence of competition,

so regulators should tread carefully when faced

with complaints that such conduct is leading to

anticompetitive effects.

119

In online markets, the competitive harms that

could arise from large rms’ access to extensive

user data usually exist only in the realm of theo-

ry. To properly evaluate claims that a rm’s con-

trol over data operates as a barrier to entry, one

must understand how data does and does not

work for a particular online business. Real-world

experience has illustrated several reasons why

even a dominant rm’s collection and use of

user data is very unlikely to create any mean-

ingful barriers to entry, for several reasons.

First, and perhaps most importantly, data

is just one of many inputs that dictate success

for online businesses, and ultimately it is not the

data, but what a company does with it, that re-

ally matters. Many online rms have successful-

ly entered markets with large, well-established

players without having access to signicant user

data. For example, Spotify was able to success-

fully enter the ad-supported streaming music

space notwithstanding Pandora’s early success,

by offering a different model. Whereas Pandora

relies on user data—which artists a user selects

and which songs a user has “liked”—to deter-

mine which songs will be played, Spotify opted

for a less curated approach that allows users

to select their own songs from a wider library,

reducing the need to rely on user data where

Pandora had a substantial advantage. Firms

118 Andres Lerner, The Role of Big Data in Online Platform

Competition (Aug. 26, 2014) at 19-20, available at http://papers.ssrn.

com/sol3/papers.cfm?abstract_id=2482780.

119 Foradiscussionofprocompetitivebenetsofuserdatacol-

lection, see id. at 10-15.

can also enter and quickly extract the data they

need from the initial user base.

120

In some cas-

es, rms have eschewed the collection or use of

personal information and have attracted priva-

cy-sensitive consumers on that basis. For exam-

ple, the search engine DuckDuckGo markets

itself as not collecting query data, and thereby

does not use that data to generate such search

results, even though other traditional search en-

gines do.

121

Second, data is non-rivalrous and non-exclu-

sive. Non-rivalrous means that one party’s use

of data does not prevent another party from

collecting and using that same data, even from

the same source.

122

When a consumer uses of

a gallon of gasoline, another driver cannot use

that same gas. As a result, it is said to be rivalrous

in consumption. When Netix uses knowledge of

a subscriber’s preferences to highlight shows she

might like, Hulu can still use knowledge of her

preferences to highlight similar shows. Non-ex-

clusive means that a rm cannot exclude others

from collecting that same data. As a result, no

single rm controls all, most, or even a signicant

amount of the total universe of user data. Many

online rms have access to vast swaths of infor-

mation about users, including data brokers who

make that information potentially available to

startups.

123

In addition, users frequently “multi-home.”

Because consumers use multiple online services,

even for the same task, multiple providers have

the ability to collect data on the same user. In

social networking, multi-homing has resulted

in Facebook continuing to ourish while also

120 Darren Tucker & Hill Wellford, Big Mistakes Regarding

Big Data, Antitrust Source, Dec. 2014, at 7, available at http://www.

americanbar.org/content/dam/aba/publishing/antitrust_source/dec14_

tucker_12_16f.authcheckdam.pdf.

121 DuckDuckGo Internet Privacy Review (Dec. 12, 2014),

https://www.expressvpn.com/blog/duckduckgo-internet-privacy-re-

view/.

122 Lerner, supra n. 118, at 21.

123 Tucker & Wellford, supra n. 120, at 3.

allowing many new entrants, which often fo-

cus on particular types of content (Instagram,

Twitter, Pinterest) or interests (Reddit, LinkedIn),

rather than offering a broader Facebook-like

approach. But this also plays out across static

market boundaries too, as with the example of

using either Google, Yelp or an alternative ser-

vice to nd a restaurant for dinner.

Third, data quickly becomes stale. As Tuck-

er and Wellford note, “[h]istorical data can be

analyzed for trends but has comparatively little

value when used for real-time decisions, such as

which ad to serve. As one study found, ‘90% of

the data in the world today has been created in

the last two years. . . . 70% of unstructured data

is stale after only 90 days.’”

124

A new entrant

need not replicate an incumbent’s vast store-

house of old, stale data. What an upstart cares

about—if it cares about data at all—is the fresh

material that the new entrant has the same op-

portunity to obtain as the incumbent. Thus any

data advantage by incumbent rms is likely to

be ephemeral at best.

Fourth, data has diminishing returns to scale.

As Lerner discusses in detail, there is substantial

empirical evidence of rapidly diminishing returns

to scale in two markets that have been cited

as potential areas where data could be a barri-

er to entry: online search (where user data can

improve the quality of search results) and ad-

vertising (where better user data might allow

better targeting of ads to users).

125

Because of

diminishing returns, the relevant question for an

entrant is not whether it can obtain the same

amount of data as the incumbent, but merely

whether it can obtain enough data to make

its product work. In online markets, there is fre-

quently a vast gap between what might be had

and what is needed.

Online Dating: A Case Study

124 Id. at 4.

125 Lerner, supra n. 118, at 35-45.

Online dating has become an almost ubiq-

uitous feature of the social lives of singles. The

National Academy of Sciences reported in

2013 that more than a third of people who mar-

ried in the U.S. between 2005 and 2012 met their

partner online, and half of those met on dating

sites. Online dating sites now comprise a $2.1

billion per year industry in the U.S. alone.

126

Besides helping to nd a potential partner,

online dating is a good match for those wanting

to examine competitive dynamics among Inter-

net companies for two reasons: (1) it is one of

the longest-existing online industries; and (2) on

the surface, online dating exhibits many of the

theoretical bases that should have caused the

market to “tip” to a single dominant rm long

ago.

•

Network Effects. Single, heterosexu-

al men want to nd an online dating plat-

form that has a lot of female members, and

vice-versa. This is a classic example of net-

work effects, specically cross-side network

effects in a two-sided market.

127

A site that

has a large user base should, therefore,

have a strong advantage in attracting new

users.

•

Data. User data has long played a role

in online dating, through various forms of

“algorithmic matchmaking.”

128

This takes

place in a variety of ways. Some sites allow

(OkCupid) or require (eHarmony) users to

126 Steve Yoder, How online dating became a $2 billion in-

dustry, The Week (Feb, 19, 2014), http://theweek.com/articles/450841/

how-online-dating-became-2-billion-industry (“Yoder”).

127 See, e.g., Voigt & Hinz, Network effects in two-sided

markets: why a 50/50 user split is not necessarily revenue optimal

(July 2015), available at http://download.springer.com/static/pdf/118/

art%253A10.1007%252Fs40685-015-0018-z.pdf (“Users may derive

positive cross-side network effects (CNEs) from the participation of

members on the other side of the market, which means the larger the

installed user base on one side of the platform, the more attractive the

service for the opposite side’s users.”).

128 James Bridle, The algorithm method: how internet dat-

ing became everyone's route to a perfect love match, The Guardian

(Feb. 9, 2014), http://www.theguardian.com/lifeandstyle/2014/feb/09/

match-eharmony-algorithm-internet-dating.

answer extensive questionnaires designed

to help the users nd compatible matches.

In addition to data provided by users, on-

line dating sites also have data about what

those users actually do—whose proles they

browse, for how long, who they do (and do

not) message, and with what results. They

use that information, to varying degrees, to

recommend potential matches to their us-

ers, theoretically improving outcomes for ex-

isting users and thereby making the platform

more attractive to potential future users.

Given these characteristics, a regulator that

believed that network effects and data barriers

cause platforms to tip to a single dominant sup-

plier might well have concluded that the online

dating market has been ripe for intervention. In

late 2006, for example, Yahoo! Personals had

“grabbed the top rank on every measurement”

and had almost twice as many active users as

its next-closest competitor.

129

But rather than rid-

ing a wave of permanent dominance, Yahoo!

Personals folded less than four years later. And

Yahoo! was not alone. According to the Pew

Internet Survey, 6 of the 10 most popular on-

line dating sites in 2005 were defunct as of 2013

(and, conversely, 5 of the 10 most popular on-

line dating sites in 2013 did not exist in 2005).

130

Today the market is allegedly “dominated

by big name, mass audience sites, like Match.

com and eHarmony.”

131

Should we now be

concerned that a “dominant” site like eHarmo-

ny is abusing a dominant market position by

making its users answer private questions about

their wants and needs in order to communicate

with other users? Or that knowing the answers

to those questions will somehow entrench eHar-

129 Max Freiert, Beyond Skin Deep: The real top dating sites,

Compete (Dec. 28, 2006), https://blog.compete.com/2006/12/28/top-

online-dating-sites-yahoo-personals-true-match-plenty-of-sh/.

130 Pew Research Center, Online Dating & Relationships (Oct.

21,2013),at50,availableathttp://www.pewinternet.org/les/old-me-

dia/Files/Reports/2013/PIP_Online%20Dating%202013.pdf.

131 Yoder, supra n. 126.

mony as a durable monopolist that would not

face a threat of new entry? It is hard to see

how, given that there currently are more than

2,500 other online dating sites operating in the

U.S., and as many as 5,000 worldwide, with new

sites popping up seemingly daily.

132

How have competitors been able to consis-

tently to crack into this business despite what

some might consider an impenetrable advan-

tage held by large incumbents? The structur-

al advantages that appear so strong in theory

appear not to have mattered much in the real

world. The potential network effects in online

dating are substantially weakened by wide-

spread multi-homing by users. Many, if not most,

online daters maintain proles on multiple plat-

forms.

133

Because of this, many platforms are

able to generate a sufciently large user base

to be attractive to potential users. Apparent

network effects are also weakened by the di-

versity among the user base. If a user is looking

for a male aged 21-28, a large pool of men over

the age of 40 does not increase that site’s value

to that user.

Moreover, despite access to substantial

amounts of user data, online dating sites might

not have gained any particularly signicant in-

sights from it.

134

As one recent study conrmed,

“no compelling evidence supports match-

ing sites’ claims that mathematical algorithms

work.”

135

If the data does not provide com-

132 Simon Davies, The Online Dating Market is Not Oversat-

urated, Tech.Co (May 9, 2015), available at http://tech.co/online-dat-

ing-market-not-oversaturated-2015-05.

133 Molly Wood, Led by Tinder, a Surge in Mobile Dating

Apps, N.Y. Times (Feb. 4, 2015), available at http://www.nytimes.

com/2015/02/05/technology/personaltech/led-by-tinder-the-mobile-

dating-game-surges.html.

134 See, e.g., Finkel, et al, Online Dating: A Critical Analysis

From the Perspective of Psychological Science, Psychological Science in

the Public Interest (Jan. 2012), available at http://psi.sagepub.com/con-

tent/13/1/3.full?ijkey=cK9EB6/4zQ0AM&keytype=ref&siteid=sppsi.

135 Dating Site Factors concerning Imperfect Preferences for

EfcientMatchings, CornellUniversity Networks IICourse Blogfor

INFO 4220 (cited by John Tierney, A Match Made in the Code, N.Y.

Times (Feb 11, 2013), available at

pelling value to the incumbent, competitors

should be able to enter successfully without it.

And contrary to the claim that data functions

as a barrier to entry, for online dating, data has

actually facilitated entry by making it easier for

users to multi-home. Now, rather than ling out

time-consuming proles, many dating sites allow

users to share information from Facebook and

other platforms.

But perhaps most importantly, the key to suc-

cess for online dating services, as it has been in

many Internet industries over time, has been to

build a better mousetrap by approaching the

problem in a new or different way. Rather than

focusing on broad-based market leaders like

eHarmony and Match, many competitors have

succeeded by focusing on smaller niche mar-

kets. Today, daters can nd sites that match

based on age (e.g., OurTime.com), race (Black-

PeopleMeet.com), occupation (FarmersOnly.

com), and a host of other factors. Those sites

have overcome network effects by focusing on

quality over quantity of matches: the self-select-

ing nature of these sites eliminates a lot of the

“chaff” users that are not of interest.

The well-known app Tinder chose a different

approach, avoiding matching based on algo-

rithm, and focused instead on proximity, show-

ing users other nearby users and offering them a

simple “yes or no” choice, and became so pop-

ular that “Swipe Left” and “Swipe Right” have

now entered common vernacular. Tinder’s

“secret sauce”—the “double opt-in,” where us-

ers declare secretly who they are attracted to

and are only matched after both say yes—was

based on the idea that “you feel more com-

fortable approaching somebody if you know

they want you to approach them.”

136

Now, Tin-

der has millions of users and facilitates 26 million

http://www.nytimes.com/2013/02/12/science/skepticism-as-eharmo-

ny-defends-its-matchmaking-algorithm.html?pagewanted=all).

136 Emily Witt, Love Me Tinder, GQ (Feb. 11, 2014), avail-

able at http://www.gq.com/story/tinder-online-dating-sex-app?current-

Page=4#1.

matches a day.

137

But Tinder can’t rest on its laurels. It, too, fac-

es the threat of dynamic competition, includ-

ing sites that have not even launched yet, and

quite possibly propelled by rather than inhibited

by the collection of user data. As just one ex-

ample, several sites have proposed improving

matches by tapping into Netix and Pandora

user data to identify potential matches with sim-

ilar tastes in movies and music.

138

We cannot know what the online dating mar-

ket will look like in ve years, but one thing we

can condently predict is that it will be vastly dif-

ferent than what we see today.

VI: Conclusion

Compared to traditional markets, the Inter-

net has unleashed a boom of entrepreneurship

and competition which is arguably unrivaled

in history and denitely unrivaled in the mod-

ern era. We continue to see robust investment,

huge shares of revenue aimed at research and

development, and the successful entry of new

competitors—hardly characteristic of stagnant

markets under the grip of dominant players.

Investment in the European e-commerce

sector, the same market that European Com-

mission focused on in its e-commerce sector

inquiry

139

and its Statement of Objections in the

Google search investigation,

140

has exploded

137 Tinder, Inc. Press and Brand Assets, https://www.gotinder.

com/press (last accessed Apr. 8, 2016).

138  See, e.g., G. Clay Gallagher, Could Netix be the Best

Untapped Dating Network Online?, Popular Science (Sept. 14, 2015),

available at http://www.popsci.com/could-netix-be-best-untapped-

dating-network-online.

139 European Commission Press Release, Antitrust: Commis-

sion launches e-commerce sector inquiry (May 6, 2015), http://europa.

eu/rapid/press-release_IP-15-4921_en.htm.

140 European Commission Fact Sheet, Antitrust: Commission

sends Statement of Objections to Google on comparison Shopping ser-

vice (Apr. 15, 2015), http://europa.eu/rapid/press-release_MEMO-15-

4781_en.htm.

over the last few years. An analysis of a diverse

sampling of publicly traded European e-com-

merce companies showed that “the European

ecommerce market is in excellent health, with

strong levels of investment and innovation, and

with a marked upwards trend during the period

2012-2015” with investment increasing by a fac-

tor of 27 during that three year period.

141

According to the National Venture Capital

Association, investment in the software and In-

ternet sectors reached their highest level in over

a decade in 2014, with nearly $30 billion invest-

ed in over 2,800 separate deals.

142

Out of all in-

dustries surveyed by PricewaterhouseCoopers,

the “software and Internet sector” exhibited

the highest rate of growth in R&D spending in

2014.

143

According to a Battelle study featured

in R&D Magazine, the ICT sector was the largest

R&D spender in 2014 and accounted for one-

third of all R&D spending in the U.S. In particu-

lar, Battelle noted that “cloud computing and

technologies built on it will remain a major R&D

thrust for the foreseeable future.”

144

Furthermore, the Internet is expanding rapid-

ly, and the rest of the world is catching up to

the U.S. In 1995, there were 35 million Internet

users and the U.S. accounted for 61 percent of

that total. As of 2014, there were 2.8 billion In-

ternet users, and the U.S. only accounted for 10

141 European Ecommerce, Investment & Innovation A Com-

petitive Marketplace? 4 (Nov. 2015), http://cdn.ccianet.org/wp-content/

uploads/2015/11/European-Ecommerce_Investment_Study.pdf.

142 National Venture Capital Association Press Release, An-

nual Venture Capital Investment Tops $48 Billion in 2014, Reaching

Highest Level in Over a Decade, According to the Money Tree Report

(Jan. 16, 2015), http://nvca.org/pressreleases/annual-venture-capital-in-

vestment-tops-48-billion-2014-reaching-highest-level-decade-accord-

ing-moneytree-report.

143 Total R&D spending growth at large companies fell last

year to the second lowest rate in a decade, according to new Strategy&

study, Strategy& (Oct. 28, 2014), http://www.strategyand.pwc.com/

global/home/press/displays/proven-paths-to-innovation-success.

144 2014 Global R&D Funding Forecast, Battelle & R&D Mag-

azine, at 24 (2013), https://www.battelle.org/docs/tpp/2014_global_rd_

funding_forecast.pdf.

percent of that total. Whereas 13 of the top 15

publicly traded Internet companies were Amer-

ican in 1995, now nearly half of the top 20 pub-

licly traded Internet companies are from outside

of the U.S.

145

Instead of exhibiting signs of market failure

that call for extensive regulation, the Internet is

a thriving, enormously competitive ecosystem.

That is not to say that regulators should turn a

blind eye to deceptive or anticompetitive mar-

ketplace conduct—quite the opposite. If any-

thing, at this inection point two decades after

the commercialization of the Internet, it is criti-

cal for policymakers and regulators to better un-

derstand how these markets work and how they

differ from their brick-and-mortar predecessors.

Competition enforcement remains an import-

ant tool to deter and remedy anticompetitive

activity in the online sector, as in any other sec-

tor. Competition enforcers must, however, in-

corporate new economic understandings and

avoid presumptions about how markets work,

how rms compete, or why rms might engage

in conduct without precedent in a traditional

market. Enforcers should also consider employ-

ing what FTC Commissioner Maureen Ohlhau-

sen calls “regulatory humility” when faced with

conduct that is alleged to be anticompetitive

but that also has plausible benets for dynamic

competition or consumers, even if these bene-

ts cannot be quantied with specicity. Chill-

ing the incredible dynamism in these markets

should be the last thing any regulator or com-

petition enforcer should want to do.

145 Meeker Report, supra n. 20.

The Move to Smart

Mobile and Its

Implications for

Antitrust Analysis

of Online Markets

Hemant Bhargava, David S. Evans, and Deepa Mani

Abstract

Online markets have changed as a result

of people shifting massively from using person-

al computers and browsers to using techno-

logically powerful mobile devices and apps.

These changes cover leading online players,

consumer behavior, and products. The use of

smartphones and mobile apps, and the speed

of change, vary between countries and in par-

ticular between countries based on their stage

of development. Mobile app use is lower in

fast-growing countries, such as India, than in de-

veloped ones, such as the United States. Howev-

er, as smart mobile phones with mobile broad-

band connections become ubiquitous among

consumers in developing countries, mobile app

use in these countries is likely to leapfrog the use

of personal computers and browsers.

Bhargava is the Jerome J. and Elsie Suran Chair in Technol-

ogy Management, University of California at Davis; Evans is the Ex-

ecutive Director, Jevons Institute for Competition Law and Economics

and Visiting Professor, University College London and Lecturer, Uni-

versity of Chicago Law School; Mani is the Joint Executive Director,

Srini Raju Centre for IT and the Networked Economy and Assistant

Professor, Indian School of Business. The authors gratefully acknowl-

edge research support from Google.

As a result of the movement to smart mobile,

the analysis of markets that might have made

sense several years ago, does not today, and

will make even less sense several years hence.

The widespread adoption of smart mobile has

caused, and continues to result in, signicant

market disruption, including for incumbent In-

ternet-based companies, which are themselves

young compared to the traditional compa-

nies they disrupted. These dramatic and un-

predictable changes pose several issues for

antitrust. They show that antitrust analysis that

focuses on static markets is highly prone to er-

ror when it comes to dynamic online industries,

that authorities risk making assumptions during

investigations that are disproven by the markets

soon after they have brought charges or decid-

ed a case, and antitrust remedies are prone to

be ineffective or harmful because they are de-

veloped for markets during the investigation but

are radically different by the time the remedies

are implemented.

I. Introduction and Summary

The rapid growth of smart mobile devices

is changing online markets, and the nature of

competition, drastically, around the world. The

impact is apparent in developed countries.

The introduction of fast and capacious mobile

broadband networks in the early 2010s in coun-

tries such as the United States spurred the adop-

tion of smart mobile phones for online activities.

In the US more than 77 percent of adults had

a smart mobile phone as of August 2015.

The

average American spent more than an hour a

day using their smart mobile phones mainly for

online activities.

In the US, people spent 50 per-

1 comScore, “comScore Reports August 2015 U.S. Smart-

phone Subscriber Market Share”, https://www.comscore.com/Insights/

Market-Rankings/comScore-Reports-August-2015-US-Smartphone-

Subscriber-Market-Share.

2 ThisgureisacrossallUSresidents18andolderregardless

of whether they own a smart phone. Nielsen, “The Total Audience Re-

port:2015Q22015”,http://s1.q4cdn.com/199638165/les/doc_presen-

cent more time on their mobile phones than on

their personal computers as of mid 2015 and,

when on their mobile phones, they spent 87 per-

cent of their time using mobile apps compared

with just 13 percent of their time using a mobile

browser as of June 2015.

That shift from using

the personal computers and browser to using

mobile devices and apps continues at a rapid

clip.

Already, smart mobile phones have led to im-

mense changes in consumer behavior. People

have their smart mobile phones with them most

of the day and have come to depend on them

for shopping, communication, entertainment,

and more. These mobile devices are chang-

ing how people buy goods and services online,

and in physical environments, as reected by

the spread of ride-sharing apps globally; how

people communicate with each other as seen

in the widespread use of diverse messaging

apps; and how they consume entertainment

as people adopt streaming music and video

apps. For example, around 60 percent of the

visits American made to websites on November

26, 2015—Thanksgiving Day—were from mobile

devices.

The move to smart mobile has resulted in sig-

nicant changes in the competitive dynamics

of the online economy. The increasingly wide-

spread use of mobile apps has accelerated

the growth of other companies, from publicly

tations/2015/Total-Audience-Report-Q2-2015.pdf.

3 Nielsen Total Audience Report 2015Q2; https://www.com-

score.com/Insights/Presentations-and-Whitepapers/2015/The-2015-

US-Mobile-App-Report; http://es.slideshare.net/comScoreInsights/the-

us-mobile-app-report-comscore-53067374.

4 comScore, “2015 U.S. Mobile App Report”, https://www.

comscore.com/Insights/Presentations-and-Whitepapers/2015/The-

2015-US-Mobile-App-Report.

5 Based on 180 million visits to over 4,500 United States

websites on November 26, 2015, as reported by Adobe. Hiroko Tabu-

chi, “Black Friday Shopping Shifts Online as Stores See Less Foot

Trafc,” NewYork Times, November 27, 2015, http://www.nytimes.

com/2015/11/28/business/black-friday-shopping-shifts-online-as-

stores-see-less-foot-trafc.html?_r=0.

traded ones such as Facebook, which secures

78 percent of its advertising revenue from mo-

bile,

to startups such as Uber that rely entirely

on mobile devices for delivering services to driv-

ers and riders. Meanwhile, in the last ve years

Apple has vaulted to preeminence.

In devel-

oped countries such as the US around two-thirds

of online activity on mobile devices takes place

on Apple’s mobile devices using its iOS mobile

software platform. Meanwhile, companies that

were considered central to the online econo-

my a few years ago, such as the Yahoo ad-sup-

ported web portal, have struggled to make the

transition from the desktop to mobile.

The story in developing countries will be simi-

lar as they become wealthier and deploy faster

mobile broadband networks. The penetration

and adoption of advanced technologies is

growing rapidly and will soon reach the critical

levels necessary for igniting the smart mobile

ecosystem. In India, for example, only 14 per-

cent of adults have smart mobile phones.

But

this number represents a 121 percent increase

between 2013 and 2015.

Only 5.5 percent of

households were served by mobile operators

that have fast enough broadband for most on-

line activities in 2014, but that too is increasing

rapidly—72 percent from 2014 to 2015.

Indian

6 Based on data for 2015 Q3. Facebook Inc., “10-Q for Period

Ending September 30, 2015,” p. 40.

7 Anuj Srivas, “The Rise and Rise of Apple,” The Hindu,

February 1, 2015, http://www.thehindu.com/business/Industry/the-rise-

and-rise-of-apple/article6845491.ece.

8 PewResearchCenter,“InternetSeenasPositiveInuence

on Education but Negative on Morality in Emerging and Developing

Nations,” March 2015, http://www.pewglobal.org/2015/03/19/1-com-

munications-technology-in-emerging-and-developing-nations/.

9 Ambika Choudhary Mahajan, “Worldwide Active Smart-

phone Users Forecast 2014-2018L Niew than 2 Billion by 2016,”

DaazeInfo, December 18, 2014, http://dazeinfo.com/2014/12/18/world-

wide-smartphone-users-2014-2018-forecast-india-china-usa-report/.

10 Broadband Commission, “The State of Broadband 2015:

Broadband as a Foundation for Sustainable Development,” September

2015, http://www.broadbandcommission.org/documents/reports/bb-an-

nualreport2015.pdf; Broadband Commission, “The State of Broadband

2014: Broadband for All,” September 2014, http://www.broadband-

commission.org/documents/reports/bb-annualreport2014.pdf.

telecom operators are expected to increase

high-speed (3G and 4G) capacities over the

next 5 years at a compounded annual growth

rate of 125 percent.

Technology entrepre-

neurs and businesses are innovating towards

this new reality. Technology start-ups such as

OlaCabs (ride-sharing), Snapdeal (online re-

tail), near.in (at-home fulllment of local goods

and services), and Paytm (nancial payments)

are examples of recent startups that exploit the

capabilities of connected mobile devices to

deliver services and goods efciently to Indian

consumers.

The online economy in India and other de-

veloping countries will look more like the US and

other developed countries but with an import-

ant difference: the ratio of mobile technologies

and social computing to traditional comput-

ing with PCs and xed broadband is likely to

exceed the ratio in developed, more prosper-

ous, economies. This is because less developed

countries have exhibited very low penetration

of personal computers and xed broadband

connections due to the high cost of these tech-

nologies and their dependence on unreliable

electricity and xed Internet connectivity. Only

11 percent of Indian adults have personal com-

puters, for example, compared to 14 percent

that have smart mobile phones and 81 percent

that have smart or feature phones.

In contrast, mobile devices are available

over a wide spectrum of price points stretching

to the very inexpensive, are untethered from

the burden of continuous electric power and

xed-line Internet connectivity, and universal

mobile service can be provided without large

scale disruptions to civic infrastructure. For in-

stance, the online shopping site Snapdeal.com

11 Indian Tower Industries,” The Future is Data,” Deloitte In-

dia, June 2015, p. 28.

12 PewResearchCenter,“InternetSeenasPositiveInuence

on Education but Negative on Morality in Emerging and Developing

Nations,” March 2015, http://www.pewglobal.org/2015/03/19/1-com-

munications-technology-in-emerging-and-developing-nations/.

lists the 4G-capable Moto E phone running the

Android 5 operating system for Rs. 6,500 (about

half the price of the most inexpensive desk-

top-PC system), and most mobile operators

provide voice and data service for a couple

hundred rupees per month.

As a result, smart

mobile devices will leapfrog personal comput-

ers in developing countries. In a few years, peo-

ple in India, and other fast-growing countries,

will rely relatively more on smart mobile phones,

and mobile apps, than on personal computers

and browsers than people in the US and oth-

er developed countries. Respected analysts

and consulting rms forecast that, in India, the

percent of adults with smart mobile phones will

reach 22 percent in 2018 from the current 13

percent today.

The growing size and signicance of the on-

line economy has attracted the interest of com-

petition authorities, particularly over the role of

large multi-national online platforms. The dra-

matic and continuing shift by online consum-

ers and businesses to smart mobile devices has

important implications for the analysis of online

markets that would be prudent for competition

authorities and courts to consider. The analysis

of market denition and market power needs

to account for rapid dynamic change result-

ing from the simultaneous change in consumer

behavior, development of new technologies,

entry of new players, and integration of online

and ofine markets.

The enormity of these changes is apparent

just by comparing the state of online compe-

tition in 2015 versus 2005 in a developed coun-

try like the US. Ten years ago, Google was the

leading provider of online search on the web,

Microsoft controlled desktop computing, four

13 SnapDeal.com, http://www.snapdeal.com/. A rupee equals

about US0.015 or 1.5 cents as of November 29, 2015.

14 Smashing Boxes, “The Smartphone’s Impact on Social

Change,” January 30, 2015, http://smashingboxes.com/ideas/the-smart-

phones-impact-on-social-change.

telecommunication giants exerted signicant

control over most aspects of mobile phones,

Facebook was a nascent social network used

mainly in colleges, and Amazon was an online

retailer. Today, search has migrated to online

marketplaces, with Amazon serving as a search

engine in front of a giant electronic mall, and

to social networks, where Facebook is greatly

sought after for online advertising because it

enables social search and recommendations.

Apple and Google have become signicant

players in the mobile ecosystem, and mobile

phones have eroded the sales of PCs and Mic-

rosoft Windows. Amazon has become a prom-

inent vendor of cloud computing resources

used by many of the websites and apps behind

the online economy.

Given the pace of change it is likely that the

state of competition will be vastly different in

2020 than in 2015. Changes are likely to occur

even more rapidly in fast-growing places like

India. While India has lagged the US and oth-

er developed nations in adoption and pene-

tration of landline phones, xed broadband

Internet, and PCs—the old technologies—it

has leapfrogged into mobile phones, mobile

high-speed Internet, and mobile computing,

suggesting that these newer technologies will

be even more consequential, than they have

been in the US, in shaping competition in the

online economy. Consequently, the analysis of

market denition and market power analysis

need to be less rigid, analyze a broader range

of competitive dynamics, and be more forward

looking.

The move to smart mobile, and the disrup-

tions that shift is causing to the online economy,

create four implications for antitrust analysis.

1) Changes in consumer behavior, on-

line entry based on mobile apps, and

increased competition between mo-

bile app-centric and website-brows-

er centric businesses, lead to crossing

and overlap between previously-sep-

arate markets, and are likely to re-

duce the extent to which online pro-

viders possess market power.

2) Rapid changes in consumer behav-

ior and online entry increase the like-

lihood of making mistakes in market

denition and market power analysis.

It has become increasingly difcult to

predict the future even a few years

ahead.

3) The rapid and unpredictable shifts

in competitive dynamics, and tech-

nologies, caused by the shift to smart

mobile make it more difcult to de-

sign remedies, which are effectively

shooting at a moving target.

4) There is a greater likelihood of rem-

edies having negative unintend-

ed consequences by, for example,

limiting competition by incumbents

against fast-moving entrants who

quickly emerge as powerhouses. That

is particularly so during these times of

intense disruptive innovation resulting

from the move to using mobile apps.

As the online economy produces innovative

new technologies, services, and business mod-

els, spurred by the move to smart mobile, it is

vital that policy and antitrust analysis account

for these four implications, in order to ensure

that these innovations continue to improve the

functioning of society, business, and the econ-

omy.

This paper has four sections including this

Introduction. Section II documents the growth

of smart mobile and its impact on online and

ofine businesses and on consumer behavior.

It focuses mainly on what’s happened in the

US and other developed countries. Section III

presents a detailed analysis of how smart mo-

bile is affecting consumers and businesses in

the fast-growing Indian economy. India exem-

plies countries in which smart mobile is likely to

leap frog traditional Internet industries based

on using a PC-based browser to access web-

sites. Section IV examines the implications of

disruptive innovations surrounding smart mobile

for antitrust analysis.

II. How Smart Mobile Has Changed

Online and Ofine Businesses and

Consumer Behavior

During the 2000s mobile feature phones

became ubiquitous around the world. They

changed how people communicated with

each other by untethering phones from xed

lines, by providing text messages, and by mak-

ing it easy to take and share photos. Initially,

these devices primarily served voice communi-

cation and a smattering of other features such

as text messaging, music, and FM radio. This

limited role of mobile devices was transformed

around 2005 with the development of “smart”

phones. Smartphones employ a full-blown op-

erating system and have essentially unlimited

capabilities, because they enable end-us-

ers to install third-party application programs

(or apps). Leveraging Internet standards and

cloud computing, these small portable devic-

es imported the full power of modern comput-

ing, through easy-to-use apps connected with

powerful computers, software, and data over

the Internet. Building on that foundation the

rapid spread of smartphones starting around

2010 is having profound changes in every facet

of the online economy.

A. Why the World is Moving Rapidly to

Smart Mobile

Mobile computing is characterized by a busi-

ness-technology-ecosystem. A high level of

adoption and usage requires a combination

of a massive user base, affordable but highly

capable hardware, fast data networks with af-

fordable access plans, and a large collection

of useful services and applications. Four devel-

opments have resulted in the widespread use of

mobile devices for online activities. In the 2000s,

mobile broadband technologies were devel-

oped that made it possible for mobile network

operators to build mobile broadband networks

that were fast enough for conducting online

activities. These technologies included meth-

ods for making more efcient use of the wireless

spectrum for transmitting data and chip tech-

nologies for mobile handsets that worked with

the new mobile broadband standards and pro-

vided powerful computing devices. Although

these mobile broadband networks didn’t be-

come widely available in developed countries

until after 2010 it was apparent earlier in the

2000s that they would be forthcoming.

That anticipation led to the second major

development. A number of technology com-

panies started investing in developing mobile

phones that could provide online access as

well as make basic phone calls. This required

developing mobile handsets, mobile operating

systems that could run those handsets and sup-

port applications, and applications that could

make the handset particularly useful. Apple

adopted a vertically integrated approach in

which it produced the mobile operating system

(iOS) and handset (iPhone). Google focused

on developing an open source operating sys-

tem (Android), providing a standard hardware

and software framework for third-party handset

makers, and organizing the Open Handset Alli-

ance (OHA) to produce mobile phones. Both

approaches worked well. Apple launched the

iPhone, based on the iOS operating system in

June 2007. To help kick off the Android ecosys-

tem, Google, working with a HTC, introduced

the HTC Dream phone based on the Android

operating system in October 2008.

Mobile apps made available through “app

stores” were the foundation for the third de-

velopment. Apple, Android, and other mobile

operating system providers made it possible for

developers to write apps that used the features

and functions of the phone, including loca-

tion-based services, and availed themselves of

a connection to the Internet that was usually

on so long as the phone was turned on. These

mobile apps also enabled people to consume

content on the app ofine as well as online. The

mobile operating system vendors then created

“app stores” to provide a convenient way for

developers to distribute apps to users and for us-

ers to obtain these apps. Consumers and devel-

opers saw the iOS and Android-based devices

as superior to existing offerings from Blackberry,

Microsoft, and Symbian. The positive feedback

between consumer adoption and app devel-

opment propelled Apple and Android mobile

adoption. By June 2015, there were around 1.5

million apps for each of these platforms.

Con-

sumers had downloaded more than 100 billion

apps by June 2015.

The fourth development was cloud com-

puting whose power goes hand-in-hand with

the development of faster mobile broadband.

There are now 5.3 million server computers con-

nected to the Internet.

Companies own or

lease servers, on which they store the content

that they make available to their customers and

15 David Curry, “Apple iOS App Store Reaches 1.5 Million

Apps,” June 15, 2015, http://www.itproportal.com/2015/07/15/ap-

ple-app-store-milestone/; AppBrain, http://www.appbrain.com/stats/

number-of-android-apps; Jose Maria Delos Santos, “Top 5 Best Native

Apps for Project Management,” project-management.com, August 18,

2015, http://project-management.com/top-5-best-native-android-apps-

for-project-management/.

16 Harrison Weber, “Apple’s App Store Passes 100B Down-

loads and $30B Paid to Developers,” VentureBeat, June 8, 2015, http://

venturebeat.com/2015/06/08/apples-app-store-passes-100-billion-

downloads/.

17 Netcraft, “June 2015 Web Server Survey,” http://news.net-

craft.com/archives/2015/06/25/june-2015-web-server-survey.html.

others. Together, these servers, which sit on the

edge of the physical network of networks, are

called the “cloud.” They provide the computa-

tional resources and data for what people do

on their personal computer and smart phones.

As Internet speeds have increased, there is little

difference between using an app on a mobile

device connected to the cloud and using a

software application installed on a hard drive

on a PC. Lightweight apps, many of which can

t on a smart phone, can function as front-ends

to elaborate software systems with extensive

databases and deep algorithmic search. For

instance, ride-sharing services such as Uber use

large amounts of historical data (such as about

trafc patterns and sharing patterns) as well as

real-time data (such as about trafc conditions

and the location and preferences of riders) as

the fuel for intelligent algorithmic search and

optimization programs that produce ride-shar-

ing allocations in real-time.

The rate of adoption of smart mobile devices

in the US accelerated with the rollout of fast-

er mobile broadband in the US. The number of

mobile broadband subscriptions with speeds

of 256 Kbps or more per inhabitant increased

from 65.5 in 2011 to 97.9 in 2014.

Over those

same years, Pew Research Center reports

that the percent of people 18 and older with

smartphones increased from 35 percent to 64

percent.

Likewise, comScore reports that the

18 This number overstates the penetration rate of mobile broad-

band since some inhabitants have more than one mobile broadband

subscription. Broadband Commission, “The State of Broadband 2015:

Broadband as a Foundation for Sustainable Development,” September

2015, http://www.broadbandcommission.org/documents/reports/bb-an-

nualreport2015.pdf; Broadband Commission, “The State of Broadband

2014: Broadband for All,” September 2014, http://www.broadband-

commission.org/documents/reports/bb-annualreport2014.pdf; Broad-

band Commission, “The State of Broadband 2012: Achieving Digital

Inclusion for All,” September 2012, http://www.broadbandcommission.

org/documents/reports/bb-annualreport2012.pdf.

19 Aaron Smith, “Smartphone Adoption and Usage,” Pew Re-

search Center, July 11, 2011, http://www.pewinternet.org/2011/07/11/

smartphone-adoption-and-usage/; Aaron Smith, “U.S. Smartphone Use

in 2015,” April 1, 2015, http://www.pewinternet.org/2015/04/01/us-

smartphone-use-in-2015/.

number of people 13 and older with smart-

phones increased from 98 million in the three

months ending December 2011 to 182 million

in the three months ending December 2015.

Figure 1 shows the rapid growth between 2007

and 2015 in the percent of American adults

with a smart mobile phone.

Figure 1: Evolution of US Smartphone Owner-

ship and Access to Mobile Broadband

Sources: Pew Research Center, “Device Ownership over Time,” July

2015, http://www.pewinternet.org/data-trend/mobile/device-owner-

ship/.

Less prosperous, but now rapidly growing,

countries are following a similar path. Until

now, although mobile phones have become

ubiquitous in developing countries, relatively

few consumers use smart mobile phones—only

about 30 percent of Indians who have mobile

20 comScore, “comScore Reports December 2011 U.S. Mo-

bile Subscriber Market Share,” February 2, 2012, https://www.com-

score.com/Insights/Press-Releases/2012/2/comScore-Reports-Decem-

ber-2011-US-Mobile-Subscriber-Market-Share; comScore, “comScore

Reports December 2014 U.S. Smartphone Subscriber Market Share,”

February 9, 2015, https://www.comscore.com/Insights/Market-Rank-

ings/comScore-Reports-December-2014-US-Smartphone-Subscrib-

er-Market-Share.

phones have smartphones.

Many of the peo-

ple who do have smart mobile phones do not

use them much for online access because their

mobile network operators (MNOs) do not offer

sufciently fast, or cheap, mobile broadband.

For instance, as 2014 Q4, the slower technology

dominated data networks in India; high-speed

data networks accounted for only 11 percent

of unique mobile phone connections in India vs.

85 percent in the US.

In India, one megabyte

per second costs around $61, which makes it

very expensive for the average person.

Three phenomena in historically poorer but

rapidly growing countries, however, are chang-

ing this situation rapidly. First, the MNOs are roll-

ing out fast mobile broadband networks across

those countries and the cost of mobile broad-

band is declining for their residents. For mobile

operators in India, cut-throat competition in

the voice market, where a preference for pre-

paid plans has driven the cost of calling below

50p per minute, provides an imperative to in-

vest heavily in high-speed data networks and

promote smartphones for data-intensive tasks

such as entertainment, shopping and multime-

dia communication. In India, for example, ac-

tive mobile broadband subscriptions increased

from 1.9 percent in 2011 to 5.5 percent in 2014

while telecom operators like MTS announced a

price reduction of about 33 percent of its mo-

bile broadband tariffs.

21 Kishalaya Kundu, “HSBC: India Surpasses China in Pre-

mium Smartphone Sales,” Android Headlines, July 11, 2015, http://

www.androidheadlines.com/2015/07/hsbc-india-surpasses-china-pre-

mium-smartphone-sales.html.

22 The slower networks are 2G and the faster ones are 3G and

4G/. GSMA Intelligence, https://gsmaintelligence.com/data/.

23 Darrell M. West, “Digital Divide: Improving Internet

Access in the Developing World through Affordable Services and

Diverse Content,” Center for Technology Innovation at Brookings,

February 2015, http://www.brookings.edu/~/media/research/les/

papers/2015/02/13-digital-divide-developing-world-west/west_inter-

net-access.pdf.

24 Broadband Commission, “The State of Broadband 2015:

Broadband as a Foundation for Sustainable Development,” September

2015, http://www.broadbandcommission.org/documents/reports/bb-an-

80%

70%

60%

50%

40%

30%

20%

10%

May-11

Aug-11

Nov-11

Feb-12

May-12

Aug-12

Feb-13

May-13

Aug-13

Nov-13

Feb-14

May-14

Aug-14

Nov-14

Feb-15

May-15

Second, as average incomes increase, as

the price of smart mobile phones decline, and

as mobile broadband become cheaper and

faster, more people are adopting smart mobile

phones. Third, with higher incomes and lower

rates for mobile broadband, smartphone users

have increased their consumption of mobile

apps, data services, and mobile browsing.

B. How Mobile Changes Online Access

Before smart mobile devices became wide-

spread people used their personal computers

to engage in online activities. In the US, and

other developed countries, local cable and

telecom companies extended wires to house-

holds and businesses that connected these

households and businesses to the network of

networks that comprise the physical Internet. By

2010, about 23.7 percent of American house-

holds had xed broadband connections with

speeds of 3 Mbps or more.

People then used

a browser on their personal computer to access

websites from personal computers at home or

work. Starting with the birth of the commercial

Internet in the mid 1990s online businesses de-

veloped websites that people could access

from multiple browsers.

In retrospect there were two major limitations

to the “PC-xed broadband-browser” model.

The rst was the costs. The average price of a

personal computer in 2010 with a typical suite of

consumer software was $615 and the average

annual cost of a xed broadband connection

nualreport2015.pdf; Broadband Commission, “The State of Broadband

2012: Achieving Digital Inclusion for All,” September 2012, http://

www.broadbandcommission.org/documents/reports/bb-annualre-

port2012.pdf; Tech2 News, “MTS Drops Mobile Broadband Tariffs by

upto33Percent,”October6,2014,http://tech.rstpost.com/news-anal-

ysis/mts-drops-mobile-broadband-tariffs-33-percent-236836.html.

25 National Telecommunications and Information Administra-

tions (US), “U.S. Broadband Availability: June 2012 – June 2012,” May

2013, https://www.ntia.doc.gov/les/ntia/publications/usbb_avail_re-

port_05102013.pdf.

was $828.

As of 2010, 23 percent of American

households didn’t have a personal computer

and 29 percent did not use the Internet primar-

ily because they were poor households who

couldn’t afford it.

Other developed countries

were in a similar position. PC and broadband

penetration were much worse in poorer coun-

tries. In 2010, only 6.1 percent of Indian house-

holds had personal computers and only 8.5

percent had access to the Internet.

In fact,

except for the wealthiest of citizens, the online

revolution hadn’t really touched the billions of

people on earth who weren’t fortunate to live

in one of the developed countries.

The second limitation was the fact that when

it came to accessing the online world, people

were tethered to places that had xed broad-

band connections. Increasingly, locations,

such as airports and coffee shops, had “WiFi”

that provided Internet connectivity in their lo-

cations. But people generally didn’t have any

connection to the Internet when they were out

and about or even when they were wandering

around their homes.

Smart mobile phones with mobile broad-

band access changed this situation in funda-

mental ways. People got the equivalent of an

Internet-connected personal computer in a tiny

package that they could carry with them all

the time. That increased both the time during the

day, and physical places, where people could

engage in online activity. The proliferation of mo-

26 Ben Worthen, “Rising Computer Prices Buck the Trend,”

Wall Street Journal, December 13, 2010, http://www.wsj.com/articles/

SB10001424052748704681804576017883787191962; Federal Com-

munications Commission (US), “International Broadband Data Report

(Second),” May 20, 2011, https://www.fcc.gov/reports/internation-

al-broadband-data-report-second.

27 Thom File and Camille Ryan, “Computer and Internet Use

in the United States: 2013,” American Community Survey Reports,

ACS-28, US Census Bureau, November 2014, http://www.census.gov/

content/dam/Census/library/publications/2014/acs/acs-28.pdf.

28 OECD, Key ICT Indicators, 2012, http://www.oecd.org/

sti/ieconomy/Final_6.a_PC%20Households_2012.xls; Internet World

Stats, “India,” http://www.internetworldstats.com/asia/in.htm.

bile apps supported by the mobile operating sys-

tems provided innovative ways for developers to

provide services to consumers. Location-based

technologies enabled developers—through mo-

bile apps or websites accessed from the brows-

er—to provides services based on where the

consumer was or wanted to go

The adoption of smart mobile phones has re-

sulted in a high density of Internet connections

throughout physical space. Most of the people

moving around Manhattan by foot, car, or sub-

way, for example, now have a smart mobile

device and are online either actively, because

they are using an app or browsing, or passively

because they can receive alerts. A few years

ago most people could go online only if they

were sitting at a computer at their home or, in

some cases, their ofce.

Moreover, with new methods and standards

for machine-to-machine communication, enor-

mous amounts of computation and communi-

cation occur without the active participation of

the user. For instance, consider a smartphone

user who has a Nike Running app, a Facebook

account, online calendar, and a Picasa photo

account: for this user, a jog can automatical-

ly lead to a Facebook post, a camera picture

is instantly communicated to a Picasa group

or Instagram, and a calendar entry causes a

mapping service to communicate the best

travel option for the user’s next meeting.

Not surprisingly, enabling most people to be

connected most of the time, and using a pleth-

ora of apps, has had huge ramications for

people and businesses.

C. The Impact of Smart Mobile on How

People Spend Their Time

Now that people can always be online, they

go online more than they used to. Figure 2

shows the minutes a day that Americans spent

online using personal computers or mobile de-

vices from 2008 to 2015. The total number of

minutes more than doubled from 150 minutes in

2008 to 313 in 2015. The proportion of this time

spent on mobile increased from 12.7 percent in

2008 to 54.6 percent in 2015.

Figure 2: Daily Time Spent per Media Platform in

Minutes, United States

Source: David Pakman, “May I Have Your Atten-

tion, Please?” Medium, August 10, 2015, https://

medium.com/life-learning/may-i-have-your-at-

tention-please-19ef6395b2c3?curator=Medi-

aREDEF#.2rwedr27o.

Note: This data is consistent with that from oth-

er sources, including Nielsen, “The Total Audi-

ence Report: 2015Q2 2015”, http://s1.q4cdn.

com/199638165/files/doc_presentations/2015/

Total-Audience-Report-Q2-2015.pdf; eMarketer,

“Mobile Continues to Steal Share of US Adults’

Daily Time Spent with Media,” April 22, 2014,

http://www.emarketer.com/Article/Mobile-Con-

tinues-Steal-Share-of-US-Adults-Daily-Time-

Spent-with-Media/1010782.

People typically rely on apps when they use

their smartphones. The typical smartphone user

in the US used about 27 apps per month as of

the 2014Q4.

Table 1 shows the breakdown

of how people spend their time. Important-

ly, smartphones are really not used much as

phones in the US. Making phone calls and send-

ing text messages account for only 16 percent

of the time. Of the time people spend online, 90

percent is spent using mobile apps and only 10

percent is spent accessing websites using their

browser.

The story is similar for smart mobile

phone users in India.

Table 1: Mobile Time Usage in the United States

and India, 2015

United

States

India

Utility Features

and Apps

33% 20%

Multimedia Apps 16% 12%

Games 11% 9%

Other 10% 9%

Web Surng 9% 10%

Messaging 9% 2%

Phone Calls 7% 12%

Chat and VOIP 4% 26%

Source: Nielsen Informate, “International Smartphone Mobility Re-

port,” March 2015, p. 14.

D. Market Disruptions Resulting from the

Move to Smart Mobile

The shift of the time spent online to mobile

devices, and the shift from using browsers to

apps has, not surprisingly, disrupted many as-

pects of the online economy. Communication

29 Nielsen, “So Many Apps, So Much More Time for Enter-

tainment”, 2015, http://www.nielsen.com/us/en/insights/news/2015/so-

many-apps-so-much-more-time-for-entertainment.html.

30 Simon Khalaf, “Seven Years into the Mobile Revolution:

Content is King … Again,” Flurry Insights, August 26, 2015, http://

urrymobile.tumblr.com/post/127638842745/seven-years-into-the-mo-

bile-revolution-content-is.

among people is moving swiftly to messaging

apps, away from voice, text messaging, email,

and browser-based methods. According to

data from comScore, mobile apps accounted

for 52 percent of the time spent using digital

media while desktop-based digital media con-

sumption took 40 percent (mobile web brows-

ing took up the remaining 8 percent).

People

increasingly engage in social networking from

apps on their mobile devices rather than using

a browser from their personal computers. A re-

port by BI Intelligence conrms that 60 percent

of social media time is spent, not on desktop

computers, but on smartphones and tablets.

Commerce is moving from people using their

browsers on their personal computers in xed

locations to search, discover, and buy to using

apps on mobile devices often while they are

making shopping trips or just happen to be out

and about.

The changes taking place in shopping illus-

trate how quickly and dramatically the move

to smart mobile is changing the behavior of

people and businesses. Smart mobile devices

enable people to blend physical and online

shopping and retail stores are reacting to this

change in consumer behavior. People can use

specialized apps such as AroundMe to suggest

where they should shop given their current lo-

cations, using retailer apps get notications

from stores they’re near, and compare prices

using apps such as PriceGrabber.

As a result, a signicant amount of the activ-

ity surrounding the “path to purchase” for con-

sumers is moving to mobile devices in devel-

oped economies such as the US. That is seen in

31 Sarah Perez, “Majority of Digital Media Consumption Now

Takes Place in Mobile Apps,” TechCrunch, August 21, 2014, http://

techcrunch.com/2014/08/21/majority-of-digital-media-consumption-

now-takes-place-in-mobile-apps/; comScore, “The U.S. Mobile App

Report,” August 21, 2014, http://www.comscore.com/Insights/Presen-

tations-and-Whitepapers/2014/The-US-Mobile-App-Report.

32 http://www.businessinsider.com/social-media-engage-

ment-statistics-2013-12.

a number of ways. Many Americans, and most

“millennials”—those who are between roughly

20 and 35 in 2015—use smart mobile phones as

part of their process of buying goods. According

to Thrive Analytics, 90 percent of adults, and 97

percent of millennials, use smart mobile phones

as part of their typical shopping practices.

Advertisers are directing a signicant amount

of their spending to mobile devices. Facebook

earned 14 percent of its global advertising rev-

enue from mobile in 2012Q3.

As of 2015Q3 it

earned 78 percent of its advertising revenue on

mobile.

Consumers aren’t just using mobile to

help in the search for and discovery of things

to buy. Increasingly, they are consummating

their purchases on mobile devices. Americans

made 57 percent of their online purchases from

mobile devices in 2014 compared with virtu-

ally none before 2010.

As noted above, on

Thanksgiving Day, November 26, 2015, around

60 percent of US website visits were made from

mobile devices in the US.

Whereas online shopping was previously

based on computers and web browsers, online

commerce on mobile devices is largely con-

ducted via apps that enable people to buy

conveniently online or at physical stores. Sever-

al social networking and communication plat-

forms that are used predominantly on mobile

devices have transitioned from helping people

discover and evaluate products to enabling

people to buying products online. Twitter, for

example, has integrated a “Buy Now” button

33 http://www.thelsa.org/main/pressreleases/mobile-use-now-

surpasses-pcs-when-searching-for-lo-3080.aspx; http://www.defyme-

dia.com/2014/09/11/millennials-using-smartphones-store/.

34 Facebook Inc., “10-Q for Period Ending September 30,

2012,” p. 27.

35 Facebook Inc., “10-Q for Period Ending September 30,

2015,” p. 40.

36 http://www.pcmag.com/article2/0,2817,2474217,00.asp.

37 Hiroko Tabuchi, “Black Friday Shopping Shifts Online as

StoresSeeLessFootTrafc,”NewYorkTimes,November27,2015,

http://www.nytimes.com/2015/11/28/business/black-friday-shopping-

shifts-online-as-stores-see-less-foot-trafc.html?_r=0.

as part of an effort to enable people to “have

even more opportunities to discover and pur-

chase products from the brands they love on

Twitter.”

Other platforms that are taking sim-

ilar approaches include Facebook, Facebook

Messenger, Pinterest, WeChat, and SnapChat.

These changes in consumer shopping be-

havior are resulting in a revolution in retail.

Retail stores are developing “omnichannel”

approaches that integrate physical stores, mo-

bile apps, and websites and provide consum-

ers with multiple choices of how to shop and

buy.

Because consumers have more and bet-

ter ways to make their purchase decisions they

increasingly go to a store only after making a

decision to buy something there. Since they en-

gage in less comparison shopping foot trafc

to stores is declining. Retailers are reducing the

size of stores and reorganizing their businesses

to cater to this change in behavior. More re-

tailers are letting consumers order online—often

with mobile—and pick up in store.

The latest step in this revolution is “hyperlo-

cal” retail, where physical and online stores aim

to provide instant gratication through same-

day delivery, in some cases within a few hours

of order placement. Hyperlocal retail is growing

38 Nathan Hubbard, “More Ways to Sell Directly on Twitter,”

Twitter Blog, September 30, 2015, https://blog.twitter.com/2015/more-

ways-to-sell-directly-on-twitter.

39 http://outtosea.co.uk/articles/social-commercetrends-2015/;

http://a16z.com/2015/08/06/wechat-china-mobile-rst/

40 See Evans and Schmalensee, Matchmakers: The New Eco-

nomics of Multisided Platforms, Chapter 12, forthcoming 2016.

41 Jennifer Kasper (Group Vice President, Digital Media and

Multi-Cultural Marketing), “Macy´s Goes Omni-Channel,” interview,

ThinkWithGoogle, October 2014, https://www.thinkwithgoogle.com/

interviews/macys-goes-omni-channel.html; Angela Ahrendt (CEO of

Burberry), “Burberry’s Digital Transformation,” interview by Capgem-

iniConsulting,2012,https://www.capgemini.com/resource-le-access/

resource/pdf/DIGITAL_LEADERSHIP__An_interview_with_An-

gela_Ahrendts.pdf; Warby Parker, “Homepage,” accessed August

28, 2015, https://www.warbyparker.com/; Dennis Green, “Bonobos

Is Opening Retail Stores – But You Can´t Actually Take Any of the

Clothes Home,” Business Insider, July 16, 2015, http://www.busines-

sinsider.com/bonobos-opened-a-store-where-you-cant-physically-buy-

anything-2015-7.

rapidly in both developed countries such as the

US and developing countries such as India, and

is fueled by high-speed supply chains, highly lo-

calized data, and location-awareness due to

mass adoption of mobile devices.

E. Smart Mobile and Competitive Dynam-

ics of the Online Economy

The move to smart mobile is one of the ma-

jor factors in a dramatic shift in the compet-

itive landscape in the online economy. These

changes in the competitive landscape are driv-

en by a profound shift that combines techno-

logical innovation and business strategy. With

respect to market denition and competition, a

crucial feature of smartphones is that they are

based on full-edged operating systems, on ac-

count of which consumers can install third-par-

ty apps at any time after acquiring the phone.

This shift to operating systems has destroyed the

pipe-model of mobile telephony that was prev-

alent until the launch of the iPhone, in which

value and features came to consumers through

a strictly linear chain comprising feature de-

velopers, phone manufacturers and mobile

operators, and the “walled-garden” model in

which the mobile operators maintained strict

control of what was on mobile phones on their

systems.

The replacement of this linear model,

and the walled garden, with a “platform” mod-

el has transformed the market, resulting in the

demise of phone makers that could not keep

up (such as Nokia), and drastically reducing the

once considerable power of mobile network

operators in many developed countries.

To get a sense of the change in competitive

dynamics it is useful to consider what has hap-

pened to several key players.

The most dramatic change is that Apple has

become one of the most valuable companies

in the world largely based on its sales of iPhones.

42 “Why Business Models Fail: Pipes vs. Platforms,” Wired

magazine, http://www.wired.com/insights/2013/10/why-business-mod-

els-fail-pipes-vs-platforms/

One could argue that Apple is really a manu-

facturer and, except for its digital music busi-

ness, not a signicant online player at all. That

misses the essence of what Apple is and why

it sells iPhones. The Apple iPhone, and its other

mobile devices, are valuable primarily because

they provide a platform for online activity. Ap-

ple’s mobile operating system, iOS, and its App

Store anchor a vast ecosystem of mobile ap-

plications. Although Apple accounts for fewer

smartphone sales than Android, iPhones tend

to be used much more for online activity while

Android phones tend to be used much more

for voice calls and text. In the US, for example,

where Apple accounts for only about 38 percent

of the 2015 smartphone installed base, its users

accounted for around 62 percent of the time

spent on mobile and 71 percent of spending on

mobile apps occurs on iOS-based devices.

Another remarkable change involves Face-

book. In 2010, Facebook was only six years old

and two years away from its IPO. It was only

two years before, in May 2008, that Facebook

rst accounted for more pageviews than MyS-

pace, which was the leading social networking

site during the mid 2000s. As of 2015, Facebook

provides three of the 10 most popular mobile

apps as measured by downloads by American

smartphone mobile media users.

Facebook it-

self is the most popular mobile app, Facebook

Messenger is the third most popular mobile

app, and Instagram, which Facebook owns, is

the ninth most popular. Altogether, Facebook’s

apps account for 13 percent of unique visitors

on mobile phones according to a report by For-

rester Research.

Facebook makes money by

selling advertising that reaches people who vis-

it its properties. According to eMarketer, Face-

book’s share of the US mobile advertising rev-

43 Canalys, “Worldwide Smartphone/Mobile Phone Installed

Base Forecasts (Consolidated),” June 2015; comScore, “Mobile Met-

rix,” April 2015; App Annie, “Store Intelligence,” June 2015.

44 https://www.comscore.com/Insights/Market-Rankings/

comScore-Reports-August-2015-US-Smartphone-Subscriber-Market-

Share,

45 http://www.cio.com/article/2943866/mobile-apps/face-

book-and-google-dominate-time-spent-with-mobile-apps.html,

enues is expected to grow from 18.5 percent

in 2014 to 20.3 percent by 2017. Meanwhile,

Google’s share of overall advertising revenues

is projected to drop from 37.0 percent to 31.7

percent in the same period.

Microsoft’s fortunes online have faded since

2010 although it remains an immensely prot-

able rm as a result of its licenses for its Windows

desktop and server operating system and its Of-

ce productivity app. In 2007, Microsoft was the

second largest provider of operating systems

for smart mobile phones with a 13.7 percent

share of smart mobile phones; this share under-

states its importance because the leader was

Symbian, which was not a good platform for

app developers. It was widely expected by an-

alysts that Microsoft would leverage its success

on the desktop to mobile. Eight years later, as

of July 2015, rather than being the leader Mic-

rosoft accounted for only about 2.7 percent of

all smart mobile phone subscribers.

Moreover,

Microsoft has virtually no presence as a mobile

app provider—its worldwide share of mobile

app downloads on Google Play and the iOS

App Store is only 1 percent for free apps and

3 percent for paid apps, and its share of total

mobile minutes in the United States for all of its

properties is only 1 percent.

Google remains a signicant online play-

er and highly valuable company. Its various

properties account for about 12 percent of the

time people spend on smart mobile phones.49

The company, however, faces a very different

competitive environment on app-based smart

46 http://blogs.wsj.com/cmo/2015/09/08/facebook-projected-

to-narrow-mobile-ad-gap-with-google-as-emarketer-reverses-forecast/.

47 https://www.comscore.com/Insights/Market-Rankings/

comScore-Reports-July-2015-US-Smartphone-Subscriber-Market-

Share.

48 App Annie, “Store Intelligence,” June 2015; comScore,

“Mobile Metrix,” April 2015.

49 Cio, “Facebook and Google dominate time spent with

mobile apps”, http://www.cio.com/article/2943866/mobile-apps/face-

book-and-google-dominate-time-spent-with-mobile-apps.html; http://

techcrunch.com/2015/06/22/consumers-spend-85-of-time-on-smart-

phones-in-apps-but-only-5-apps-see-heavy-use/#.jkexdnu:5bJy.

mobile devices than it faced on the web-brow-

er-based desktop. The heavy use of mobile

apps, together with the natural use of voice on

smart mobile phones, has opened up a new

battleground for search. Apple, Google, and

others are developing new methods of search

that can canvass the vast amount of content

being generated within mobile apps, which are

not indexed by existing search engines, and

new methods of interacting with the mobile

device to conduct searches.

With the release

of iOS 9 in September 2015, Apple redesigned

Spotlight to include search results from content

within apps, apparently in an attempt to steer

users towards apps and away from websites.

Facebook recently launched search across

postings on its properties, which are walled off

from search engines and now include many

mobile-based postings.

While predicting the future is quite hazardous,

the history of dynamic competition in the online

economy, the rapid move from the PC-brows-

er centric model to the smart mobile-app-cen-

tric model, and the surge of investment in mo-

bile-app based startups, all suggest that 2020

will look dramatically different from 2015.

That is especially likely in historically poorer

but now fast-growing economies. What’s hap-

pening in India, which we detail next, is hap-

pening in many developing countries to vary-

ing degrees.

50 Contently, “Google and Apple Are in a War for the Future of

Search”, https://contently.com/strategist/2015/08/17/google-and-apple-

are-in-a-war-for-the-future-of-search/; http://www.ibtimes.com/apple-

incs-ios-9-spotlight-app-search-could-make-google-obsolete-indexing-

content-210083.

51 Salvador Rodriguez, “Apple Inc.’s iOS 9 Spotlight In-App

Search could Make Google Obsolete by Indexing Content From Ev-

ery App You Own 9And Some You Don’t),” International Business

Times, September 17, 2015, http://www.ibtimes.com/apple-incs-ios-

9-spotlight-app-search-could-make-google-obsolete-indexing-con-

tent-2100836.

52 Nick Statt, “Facebook is Unleashing Universal Search

Across its Entire Social Network,” The Verge, October 22, 2015, http://

www.theverge.com/2015/10/22/9587122/new-facebook-search-all-

public-posts.

III. How Mobile Is Disrupting Online in

Fast-Growing Economies: the Case of

India

India’s Internet population of 243 million is

the third largest in the world, after China with

642 million and the U.S. with 280 million, as of

July 2014.

The number of Internet users in In-

dia is expected to register exponential growth,

reaching an estimated 500 million by 2017, mak-

ing it the second largest population of Internet

users. The economic impact of this growth is

signicant. In 2013, the Internet contributed 2.7

percent of India’s GDP (USD 60 billion).

It is es-

timated to increase to over 4 percent of GDP

by 2020 and employ nearly 22 million people.

Beyond the contribution to GDP, the impact

of the Internet in India is manifest in improved

quality of life and empowerment of the coun-

try’s citizens through greater and more diverse

information consumption, improved access to

government and essential services, and greater

transparency in the delivery of these services.

Indeed, the Government of India, through its

“Digital India” campaign, has identied provi-

sion of digital infrastructure, digital literacy of

citizens56, and digitization of services as key pri-

orities for the government in the coming years.

53 http://www.internetlivestats.com/internet-users-by-coun-

try/. Other sources give somewhat higher estimates for the number of

Internet users in India. Internet and Mobile Association of India (IMAI)

and KPMG, “India on the Go: Mobile Internet Vision 2017,” http://rtn.

asia/wp-content/uploads/2015/07/Report.pdf. It is estimated at over

340 million by December 2015 according to http://www.internetlives-

tats.com/internet-users/india/.

54 Boston Consulting Group and IMAI, “India@Digital.

Bharat: Creating a $200 Billion Internet Economy,” January 2015,

http://www.bcgindia.com/documents/le180687.pdf.

55 McKinsey, “Online and Upcoming: The Internet’s Impact on

India,” March 2013, http://www.mckinsey.com/~/media/mckinsey%20

ofces/india/pdfs/online_and_upcoming_the_internets_impact_on_in-

dia.ashx. .

56 The National Digital Literacy Mission (NDLM) of the Gov-

ernment of India envisions imparting IT training to nearly 5.2 million

persons in the country so as to enable them to actively and effectively

participate in the democratic and developmental process and also en-

hance their livelihood.

For India, though, online is now, and will be,

centered on mobile.

A. Role of the Mobile Internet

The mobile Internet has led the exponen-

tial growth in the online economy in India. As

shown in Figures 3a and 3b, wireless Internet

has leapfrogged the wireline Internet access.

Mobile devices are the dominant ramp to the

Internet in India with 34 percent of Indians ac-

cessing the Internet exclusively from mobile de-

vices in 2014. As of March 2014, the PC pene-

tration rate in India was 5 percent in contrast to

75 percent for mobile devices.

Not surprisingly,

industry esti

mat

suggest that mobile Internet

users comprised over 60 percent of the online

population in 2014 and are expected to com-

prise 70-80 percent of the online population in

2018, representing a CAGR of 27.8 percent for

the period 2014-2018.

Figure 3a: Growth in Wireline versus Wireless

Internet Connections

450

400

350

300

250

200

150

100

2013 2014 2015 2016 2017

Internet Connections in India (in millions), 2013-2017(P)

150

24 20 22 25

210

Wireline

Wireless

273

337

402

Source: Internet and Mobile Association of India (IMAI) and KPMG,

“India on the Go: Mobile Internet Vision 2017,” http://rtn.asia/wp-con-

tent/uploads/2015/07/Report.pdf.

57 Ericsson, “Ericsson Mobility Report,” June 2014, http://

www.ericsson.com/res/docs/2015/ericsson-mobility-report-june-2015.

pdf.

58 Avendus,“India’s Mobile Internet: the revolution has be-

gun”, 2013, http://www.avendus.com/media/1366/avendus_india_mo-

bile_internet.pdf.

Figure 3b:

Growth in Mobile Internet Subscriptions

Source: Brookings (India), http://www.brookings.edu/blogs/

techtank/posts/2015/03/18-mobile-technology-india.

1998

1999

2000

2001

2002

2003

2004

2005

2006

2007

2008

2009

2010

2011

2012

2013

Internet in India

Year

Mobile Internet

Subscriptions per 100 people

BROOKINGS

At the heart of the growth in the mobile Inter-

net population in India are two market factors,

notably, growth in affordable smartphones and

tablets, and improved performance of network

infrastructure of telecom operators at lower

costs of ownership. As shown in Figure 4, smart-

phone sales in India have had very high growth

over the last ve years, making the country the

third largest smartphone market in the world.

Although the smartphone penetration rate in

India is low at around 19 percent the average

annual growth rate in smartphone sales be-

tween 2013 and 2015 estimated at 121 percent

is very high.

The equivalent growth in China during the

same period is estimated at 31 percent. Re-

markably, annual smartphone shipments in-

creased almost ten-fold between 2012 and

2015. Helped by declining average selling pric-

es, increased competition and rising disposable

incomes, the Indian smartphone industry is likely

to sustain these high growth levels.

Figure 4: Growth in Smartphone Shipments,

2009-2014

59 http://www.bgr.in/news/smartphone-shipments-in-india-

grew-229-percent-in-q3-2013-idc/; http://www.digitimes.com/news/

a20151125PB200.html

120

100

2009 2010 2011 2012 2013 2014 2015(P)

Smartphone Shipments in India (millions)

2.2

5.5

10.1

15.1

41.1

114

Source: International Data Corporation (IDC), KPMG

In addition to increasing smartphone sales,

the upgrading of network infrastructure and the

improved affordability of data services of tele-

com operators are key contributors to growth in

mobile Internet in India. Government policies,

including the National Telecom Policy of 2012

that seeks to achieve rural teledensity of 100 by

2020 and permits foreign rms to own up to 100

percent of the equity of telecom rms, act as

catalysts for investments in network infrastruc-

ture, especially in rural areas. While growth in

rural connectivity is likely to occur through slow

2G technologies, the latter is on the decline in

urban areas as an increasing number of cus-

tomers migrate from slow 2G to faster 3G ser-

vices.

McKinsey’s Digital Consumer survey nds that

nearly half of mobile users want to upgrade to

a smartphone and a third of these will adopt

3G solutions giving them reasonably fast broad-

band. There were approximately 82 million 3G

subscribers in India at the end of 2014 but this

number is projected to grow at a CAGR of

61.3 percent to reach 284 million by the end of

2017.

The amount of 4G data trafc in India is

60 KPMG-FICCI M&E Industry Report, “The Stage is Set,”

2014, https://www.kpmg.com/IN/en/Topics/FICCI-Frames/Docu-

ments/FICCI-Frames-2014-The-stage-is-set-Report-2014.pdf.

also expected to grow at a CAGR of 176 per

cent from 2014 to 2019.

The increased adop-

tion of smartphones is correlated with growth in

mobile Internet as well as increased revenues

since the average smartphone user spends al-

most twice the amount that an average mobile

user does on mobile Internet.

B. Key Online Players

As these trends suggest, mobile has become

important for online activity in India. For in-

stance, nearly 42 percent of e-commerce sales

occur through mo-

bile devices in India

compared with 15

percent for the US. As

shown in Figure 5 be-

low, in 2014, leading

Indian e-commerce

companies, including

Flipkart and Snap-

deal, obtained more

than 70 percent of

their sales (measured

by “gross merchan-

dise value” (GMV))

from mobile devic-

es as opposed to

Chinese companies

Alibaba.com and

JD.com that derived

40-50 percent of their

sales from mobile de-

vices. Similarly, more

than 90 percent of

Facebook's Indian us-

61 Cisco Visual Networking Index, VNI Mobile Forecast

Highlights, 2014-2019, http://www.cisco.com/c/dam/assets/sol/sp/vni/

forecast_highlights_mobile/index.html#~Country.

62 McKinsey, “Online and Upcoming: The Internet’s Impact on

India,” March 2013, http://www.mckinsey.com/~/media/mckinsey%20

ofces/india/pdfs/online_and_upcoming_the_internets_impact_on_in-

dia.ashx.

ers

and 60 percent of Amazon’s Indian users

access the sites through mobile devices. The re-

cent launch and explosive growth of Paytm, a

digital payments company, is another catalyst

for increasing Internet and mobile commerce

in India.

The importance of the mobile Internet is re-

ected in strategies and investments of key

online companies that are idiosyncratic to the

Indian market. For example, leading online

fashion retailer Myntra has adoped an app-fo-

cused strategy, noting that 90 percent of trafc

and 70 percent of sales were coming from its

mobile app. Conducting business via mobile

63 http://www.bgr.in/news/90-of-facebooks-132-million-users-

from-india-come-from-mobile-phones/

64 http://www.afaqs.com/interviews/index.html?id=469_Over-

60-per-cent-of-our-trafc-comes-through-mobile-Manish-Kalra-Ama-

zon-India

Figure 5: Share of Mobile Trafc of Leading E-Commerce Players in India

apps enables the rm to leverage information

on the consumer’s location and social circle to

provide customized product recommendations

and timely promotions notications. Some rms

have given up on web commerce and have

decided to focus exclusively on conducting

business via apps. TruelyMadly, a dating and

matchmaking site, Tiny Owl, a food delivery

site, and RoomsTonite, an on-demand hotel

booking platform, are all examples of fast grow-

ing Indian online rms that have shut down their

websites, citing higher mobile trafc rates and

superior conversion rates for mobile Internet

customers in support of their strategy.

Firms that have not adopted an app-only

strategy have launched websites optimized for

the mobile environment, in general, and slow-

er 2G networks and areas with limited connec-

tivity, in particular. For instance, in November

2015, Flipkart launched Flipkart Lite, a website

optimized for mobile devices. The initiative was

quickly replicated by its competitor, Snapdeal

that launched Snap-lite. Similarly, Facebook

too has developed Facebook Lite, a mobile

website that uses less data, for India and other

emerging markets. Google meanwhile has de-

veloped ofine versions of YouTube and Maps

that people can use on their mobile phones

without consuming data and in places with

poor coverage. Browser companies too have

customized their offerings for the Indian mobile

environment. UC Browser, which has more than

50 percent share of browsers in India, incor-

porates a slew of features to compress data,

increase navigation speeds, and improve

download quality, all of which are aligned with

constraints in the Indian mobile environment, in-

cluding dropped connections, poor availability

and slow network speeds.

The importance of the mobile Internet in In-

65 Techstory, “Why Is Indian E-Commerce Adopting Mobile

Strategy & What It Means For Indian Consumers”, http://techstory.in/

app-only-strategy/.

dia is also seen in the mobile-focused business

models that characterize entrepreneurial ac-

tivity in the country. To illustrate, a sector that

has witnessed signicant start-up activity and

investment is hyperlocal businesses that rely on

geolocation awareness using the mobile phone

to enable local ofine services from anywhere,

anytime. In just over the past six months, over

$140 million has been raised by this class of busi-

nesses and over 28 funding deals closed.

Zopper, shopping marketplaces, PepperTap

and Grofers, grocery delivery rms, Swiggy, a

food delivery service, and UrbanClap, a ser-

vices marketplace, are all examples of start-ups

in the fast-growing hyperlocal space. Niche on-

line marketplaces such as those in the furniture

and decor category too are going hyperlocal

by partnering with ofine stores for a commis-

sion-based model rather than building a net-

work of designers, manufacturers and logistics to

create an online brand.

This category of busi-

ness models is but an illustration of mobile-rst

start-ups that are disrupting the traditional Inter-

net business in India. This transformation will be

facilitated by available of mobile-based digi-

tal payment services such as Paytm, which re-

ceived a $680 million investment from Alibaba

and Ant Financial in September 2015.

The rise of the mobile Internet, and the reduc-

tions in entry barriers has also made it easier for

domestic companies to challenge internation-

al ones. While it is too soon to know how com-

petition will play out in India Ola is challenging

Uber in the ride-sharing business and Flipkart

and Snapdeal are challenging Amazon.

C. Paths to Purchase in India

66 http://economictimes.indiatimes.com/small-biz/startups/

startups-like-urban-ladder-fab-furnish-goes-hyperlocal-partners-with-

ofine-stores-for-commission-based-model/articleshow/49574337.cms

67 http://economictimes.indiatimes.com/industry/banking/

nance/banking/alibaba-ant-nancial-invest-about-680-million-in-pay-

tm-up-stake-to-40/articleshow/49148651.cms

The increasingly widespread deployment of

mobile Internet in India, relying on smart mobile

phones, running apps, has important implica-

tions for buyer behavior and paths to product

search, discovery and purchase. Several in-

dustry reports worldwide nd that mobile apps,

such as those from large retailers such as Am-

azon and Flipkart, are used for search and dis-

covery by people with mobile phones in place

of using the websites, browsers and search en-

gines used in the desktop environment.

Indeed, a study by Criteo

nds that, world-

wide, apps convert at a rate 3.7 times higher

than mobile browsers in mobile commerce.

Similarly, a study by xAd/Telmetrics

on mo-

bile paths to purchase found that mobile users

in the U.S. that conduct retail research online

nalize their decisions with the help of sites as

Amazon, establishing the latter as a top refer-

ence for smartphone users narrowing their re-

tail decisions. The same

study found of the time

people spent on Am-

azon on their mobile

phones 79 percent was

spent using of apps ver-

sus 21 percent using the

web as of 2012.

Preliminary data from

India conrms these dif-

ferential paths to pur-

chase in Indian mobile

commerce too. Figure

6, adapted from a 2013

Nielsen consumer sur-

vey, documents the

“social” nature of the

mobile Internet in India,

68 e.g. http://www.mmaglobal.com/les/casestudies/xAd_Mo-

bile_Path_to_Purchase_Retail_FINAL.pdf

69 http://www.criteo.com/resources/mobile-commerce-report/

70 http://www.mmaglobal.com/les/casestudies/xAd_Mobile_

Path_to_Purchase_Retail_FINAL.pdf

where SMS, social networking and email and in-

stant messaging are the leading uses of smart-

phones.

Interestingly, the study nds that only

15 percent of Indian smartphone users use the

device to browse the web. This estimate is in

contrast to the case of China and U.S., where

the proportions of smartphone users who use

the device to browse the web are 75 percent

and 82 percent respectively. The Nielsen nd-

ings conrm the limited role of the mobile web

in Indian mobile commerce, and are consis-

tent with those of other reports, which nd that

WhatsApp and Facebook are the most down-

loaded apps in the Indian context.

Figure 6: Distribution of Smartphone Activity

by Proportion of Engaged Users

71 Nielsen, “The Mobile Consumer,” 2013, http://www.niel-

sen.com/content/dam/corporate/uk/en/documents/Mobile-Consum-

er-Report-2013.pdf

72 TNS, “Connected Life,” October 7, 2015, http://www.indi-

atechonline.com/it-happened-in-india.php?id=2011.

Activities performed on smartphones (India)

India

SMS

45%

17%

15% 15%

26%

13%

11%

Source: Nielsen mobile consumer report, a global snapshot 2013

Instant Messaging

Blowsing

VideoMobile TV

Apps

Streaming online music

Social Networking

Shopping

Mobile Banking

The buyer behavior discussed above is con-

sistent with differential paths to purchase docu-

mented in Indian mobile commerce. As shown

in Figure 7 below, mobile search drives a mini-

mal component of the trafc to leading online

retailers such as Flipkart, Snapdeal and Ama-

zon; the bulk of the trafc to these sites is driven

either directly through their app or through re-

ferrals. We expect that much like the U.S., where

Amazon is a rst stop for 43 product of consum-

ers for product search (compared with Google

for 34 percent of users), as the share of sales

made by online retailers increases in India, such

direct and referral trafc will increase many fold

relative to that driven through search.

D. Evolution of Smart Mobile Ecosystem

Growth in the data services of Indian tele-

com operators, mobile content service provid-

ers, and most important, smartphone shipments

is likely to drive accelerated adoption of the In-

dian mobile Internet in the future. A 2015 report

by KPMG nds that smartphone sales in India

are expected to grow at a projected CAGR of

53.8 per cent from 2013 to 2017, a growth rate

that is signicantly higher than that in other de-

veloped and developing markets. Further, while

the high growth rate of smartphones in the In-

dian market has primarily been an outcome of

domestic players that serve the low cost smart-

phone segment, it is expected that the country

will also be a prime market for more high-end

manufacturers, notably, Apple.

Figure 7: Source of Trafc for Major Indian

E-Commerce Platfor

Recent reports suggest that for the quarter

ending June 2015, Apple grew at 93 percent (al-

though off a signicantly lower base).

Apple

73 http://articles.economictimes.indiatimes.com/2015-07-23/

news/64773018_1_iphone-idc-india-tim-cook

Source: http://trak.in/tags/business/2014/06/04/top-10-indian-e-commerce-sites-comparison/.

has engaged in aggressive pricing and market-

ing strategies in India to stoke this growth. It has

discounted the iPhone4S and has introduced

buybacks and upgrade offers for the new iP-

hone 6 and iPhone 6 Plus. It has also been offer-

ing discounts and increasing its retail presence

through smaller-sized stores that target second

and third tier cities in the Indian market.

IV. Implications for Antitrust Analysis

The preceding discussion highlights funda-

mental changes in technology infrastructures,

technology adoption, technology use, and

conduct of business activity over the Internet.

These changes are transforming boundaries

between existing industries, creating overlap

between previously separate markets, and al-

tering competitive forces and actors in these

markets. These shifts have occurred not only

with respect to ofine vs. online commerce but

also within different forms of online commerce.

Competition authorities in developed and de-

veloping countries should factor in these chang-

es in how they monitor and govern the online

economy. Specically, the analysis of market

denition and market power needs to recog-

nize the impact of these changes, particularly

the move to smart mobile on demand-side and

supply-side substitution.

Changes in consumer behavior—in particular

how people discover, search and buy products,

how people communicate with each other

and with businesses, and how people consume

content—have important implications for ana-

lyzing substitution between various online and

ofine services. Consumers also have access to

many more options, in many more places, over

a much greater space of the day than they

had ve years ago. Consider something as sim-

ple as buying a new television.

Ten years ago a typical consumer in the US

probably did a search on their computer at

home, perhaps on a Saturday morning, of on-

line sellers and in doing that might have seen

some advertisements for physical sellers. That

consumer may have bought from one of the

online retailers or gone to some physical stores

to see what they had to offer and to inspect

some of the televisions available online. Today,

a typical consumer could easily do this on the

train on the way home or on their lunch break

using their mobile device. They could search

on Amazon, where the consumer could look

at offerings from Amazon itself as well as many

merchants that sell on Amazon Marketplace, or

check out information on their social network.

If they went to physical stores they could use

their mobile phone to compare prices, using

Amazon, Google, or various price comparison

apps, and decide based on this, while they are

standing at the physical store, to buy online or

at the store. Increasingly, people can also use

various apps to nd products, buy them online,

pick them up at the store, including televisions.

Competition authorities also need to consid-

er how the move to smart mobile and devel-

opment of cloud computing has altered the

supply-side of the equation. De novo entry is

far easier and cheaper as a result of the abili-

ty of entrepreneurs, as well as established rms,

to develop apps for mobile phones, that rely

primarily on the cloud-based delivery of ser-

vices and data analytics, and distribute them

globally to billions of people easily. The move

to smart mobile has, moreover, enhanced op-

portunities for entry and disruptive innovation.

For example, the rapid rise of Facebook as a

powerhouse in mobile advertising, challenging

Google, has resulted from the fact that Face-

book was able to develop a highly successful

mobile app and the fact that smart mobile is a

much better platform for social communication

since people can use it anywhere, all day.

The extent to which demand-side and sup-

ply-side substitution affect the analysis of mar-

ket denition and market power is ultimately an

empirical matter that needs to be addressed

on a case-by-case basis. The move to smart

mobile, along with other rapid changes in the

online economy, however, introduces new

considerations and expands the range of de-

mand-side and supply-side substitutes that

should be considered for that analysis. Function-

al approaches to market denition and market

power, which rely on comparing the detailed

features and functions offered by products,

are also increasingly less reliable. Consumers

and businesses use the features and functions

of apps, and mobile technology, in new and

creative ways that result in apps, which appear

very different, being used to accomplish the

same purpose. For instance, WhatsApp, which

is a messenging app, is being used in India for

e-commerce, and substitutes for Amazon and

Flipkart.

Market denition and marker power analysis

also needs to recognize the fact that the move

to smart mobile is changing the competitive

environment very rapidly and in unpredictable

ways as we saw above. For many areas of the

online economy in developed countries the set

of signicant players one would have identied

as demand or supply-side substitutes in 2010 is

very different that the set one would identify in

2015 or the set of players one would have iden-

tied in 2005. An analysis based on information

in 2005 would not have identied mobile-based

advertising as a competitive to web-based ad-

vertising and would not have anticipated that a

year-old company, Facebook, would become

one of the largest online advertising compa-

nies. Likewise, in 2010, few would have predict-

ed the extent to which messaging apps such as

WhatsApp and WeChat would obtain massive

global user bases and provide strong substitutes

for other methods of communication.

These changes entail that the analysis of

market denition and market power needs to

be forward looking—anticipating what is likely

to happen as this will affect the ability and in-

centive of rms to engage in abuse of domi-

nance—and modest and exible—since the

ability to predict the evolution of competition

has proven to be extremely difcult, and will

become more so as the smart mobile disruption

continues. This point is even truer in fast-grow-

ing economies such as India because the on-

line economy is developing much more rapid-

ly, from low levels, as mobile broadband and

smartphones reach critical levels.

The move to smart mobile, and the consid-

erations we have discussed, have four major

implications for antitrust analysis. In each case

we are recommending that competition au-

thorities exercise greater caution, not that they

adopt a laissez-faire approach.

First, apparent market power may not reect

real or durable market power because chang-

es in consumer behavior in response to new

technologies and the entry of new mobile apps

are likely, based on past experience, to provide

strong competitive constraints. It is increasing-

ly easy to develop new mobile apps to attack

market inefciencies, including that resulting

from market power. Previously separate markets

easily cross and overlap as digital technologies

lead to convergence. These changes expose

market leaders to competition from new play-

ers who speedily achieve huge penetration, as

well as to powerful players in adjacent markets.

Second, rapid changes in consumer behav-

ior and entry increase the likelihood of making

mistakes in the analysis of market denition and

market power. In static markets with well-de-

ned differences between rms competition au-

thorities can make judgments based on known

facts with great condence. In markets under-

going disruptive innovation, as is occurring in

the online economy generally, but particularly

as a result of the move to smart mobile, market

relationships are uid and are changing rapid-

ly. That disruption may give rise to an increased

number of complaints to antitrust agencies, but

the concerns of a traditional, potentially less ef-

cient, supplier may not be indicative of a risk

to consumer harm. Rather the rapid innova-

tion, often the cause of the disruption, may itself

be a sign of a competitive marketplace, with

consumers directly benetting from the lower

prices and better products brought about by

that innovation.

Third, there is a greater likelihood of rem-

edies having negative consequences. Since

we do not really know what the competitive

landscape will look like there is a lower chance

that any intervention designed based on cur-

rent knowledge will x a problem and there is a

higher chance that it will cause problems that

we cannot envision today. Put another way,

even if there could be a problem, any reme-

dy, in such conditions, may cause more harm

than good. Restraints on an incumbent rm

that is perceived to have abused a dominant

position, for example, could prevent that rm

from challenging even more powerful entrants

or other incumbents. Competition authorities

face a particular risk of negative consequenc-

es from remedies sought by companies that

have incentives to slow down fast-moving inno-

vative rivals.

Fourth, these same considerations make it

difcult to design remedies, or other interven-

tions, to correct perceived abuses of domi-

nance. A remedy that looks sensible from a

backward looking perspective may make no

sense in a few years, or possibly a few months,

after it has been put into place. The move to

smart mobile, for example, is at an extremely

early stage even in developed countries. Three

years ago people were still spending far more

time on the PCs than on their mobile devices.

There is no reason to believe that anyone can

predict what the world will look like in anoth-

er three years. Anyone designing interventions

needs to keep the limits of our knowledge in the

face of disruptive innovation rmly in mind.

We do not, however, want to overstate these

implications. Competition authorities need to

examine abuse of dominance allegations on

a case-by-case basis. There is no reason they

should stand down when it comes to the online

economy. They should monitor this important

sector of the economy vigilantly. There may well

be situations in which targeted interventions are

warranted. At the same time it would make no

sense to ignore the fact, which is apparent to

most people who use smart mobile phones in

their daily lives, that the online economy is un-

dergoing massive ux and that competition au-

thorities must be mindful of that in their analyses.

Failed Analogies:

Net Neutrality

versus “Search”

and “Platform”

Neutrality

Marvin Ammori

While many have lamented that the term

“network neutrality” is boring and unclear,

that

concept has inspired millions around the world

to le comments with national regulators,

and

led those regulators to take action in much of

the Americas, Europe, and Asia.

Perhaps as a

sign of net neutrality’s success in public debate,

some thinkers have started borrowing the word

“neutrality” for concepts that are supposedly

analogous to net neutrality, but really have very

little in common with it. The two best-known ex-

pressions are “search neutrality” and “platform

neutrality” (which apparently also encompasses

“app store neutrality), all of which have prompt-

ed discussion before regulators.

 AfliateScholar,StanfordLawSchool’sCenterforInternet

&Society.MylawrmadvisestechnologycompaniesincludingGoo-

gle. Nonetheless, the views expressed in this paper are my own views

and should not attributed to Google or any other client.

1 John Oliver’s jokes about the term are perhaps the most

humorous—while also being accurate. See Marvin Ammori, “John Ol-

iver’s Hilarious Net Neutrality Piece Speaks the Truth,” Slate, June 6,

2014.

2 Todd Shields, “It Took Four Million E-Mails to Get the FCC

to Set Net-Neutrality Rules,” Bloomberg, Feb. 3, 2015; Prasanto K Roy,

“India’s Fight for Net Neutrality,’ BBC News, April, 18 2015 (noting

over 800,000 emails on the issue).

3 Trey Williams, “Here’s How Other Countries Are Address-

ing Net Neutrality,” Marketwatch, Feb. 26, 2015

Whatever the potential merit of these two

concepts on their own (and there is good rea-

son to doubt their merit),

the supposed analo-

gy to net neutrality is awed and cannot justify

them. As I explain in this short paper, net neu-

trality is a specic concept backed by over a

decade of research and debate, descending

from common carriage, rooted in the partic-

ulars of the economics of Internet networks.

Search neutrality and platform neutrality lack

rigorous research or debate, are incoherent

new concepts unrelated to long-standing legal

doctrines, and the economics of “search” and

“platforms” are much different from those ani-

mating net neutrality.

The only thing that net neutrality has in com-

mon with search neutrality and platform neutral-

ity is the word neutrality. As a result, one would

have to justify search neutrality or platform neu-

trality based on different arguments than the ar-

guments that apply to network neutrality.

I. Net Neutrality

Net neutrality is the concept that Internet ser-

vice providers, which are mainly phone and ca-

ble companies in the US, should not be allowed

to block any websites or online software, nor to

technically discriminate against or in favor of

any of these sites or software, and should not

charge a fee for sites or applications to reach

users. In February 2015, the US Federal Commu-

nications Commission has adopted bright-line

rules against blocking, discriminating, and paid

prioritization. It also decided that access to the

Internet is a “telecommunications service” un-

4 Moreover, beyond failing as analogies to net neutrality, I

doubteither“search”or“platform”neutralitycanbejustiedonother

terms. See, e.g., Ryan Heath, CCIA, “The Wrong Suspect: A Debate

About Platform Neutrality Finds Little Evidence Of Consumer Harm,”

Project Disco, March 16, 2015; James Grimmelmann, “Some Skepti-

cism About Search Neutrality,” in The Next Digital Decade: Essays On

The Future Of The Internet (Berin Szoka & Adam Marcus, eds., Tech-

Freedom, January 2011).

der the US Communications Act, and therefore

subject to traditional common carrier rules that

applied to the phone networks. Phone networks

had been subject to rules against blocking calls

and discriminating, and network neutrality is ap-

plying that traditional concept to Internet ser-

vice providers.

Network neutrality is marked

by several facts.

First, there has been a lot of academic think-

ing about net neutrality. In the US, there have

been many, many academic conferences at

law schools, at economic schools, at business

schools; there have been dozens of expert

panels at the Federal Communications Com-

mission, the Federal Trade Commission, and the

United States Congress; there have been many

books, some very good, written on the subject;

hundreds of articles in law reviews; testimony by

the engineers who invented the Internet and the

World Wide Web. These experts have analyzed

the technical arguments; they have analyzed

the economic arguments, the competition ar-

guments, and the free expression arguments,

which do matter to many of the net neutrality

advocates. These experts have catalogued net

neutrality violations around the world, including

in the United States and in Europe. The research

is voluminous.

Second, the economic arguments for apply-

ing net neutrality rules are highly specic. They

include (rst) the notion that telecom networks,

particularly facilities-based providers, are likely

natural monopolies. Moreover, the US broad-

band market is marked by actual monopolies,

5 See In re Protecting and Promoting the Open Internet, Re-

port and Order on Remand, Declaratory Ruling, and Order, 30 FCC

Rcd. 5601 (2015) [hereinafter 2015 Open Internet Order].

6 Two good resources for arguments and citations are Barba-

ra van Schewick, Internet Architecture and Innovation (2010); Barbara

van Schewick, “Network Neutrality and Quality of Service: What a

Nondiscrimination Rule Should Look Like,” 67 Stan. L. Rev. 1 (2015).

as over seventy-ve percent of Americans have

only one choice (or zero) of a provider offer-

ing over 25 megabits per second of capacity.

In the US, only the incumbent companies that

were once monopolies—like Verizon and AT&T

and Comcast—dominate the markets even to-

day. There is no likelihood of that changing any

time soon and it has not changed for decades.

The monopoly (or at best oligopoly) on net-

works is durable. There are huge barriers to entry

in the market, both economic and governmen-

tal. On the wired side, a new entrant has to ne-

gotiate with every city and many private com-

panies for access to rights of way to lay wires in

the ground or to string them on utility poles. Even

with those permissions, the xed costs of serving

a community are often too high. One reason for

these monopolies is the overwhelming econo-

mies of scale deriving from very high xed costs,

although other factors include the difculty of

offering a compelling bundle of TV and broad-

band.

Indeed, in light of these costs, both the

phone wireline monopolies and the cable mo-

nopolies in the US were built under monopoly

regulation and with the federal and local gov-

ernments guaranteeing that the monopolist

would make a return on their investment. On

the wireless side, companies need permission

from the government, in the form of a wireless

license, which costs billions of dollars; after they

buy the license they have to build a network of

antennas and wires connecting them.

Moreover, these service providers have what

economists call a “terminating access monop-

oly” over the user. This economic circumstance

is the one that most experts, such as Barba-

ra van Schewick, emphasize, and the factor

7 FCC Chairman Tom Wheeler: More Competition Needed in

High-Speed Broadband Marketplace, https://apps.fcc.gov/edocs_pub-

lic/attachmatch/DOC-329160A1.pdf.

8 Daniel Frankel, “Google Fiber Chief: Program Rights Are

‘Biggest Impediment’ to Deployment,” FierceCable, Oct. 7, 2014.

is central to the FCC’s decisions in this space.

The most efcient way for me to get to Amazon.

com on my desktop computer is through my

Internet Service Provider. Once I choose one

company as my home Internet Service Provid-

er, the most likely way a website will reach my

desktop is through that provider. For many high-

data-use applications such as streaming video,

wireline Internet access is the only cost-effective

solution, meaning that the wireline provider fac-

es little or no competition from wireless solutions.

Relatedly, users usually do not choose multi-

ple providers of home Internet service. They usu-

ally have one home Internet service provider

and one wireless provider. And they treat these

two primarily as complements, not substitutes.

Because wireless ISPs generally have lower

data caps or provide throttled video, and Net-

ix even throttles its own video on some wireless

networks, users generally have to rely on wireline

connections for high-denition video streaming.

As a result, if a user’s wireline provider blocks

Netix, that user would likely not be able to

watch Netix in high-denition without chang-

ing wireline providers.

Finally, the switching costs are quite high, as

the FCC concluded. If my provider blocks Ama-

zon or Facebook, I can try to switch Internet pro-

viders. But I have to break a long-term contract,

and pay fees for canceling. Moreover, as noted

above, most Americans do not have options for

high-speed access, even if the switching costs

were low

That’s the economics.

Net neutrality also reects a policy judgment.

The judgment is that the users and creators of

9 The FCC’s reliance on this rationale was upheld in court.

See Verizon v. FCC, 740 F.3d 623 (D.C. Cir. 2014).

10  Sam Gustin, “Netix’s Shady Mobile Throttling Policy

Doesn’t Break Net Neutrality Rules,” Vice, Mar. 28, 2016.

applications, not the owners of the networks,

should decide the applications on the net-

works. The Internet should function much like

the electricity grid, where the users and the de-

vice-makers can plug in whatever they want,

without permission from the electricity networks.

That approach has led to great competition

and innovation.

(This is why, for example, net

neutrality applies to all Internet Service Provid-

ers—whether or not the provider has signicant

market share.)

Net neutrality is also a solution to an existing

problem. There have been violations around

the world.

And that is why the US has adopted

rules, and so have countries in other continents

such as South America and Europe.

Finally, those advocating for net neutrality in

the US have been primarily consumers, nonprof-

it organizations, small business, smaller startups,

churches, and democracy and civil society

groups. Somewhat famously, the FCC explained

that Google and Facebook were hardly en-

gaged on the issue, leaving it to small players to

advocate for the issue.

II. Search neutrality

Search neutrality is much different from net

neutrality. The concept is not extraordinarily

well-dened, but the notion is that search en-

gines, especially dominant search engines,

should not favor their own websites in the main

search results.

In addition, they should not favor

11 See Barbara van Schewick, Internet Architecture and Inno-

vation (2010).

12 See, e.g., Marvin Ammori, “Net Neutrality: A Solution to an

Existing Problem,”

Ammori.org, Apr. 16, 2014

13 JeffRoberts,“FCCOfcialSaysGoogle,FacebookHadLit-

tle Say on Net Neutrality,”

GigaOm, Mar. 3, 2015.

14 See Oren Bracha & Frank Pasquale, “Federal Search Com-

their own thematic or vertical searches through

boxes and answers.

For example, when

someone searches for a location in a general

search tool, search engines respond with their

own mapping results; Google Maps for Google

Searches and Bing Maps for Bing searches. Oth-

er companies offering maps would prefer if their

maps were shown instead.

The main object of search neutrality com-

plaints has been Google. A weather website

complained in Germany, a shopping site in

Brazil, and a restaurant comparison site among

others in the US, and each time the complaint

failed because Google was able to demon-

strate that including boxes and thematic an-

swers was good for consumers and there was

no evidence of Google manipulating results.

The leading article dismantling the concept of

search neutrality is perhaps James Grimmel-

man’s demonstration of the concept’s incoher-

ence.

The FTC rejected calls for search neutral-

ity, deciding that many of Google’s practices

reected those of its competitors and were, in

fact, pro-consumer.

mission - Access, Fairness, and Accountability in the Law of Search,”

93 Cornell L. Rev. 1149 (2008). See also James Grimmelmann, “Some

Skepticism About Search Neutrality,” in The Next Digital Decade: Es-

says On The Future Of The Internet (Berin Szoka & Adam Marcus, eds.,

TechFreedom, January 2011).

15 What it means to “favor” one’s own service is not a sim-

ple concept; is “favoring” the correct term for directing users to what

the search engine genuinely believes is the best result, when it is the

search engine’s own result? Who decides what the best result is, if not

the search engine itself? But I put that question aside for purposes of this

discussion.

16 Anunofcialtranslationof the Germanorderis available

at http://united- kingdom.taylorwessing.com/leadmin/les/docs/pdf-

german/Google_Weather_InBox_-_Court_Order_2013-04- 04_Unof-

cial_Translation.pdf. See also Greg Sterling, “Google Wins Major Anti-

trust Victory In Brazil, Does It Foreshadow Broader EU & US Wins?,”

SearchEngineLand, September 10, 2012; David Goldman, “Google

Dodges Major Antitrust Bullet,” CNN, Jan. 3, 2013.

17 James Grimmelmann, “Some Skepticism About Search

Neutrality,” in The Next Digital Decade: Essays On The Future Of The

Internet (Berin Szoka & Adam Marcus, eds., TechFreedom, January

2011).

18 Statement of the Federal Trade Commission Regarding

Google’s Search Practices, In the Matter of Google Inc., FTC File Num-

Whether there is merit to the theory of search

neutrality, it is not analogous to net neutrality.

First, the academic research is far thinner and

less sustained.

The concepts are far less understood and

have less support.

Second, the history is very different. Search

is not a traditionally regulated common carri-

er service or network; search engines were not

created as government-regulated monopolies;

they were born of competition in unregulated

markets, and did not need monopolies for in-

vestment.

Third, the economics are very different.

There are lower barriers to entry for search.

Entrants need not negotiate with cities for rights

of way, dig ditches and string wires on poles, or

spend billions on licenses and build a network

of towers. There are other costs of engineering

and infrastructure, but of a lower order, with far

fewer government permissions. In telecommuni-

cations, there have been almost no successful

entrants who were not once government-pro-

tected monopolies in the US. In search, there

are several entrants in the past decade and

venture capitalists funded 60 search companies

in 2013 and 2014.

Not only are the economic barriers lower, but

multi-homing is common and switching costs are

low. Users can switch from Google to Bing—both

of which are free—without breaking a contract,

waiting for a technician for a home-visit, or pay-

ing any fees. Shoppers consult on average 10.4

sources when researching a purchase.

Eco-

nomic consultancy Oxera showed that almost

ber 111-0163, January 3, 2013; Greg Sterling, “FTC Closes Google

Antitrust Case: ‘Law Protects Competition Not Competitors,’” Search-

EngineLand, Jan. 3, 2013.

19 Conor Dougherty, “Start-Ups Try to Challenge Google, at

Least on Mobile Search,”

N.Y. Times, May 3, 2015.

two thirds of consumers in France, Germany,

Spain, and Poland use more than two websites

or apps for the same task.

Users can reach the

competition in a single click.

And there is no terminating access monopoly.

So long as the network provider is not blocking

the site, a user can come to a restaurant-rank-

ing site by typing its URL into dozens of browsers,

multiple search engines, an app download, or

through Facebook or Twitter pages.

Beyond economics, there should not be a

policy judgment that search be “neutral.” The

whole point of search is to discriminate, to nd

the very best sites based on each query, and to

show them to users.

Meanwhile, the network

would provide most value as a general purpose

network, agnostic among uses.

Finally, the advocates for search neutrality

are not consumer groups or civil society, but

competitors. The FTC noted as such in rejecting

search neutrality arguments in 2013.

Indeed,

some of the main proponents of search

neutrality are telecommunications compa-

nies.

These companies seem to be deecting

attention from their own regulation, such as net

neutrality, and strategically argue for rivals to be

regulated instead. Few consumer groups have

voiced any support for search neutrality.

III. Platform Neutrality

Finally, “platform” neutrality is the latest en-

trant to the “neutrality” circle. It is also the most

confused and incoherent of the three concepts.

Generally, platform neutrality seems to mean

that online platforms are special and distinct as

a category, and also that this category should

be regulated in a category-specic way to re-

duce their power over all sides using the plat-

forms.

These platforms are sites that connect

some people to other people, which encom-

passes a wide range of businesses, such as mar-

ketplaces like Uber or Airbnb or eBay or Amazon

and entertainment or communications sites like

YouTube or Facebook. It also includes sites that

connect users to other websites, such as search

engines, or app stores that connect users and

developers, or dating apps that connect wom-

en and men, sites that connect employers with

recruits like LinkedIn, or that connect creators

with funders such as Kickstarter, and that con-

nect startups with investors such as Angellist.

The biggest problem with platform neutrality

is that these platforms have very little in com-

mon with one another in terms of their market

dynamics and characteristics. Devising ex ante

rules for a grab bag of “platforms” doesn’t

make sense. Plus, the remedies proposed are

very diverse, so platform neutrality is less a con-

cept than a slogan.

Many American nonprots, trade groups,

and startups have argued that the concept is

incoherent. These groups led survey respons-

es with the European Commission in response to

a consultation on platform regulation. Creative

artists, blogging platforms, handcraft market-

places, crowdfunding sites, and video platforms

all led explaining that the denition of platform

was incoherent and overbroad. Rather, they ex-

21 “The Zero Moment of Truth Macro Study,” Google/Shop-

per Sciences, April 2011.

22 “BenetsofOnlinePlatforms?”,Oxera,Oct.2015.

James Grimmelmann, “Some Skepticism About Search Neutrali-

ty,” in The Next Digital Decade: Essays On The Future Of The Internet (Berin

Szoka & Adam Marcus, eds., TechFreedom, January 2011).

Barbara van Schewick, Internet Architecture and Innovation (2010).

25 Greg Sterling, “FTC Closes Google Antitrust Case: ‘Law

Protects Competition Not Competitors,’”

SearchEngineLand, Jan. 3, 2013.

26 Nate Anderson, “Search Neutrality? How Google Became

a ‘Neutrality’ Target,” Ars Technica, Apr 29, 2010.

27 See also “Principles for Europe’s Digital Ambitions,” Me-

dium (collecting articles opposing platform regulation); Tom Fairless,

“EU Digital Chief Urges Regulation to Nurture European Internet Plat-

forms,” Wall St. J., Apr. 14, 2015.

plained, generally applicable law can address

issues in these markets, and a new regulation

lumping together disparate online businesses

could likely not reect any clear underlying prin-

ciple and would do more harm than good. For

example, the Internet Association wrote that

“regulatory intervention is not warranted be-

cause less restrictive solutions are not only avail-

able but already exist in the digital realm.”

Ac-

cording to Engine Advocacy, a leading voice

for startups in public policy, trying to regulate all

“online platforms” is “troublesome because it

lumps together a large set of companies, func-

tionalities, and systems. This dilutes the mean-

ing of the phrase and depending on how this

phrase will be used going forward, there is dan-

ger in regulating a broad swath of companies

that have little more in common than that they

operate online.”

The economics of all these “platforms” is not

extremely well understood, as these examples

include multi-sided markets, where the site has

to cater to both advertisers and users or to sell-

ers and buyers. Far from being conned to on-

line players, platforms have historically included

everything from newspapers to shopping cen-

ters, credit cards, and TV networks, and their

economics are all quite different.

There has been very little discussion from con-

sumer groups or nonprot organizations or citi-

zens clamoring for platform neutrality, at least in

the US. Indeed, based on the lings to the Euro-

pean Commission, the opposite is true.

To explore this vague concept more con-

cretely, we can focus on one form of platform

neutrality here—that concerning app stores. Un-

like net neutrality, there has not been tremen-

dous research in the area of app-store neutrali-

ty. While this lack of research is one reason to be

wary of jumping to regulation, it is also a reason

why I offer the following as somewhat tentative

thoughts.

First, the barriers to entry for app stores may

be signicant for an online service, but they are

nothing like those governing telecom networks.

Apple and Google have app stores. But so do

Amazon and Microsoft. There are also dozens of

smaller and regional app stores, such as the Op-

era Mobile store which has 30 million app down-

loads a month or Wandoujia in China which has

200 million users and 30 million app downloads

daily.

So the barriers may be signicant, but

not insurmountable. None of these app stores

are historical remnants of government-regulat-

ed monopolies obviously, nor do they need to

purchase licenses at auction or petition for local

rights-of-way.

Second, users can multihome and so app

stores do not have a terminating access mo-

nopoly. On handsets with Android OS, at least,

the user can download another app store

through the web or through a browser. Users

can also use websites if an app is not in the app

store. Sometimes a hardware maker or the car-

rier will preload its own app store or an inde-

pendent app store, and there can be multiple

app stores preloaded on a phone, such as both

a Samsung and a Google app store on a Sam-

sung phone.

To be sure, antitrust issues may be possible in

any market, including with app stores or plat-

28 Internet Association Position Paper on the Regulatory En-

vironment for Platforms, Online Intermediaries, Data and Cloud Com-

puting and the Collaborative Economy, at 2, Jan. 6, 2015.

29 Evan Engstrom, “U.S. Advocates Express Concern—And

Some Hope—For EU’s Digital Ambitions,” Medium, Feb. 10, 2016

(linking to Engine comments and many others).

30 Groups including Public Knowledge and the Center for De-

mocracy&Technologyhaveled.

31 See “App Stores/Service Providers,” One Platform Foun-

dation, http://onepf.org/partners/; see also Jamie Giggs, “The Ultimate

App Store List,” Business of Apps, Feb. 24, 2015

32 Dan Rowinski, “Why Samsung Is Cloning Google Play On

Its Smartphones,” ReadWrite, Apr. 25, 2013.

forms. One can imagine contracts keeping

competing apps out of an app store in some

aggressive way, or the owner of an app store

disallowing any competitors in its app store while

having enough market share for that to matter.

But traditional competition law can handle

that issue. A new concept of platform neu-

trality is unnecessary and perhaps counterpro-

ductive

IV. Conclusion

Hopefully this short paper helps shed light on

the stark differences between net neutrality and

search neutrality and platform neutrality. In my

view, the rst is well theorized and justiable. I

am not alone in that view, as net neutrality is

popular around the world.

But even were it not, net neutrality lacks com-

monalities with search neutrality theories and

platform neutrality theories. Therefore, it should

not be a good or bad precedent for either.

Search neutrality and platform neutrality must

stand on their own underlying justications, not

merely as analogies to net neutrality. And, when

asked to stand on their own, the two new con-

cepts appear wanting.

Antitrust,

Regulation and the

Neutrality Trap:

A plea for a smart,

evidence-based

internet policy

Andrea Renda

No. 104 / April 2015

Abstract

When they look at Internet policy, EU policy-

makers seem mesmerised, if not bewitched, by

the word ‘neutrality’. Originally conned to the

infrastructure layer, today the neutrality rhetoric

is being expanded to multi-sided platforms such

as search engines and more generally online

intermediaries. Policies for search neutrality and

platform neutrality are invoked to pursue a va-

riety of policy objectives, encompassing com-

petition, consumer protection, privacy and me-

dia pluralism. This paper analyses this emerging

debate and comes to a number of conclusions.

First, mandating net neutrality at the infrastruc-

ture layer might have some merit, but it certain-

ly would not make the Internet neutral. Second,

since most of the objectives initially associated

with network neutrality cannot be realistically

achieved by such a rule, the case for network

neutrality legislation would have to stand on dif-

ferent grounds. Third, the fact that the Internet is

not neutral is mostly a good thing for end users,

who benet from intermediaries that provide

them with a selection of the over-abundant in-

formation available on the Web. Fourth, search

neutrality and platform neutrality are funda-

mentally awed principles that contradict the

economics of the Internet. Fifth, neutrality is a

very poor and ineffective recipe for media plu-

ralism, and as such should not be invoked as the

basis of future media policy. All these conclu-

sions have important consequences for the de-

bate on the future EU policy for the Digital Single

Market

I. Introduction

Recently, EU policymakers seem to have be-

come obsessed by the concept of ‘neutrality’

when discussing future digital policy. This is true

not only for the well-known and long-lasting de-

bate on ‘network neutrality’, which refers to the

impossibility for an Internet Service Provider (ISP)

to discriminate between the bits of trafc ow-

ing on the portion of the network it manages.

The past few months have seen the concept

of neutrality spread like an oil spot, giving rise

to neologisms such as ‘search neutrality’ (e.g.

in the Google antitrust case), ‘device neutral-

ity’ (as invoked by the European Parliament

already in 2011), and lately ‘platform neutrali-

ty’ (as endorsed at the end of 2014 by French

1 SeetherstcontributionintheliteraturebyT.Wu(2003),

“Network neutrality, broadband discrimination”, Journal on Telecom-

municationsandHighTechnologyLaw,Vol.2,pp.141–178.Therst

response to the net neutrality problem was given by C.S. Yoo (2004),

“Would Mandating Broadband Network Neutrality Help or Hurt Com-

petition?: A Comment on the End-to-End Debate”, Journal on Tele-

communications and High Technology Law, Vol. 3 and elaborated later

in C.S. Yoo (2005), “Beyond Network Neutrality”, Harvard Journal

on Telecommunications and High Technology, Vol. 19 and C.S. Yoo

(20018), “Network Neutrality, Consumers, and Innovation”, University

of Chicago Legal Forum. For an illustration of the network neutrality

debate, see A. Renda (2008), “I own the pipes, you call the tune: The net

neutrality debate and its (ir)relevance for Europe”, CEPS Special Re-

ports, CEPS, Brussels, and A. Renda (2011), “Neutrality and Diversity

in the Internet Ecosystem”, CEPS Special Report, CEPS, Brussels. See

also, for a literature review and a progress report, Jan Krämer, Lukas

Wiewiorra and Christof Weinhardt (2013), "Net neutrality: A progress

report", Telecommunications Policy, Vol. 37(9), pp. 794-813.

Andrea Renda is Senior Fellow in the Regulatory Affairs re-

search unit at CEPS.

and German governments).

In all this, EU pol-

iticians, and especially Members of the Euro-

pean Parliament, intuitively attach a positive

meaning to the word: they would never dare

to vote against it, as this would portray them as

enemies of the public good, and in particular of

end users. Not surprisingly, the Parliament’s vote

on the Connected Continent package in April

2014 pointed at a much stricter view of neutrali-

ty than the one originally proposed by the Euro-

pean Commission. Similarly, in the United States,

net neutrality was publicly endorsed by Barack

Obama in a message recorded at the end of

2014, in which the US President called on the

Federal Communications Commission to strong-

ly endorse net neutrality in its revision of the 2010

Open Internet Order. The FCC, formally an in-

dependent agency, eventually followed the

desiderata of the President by casting a vote

in favour of network neutrality on 26 February

2015. The unprecedented feature of that vote

was that net neutrality was presented, more

than a mere regulatory issue, as a fundamental

right of the end users, thus calling for protection

at a higher, constitutional level.

All in all, reality suggests that no politician

feels comfortable when standing against the

‘neutrality’ totem. And indeed, there are many

reasons to believe that neutrality is, in many cir-

cumstances, a useful attribute for the Internet:

but it amounts to a means, not an end in itself.

At the same time, a closer look raises doubts as

to whether neutrality, applied to the Internet

ecosystem, is always the best choice for end

2 See the European Parliament’s Resolution of 17 Novem-

ber 2011 on the open internet and net neutrality in Europe (www.

europarl.europa.eu/sides/getDoc.do?pubRef=-//EP//TEXT+TA+P7-

TA-2011-0511+0+DOC+XML+V0//EN). On platform neutrality, see

the report published by the French national digital council (https://edri.

org/french-digital-council-publishes-report-platform- neutrality/). See

also J.-H. Jeppesen, “French and German Ministers Should Not Con-

fuse Platform Neutrality with Net Neutrality”, (https://cdt.org/blog/

french-and-german-ministers-should-not- confuse-platform-neutrali-

ty-with-net-neutrality/).

3 See http://www.fcc.gov/document/fcc-adopts-strong-sus-

tainable-rules-protect-open-internet

users or for society as a whole. Moreover, the

current state of the EU debate on Internet policy

reveals that, in many domains, EU proposals are

at once sanctifying neutrality as the Holy Grail

of the Internet, at the same time sneakily pro-

posing rules that fundamentally contradict the

neutrality principle. This is leading to the birth of

an array of new monsters, including rules that

seek to achieve net neutrality through 24/7 pa-

trolling of the Internet;

rules on platform liabil-

ity coupled with search neutrality obligations;

legislation pursuing media pluralism and access

to content through neutrality obligations; and

policy that pursues neutrality at the same time

that it seeks to segment the Internet through the

imposition of cloud localisation requirements.

In this brief paper, I argue that these rules are

fundamentally awed and critically detrimental

to end users and to the Internet ecosystem as a

whole.

1. Why did we want network

neutrality in the rst place?

The word neutrality has been given many dif-

ferent denitions and interpretations over the

past decade. Interestingly, if not worryingly, it

was used as a synonym of very disparate terms.

More specically, in the net neutrality debate

the following angles have been taken by com-

mentators and advocates.

4 See A. Renda (2013), “Net Neutrality and Mandatory Net-

work-Sharing: How to disconnect the continent”, CEPS Policy Brief

No. 309, CEPS, Brussels, 18 December (on the “Stockholm syn-

drome”).

5 The debate on cloud localisation requirements emerged af-

ter the Snowden revelations and gave rise initially to commercial offers

to store data within the EU (or a given member state). The debate on

the US-EU safe harbour framework is leading to proposed legislation to

force the localisation of EU citizens’ data within the territory of the EU

or any other jurisdiction with adequate data protection legislation. See

A. Renda (2015), “Cloud privacy law in the United States and in the Eu-

ropean Union”, forthcoming in Regulating the Cloud: Policy for Com-

puting Infrastructure, Christopher S. Yoo and Jean-François Blanchette

(eds), Cambridge, MA: MIT Press.

• Anonymity. When the network neutrali-

ty debate was in its infancy, in the early

1990s, neutrality was considered as a key

safeguard to preserve a user’s anonymity

and freedom to upload and download

any content without being inspected

or prosecuted. This attribute was closely

related to the end-to-end design of the

network, which entailed that the intelli-

gence would be exclusively located at

the edges of the Internet (i.e. with end us-

ers), and not at the core.

Early legislation,

such as, inter alia, the 1998 Digital Millen-

nium Copyright Act in the United States

and the 2000 E-Commerce Directive in

the EU, reected this original design: ISPs

(Internet service providers) were consid-

ered as ‘mere conduits’, and thus could

not be held liable for the conduct of their

subscribers. They could not (and were in

any case not supposed to) monitor user

behaviour and inspect trafc, just as gov-

ernments, too, could not monitor and in-

spect trafc.

• Competition and fair business practic-

es. Since the mid-2000s, and in particular

after the 2005 Madison River case in the

US, net neutrality was framed as a prob-

lem of competition between telcos and

over-the-top (OTT) players.

The concern

raised by the ‘neutralists’ was that verti-

cally integrated ISPs had a strong incen-

tive to block OTT applications such as

VoIP (Voice over Internet Protocol), which

could potentially erode their revenues.

Even without blocking them altogether,

according to this view, ISPs may have an

incentive to intentionally degrade the

6 On the role of anonymity in net neutrality in the debate ani-

mated by David D. Clark, one of the original creators of the end-to-end

protocol on the Internet, see in particular D.D. Clark and M.S. Blumen-

thal (2007), “The End-to-End Argument and Application Design: The

Role of Trust” (http://groups.csail.mit.edu/ana/People/DDC/E2E-07-

Prepub-6.pdf).

7 See Renda (2011), “Neutrality and Diversity”, op. cit.

quality of OTT applications, in a way that

could tilt the competitive balance in fa-

vour of the ISP’s own products. This would

amount to a form of non-price discrimina-

tion, or a refusal to deal in more orthodox

antitrust terms. More recently, the debate

on anticompetitive behaviour by ISPs has

extended to so-called ‘fair business prac-

tices’ in vertical value chains: these are

most often related to the fact that, ab-

sent mandatory net neutrality legislation,

ISPs could intentionally degrade the qual-

ity of the most QoS-dependent applica-

tions, to induce them to accept to pay

a minimum QoS (quality of service) fee.

Even Tim Berners Lee, one of the founders

of the Web, recently observed that, ab-

sent neutrality legislation, innovative app

providers might be forced to “bribe their

ISPs to start a new service”.

• Innovation. Part of the debate on net

neutrality focused on its impact on in-

novation. This entails the so-called ‘next

Google’ or ‘next Facebook’ argument,

according to which, since the neutral de-

sign of the Internet has made it possible

for very small start-up companies to en-

ter the marketplace and become huge

players, modifying this feature would

jeopardise the stunning level of innovation

observed so far, raising barriers to entry in

the market and transforming the Internet

into a ring-fenced property of the ISPs. In

addition, should the Internet evolve into

a two-speed or multi-speed environment,

with some applications enjoying better

QoS than others thanks to the payment

of an ad-hoc fee, new entrants with limit-

ed nancial resources would be doomed

to occupy the ‘dirt track’ of the Internet,

and this would inevitably prevent them

8 See Brian Fung (2014), “World Wide Web inventor slams

Internet fast lanes: ‘It’s bribery’”, Washington Post, 19 September

(http://www.washingtonpost.com/blogs/the-switch/wp/2014/09/19/

world-wide-web-inventor-lashes-out-at-internet-fast-lanes-its-bribery).

from showing what they’re great at.

• User choice. Quite often the debate on

net neutrality focuses on the need to en-

sure that end users have access to all the

content and applications they want, any-

where and from any device. Blocking or

throttling applications would, of course,

reduce the amount of information that

users can have access to, at any time.

Accordingly, legislation that allows the

creation of specialised services or ‘toll

lanes’ over the Internet, and even ze-

ro-rating offers that tie the use of a device

to access to a restricted number of inter-

mediated services would be contrary to

the fundamental principles of user choice

and empowerment.

• Openness. In presenting the 2013 Con-

nected Continent proposal, the Europe-

an Commission referred to network neu-

trality as “what keeps the Internet open”.

As explained by the U.S. FCC in reviewing

its Open Internet Order in February 2015,

“an Open Internet means consumers can

go where they want, when they want”.

In more practical terms, at the infrastruc-

ture level, this means that ISPs should not

be allowed to block access to legal con-

tent, applications, services or non-harmful

9 Over the past few years, zero-rating has spread in many

OECD countries. Regulators in Chile, the Netherlands, Slovenia and

Canada explicitly prohibited zero-rating, while regulators in Germany,

Austria and Norway publicly stated that zero-rating violates network

neutrality. A scholar who has been quite active in arguing against ze-

ro-rating offers is Barbara van Schewick of Stanford Law School. She

has recently proposed that the new FCC rules should explicitly ban two

types of zero- rating: 1) zero-rating in exchange for edge-provider pay-

ment and 2) zero-rating of selected applications within a class of simi-

lar applications without charging edge providers. See B. van Schewick

(2015), “Analysis of Proposed Network Neutrality Rules”, 18 Feb-

ruary (http://cyberlaw.stanford.edu/downloads/vanSchewick2015Anal-

ysisofProposedNetworkNeutralityRules.pdf). See also Antonios Dros-

sos (2015), "Guest blog: The real threat to the open Internet is zero-rated

content", World Wide Web Foundation, 17 February (http://webfounda-

tion.org/2015/02/guest-blog-the-real-threat-to-the- open-internet-is-ze-

ro-rated-content-continued/).

10 http://www.fcc.gov/openinternet.

devices (no blocking); to impair or de-

grade lawful Internet trafc on the ba-

sis of content, applications, services, or

non-harmful devices (no throttling); and/

or to favour some lawful Internet trafc

over other lawful trafc in exchange for

consideration of any kind (no paid prior-

itisation). More generally, the FCC estab-

lished as a more general rule that broad-

band providers shall not unreasonably

interfere with or disadvantage consum-

ers' access to the Internet.

• Mediapluralism andfreedom of expres-

sion. In the past few years, net neutrality

has also been prominently described as

tightly related to media pluralism.

For ex-

ample, in its contribution to the NET Mun-

dial Conference, the European Broad-

casting Union stated: “it supports a strong

regulatory framework for net neutrality,

reecting the fact that the openness and

non-discriminatory features of the Internet

are key drivers for innovation, economic

efciency and safeguarding media free-

dom and pluralism.”

The relationship be-

tween neutrality and pluralism stems from

the simple observation that if ISPs block,

throttle or in any way discriminate against

trafc, they might lter out unwanted

media outlets or intentionally degrade

non-afliated sources of information. An

ad-hoc EU High Level Group on Media

Pluralism and Freedom of Expression pub-

lished a report in January 2013, recom-

mending that “channels or mechanisms

11 See L. Belli and P. De Filippi (2013), “The value of Net-

work Neutrality for the Internet of Tomorrow: Report of the Dynamic

Coalition on Network Neutrality”; and also L. Belli and M. Van Bergen,

(2013). “Protecting Human Rights through Network Neutrality: Further-

ing Internet Users’ Interest, Modernising Human Rights and Safeguard-

ing the Open Internet”. Steering Committee on Media and Information

Society. And on free speech, see also www.aclu.org/net-neutrality and

www.savetheinternet.com/net-neutrality-what-you-need-know-now

12 See EBU’s position relating to the Commission proposal

for a regulation laying down measures concerning the European Sin-

gle Market for Electronic Communications and to achieve a Connected

Continent, 22 November 2013.

through which media are delivered to

the end user should be entirely neutral in

their handling of this content. In the case

of digital networks, Net Neutrality and the

end-to-end principle should be enshrined

within EU law”.

2. Is current net neutrality policy

tackling these concerns?

This section looks at the policy objectives pur-

sued by net neutrality, as described in section

1 above, and assesses if current approaches

have been, or are likely to prove, effective in

addressing the related concerns.

2.1. From anonymity to Big Brother?

Concerning anonymity, it is clear that the cur-

rent debate on Internet policy does not look at

the right to surf anonymously as a policy goal

per se, with some isolated exceptions.

This is

due to a number of concurring reasons. First,

the need to ensure copyright enforcement and

protection on the Internet has led to a gradual

relaxation of the ‘mere conduit’ principle that

entailed the lack of ISP liability for the infringing

behaviour of their subscribers. Measures such

as the French HADOPI ‘three-strikes law’ and

numerous other laws introducing a graduated

response to copyright infringement effectively

considered ISPs as cyber-police.

Second, se-

curity reasons have led to the explosion of mass

surveillance activities on the Internet, undertak-

en both by public authorities alone, and in coop-

eration with private Internet intermediaries such

as ISPs. The last generation of network and in-

formation security legislation, such as the recent

Executive Order on Cybersecurity adopted by

13 http://ec.europa.eu/digital-agenda/sites/digital-agenda/les/

HLG%20Final%20Report.pdf.

14 See e.g. www.torproject.org/.

15 See A. Renda (2011), Law and Economics in the RIA

World, Amsterdam: Intersentia, section 5.8.

the White House in April 2015, critically targets

intermediaries as potential facilitators of unlaw-

ful activities and encourages them to share all

relevant information about potential threats to

the resilience of the national critical information

infrastructure.

Third, the explosion of an array

of new applications and the Internet of Things

requires extensive packet detection and trafc

management in order to ensure communica-

tions at various levels of quality and latency.

Against this background, the EU debate has

rapidly moved towards protecting net neutrality

through pervasive monitoring of quality of ser-

vice, as already foreseen (although implicitly) in

the 2009 Universal Service Directive, which intro-

duced the possibility for national regulators to

intervene and impose a minimum quality of ser-

vice, should ISPs not intentionally throttle certain

trafc. The latest ofcial version of the draft Con-

nected Continent package, currently under tri-

logue between the European Commission, the

European Parliament and the Council, de jure

forces regulators to monitor the Internet on a

24/7 basis, in all portions of their territory, to nd

out if a given bit of trafc is being discriminated

against. Against this background, ensuring the

neutrality of the network no longer means pur-

suing user anonymity; rather, it is based on Big

Brother-like patrolling of the Internet, to ensure

that non- discriminatory behaviour is detected

and sanctioned.

2.2. Competition and fair business

practices: Is there a level play-

ing eld?

16 See Executive Order -- Promoting Private Sector Cyber-

security Information Sharing (www.whitehouse.gov/the-press-of-

ce/2015/02/13/executive-order-promoting-private-sector- cybersecu-

rity-information-shari).

17 See the declarations by Nokia’s CEO (www.cnet.com/news/

nokia-knocks-net-neutrality-self-driving-cars-wont-get-the-service-

you-need/).

18 See A. Renda (2013), “”Net Neutrality and Mandatory

Network-Sharing: How to disconnect the continent”, CEPS Policy

Brief No. 309, CEPS, Brussels, 18 December (on the “Stockholm syn-

drome”).

While original concerns on the anticompeti-

tive effects of trafc management practices fo-

cused on the ISPs’ potential to abuse their mar-

ket power (i.e. Section 2 of the US Sherman Act

and Art. 102 TFEU in Europe), today the issue is

often reversed. As a matter of fact, while it is true

that dominant ISPs could potentially engage in

anticompetitive conduct, such as discrimination

or refusal to deal, resulting in instances of block-

ing, or granting different treatment to equiva-

lent transactions (throttling, paid prioritisation),

EU policymakers are gradually discovering that

market power is distributed across all layers of

the value chain, and thus that potentially a

large IT giant could exploit superior bargaining

strength vis-à-vis ISPs. In Europe, this is leading to

rather counter-intuitive situations in which mo-

bile operators are considered dominant by tele-

com regulators and, at the same time, victims

of predatory behaviour on the part of large IT

rms.

All this is inconsistent with antitrust law and

economics, as dominance must be assessed in

light of existing constraints exerted not only from

rivals, but also from upstream and downstream

players.

A market player cannot be dened as

dominant by one authority, and as dominated

by another.

Competition between an ISP and an OTT ser-

vice can also be imbalanced if the former bears

costs that the latter does not face, such as net-

work maintenance and upgrade costs. Just as

network maintenance and upgrade are nor-

mally included in access charges determined

by regulators for new entrant e-communica-

tions operators that rely on the incumbent’s in-

frastructure under the EU Access Directive, there

19 See interalia http://bgr.com/2014/12/12/apple-iphone-an-

titrust-investigation/, http://www.androidauthority.com/apple-ip-

hone-carrier-deals-europe-examination-antitrust-abuse-176512/,

and http://www.zdnet.com/article/apple-under-european-investiga-

tion-over- iphone-ipad-sales-tactics-and-4g-restrictions/.

20 See Court of Justice of the European Union in Hoffman-La

Roche,deningdominanceasthepowerto“behavetoanappreciable

extent independently of competitors, customers and … consumers”; C-

85/76 - Hoffmann-La Roche v Commission, 1978.

is no economic reason why the same cost would

not have to be charged to those operators that

provide a similar service thanks to the existence

of a pre-existing telecoms infrastructure. Further-

more, if an OTT service ends up representing half

of the IP trafc carried by a single ISP, the issue

becomes critical: while it is true that the OTT ser-

vice creates positive externalities and increases

trafc for the ISP, not being able to monetise this

additional trafc can be disastrous for the infra-

structure operator. This is the issue that led the

DC Circuit Court of Appeals to decide against

the 2010 Open Internet Order in Verizon v. Net-

ix in January 2014. Furthermore, the relationship

between net neutrality and competition ulti-

mately rests in the principle of non-discrimina-

tion. To the extent that an ISP does not discrim-

inate between types of trafc, i.e. applications

that ideally fall into the same relevant market

(including, where appropriate, the ISP’s own

vertically integrated service), there is no reason

to believe that charging for minimum service

quality would be of any relevance to antitrust

law, let alone economic regulation.

Moreover, despite the absence of ad-hoc

neutrality legislation, OTTs have been gaining

market share everywhere in Europe: players

such as Skype and Whatsapp (now adding

voice calls to its successful messaging service)

have eroded the margins of ISPs without having

to pay for the use of the bandwidth. While this

is certainly a short-term benet for the end us-

ers, one wonders whether ISPs will nd it useful to

continue investing in an infrastructure that will in-

crease prot opportunities for other companies.

As a result, end users might suffer in the long

term due to lack of sufcient incentives to invest

in new infrastructure. Against this background,

the concerns about securing a level playing

eld that were raised after 2005 have led to a

situation in which the pendulum has swung to

the other extreme, with possible consequences

in terms of infrastructure investment.

In summary, net neutrality seems neither a

sufcient, nor an essential remedy for the per-

ceived lack of a level playing eld between

telcos and OTTs. The European Commission’s

current attempt to help national regulators in

considering, where appropriate, OTT players in

dening relevant markets is a more meaning-

ful approach to inter-layer competition than a

remedy that tilts the balance in favour of OTTs.

Depending on market circumstances, market

power might be found upstream or downstream,

thus in the hands of OTTs or telcos. But the rele-

vance of this debate is now weakened, at least

in the US, since the FCC is placing emphasis on

the ‘public utility’ nature of Internet access, and

is thus imposing neutrality obligations on all car-

riers regardless of their monopoly power: this re-

alises a quantum leap from net neutrality, from

regulatory issue to constitutional right.

2.3. Innovation and neutrality:

Friends or foe?

he fact that neutrality is essential for innova-

tion is a recurring mantra, especially in Brussels.

However, reality seems to be far more nuanced

compared to the ctitious, partisan statements

we hear on net neutrality every day. We at-

tempt to explain in plain words below why this

is the case, although the issue is rather complex

in and of itself.

To be sure, the original design of the Internet

has made it possible for companies like Google

and Facebook to emerge and quickly become

Internet giants (too big, according to some EU

policymakers). At the same time, the gradual

‘platformisation’ of the Internet has gradually

shifted the most turbulent and creative areas

of the Web into higher layers. The original, lay-

ered architecture of the Internet is now being

replaced by a patchwork of multi-sided plat-

forms operating with different business models

and with differing levels of openness.

Platforms

21 K.C. Claffy and David D. Clark (2013), “Platform Models

such as Apple’s iOS, Google Android, Amazon

Web Services and Microsoft Windows/Azure

are lowering barriers to entry for smaller play-

ers wishing to enter the Internet ecosystem.

the same time, importantly, such platforms are

being commoditised by applications that are

platform-independent: in most cases, this oc-

curs when apps are downloadable for free on

any platform, and then manage their custom-

er bill and data directly from the cloud. This is

the case of very successful apps such as Uber,

Spotify and many others. This blossoming rich-

ness and diversity at the app layer is one of the

key drivers of innovation in the current Internet

ecosystem. Entry possibilities are simply shifting

to higher layers, or downstream in this complex

supply chain.

Other issues must be crucially taken into ac-

count. Importantly, certain innovative services

cannot emerge without minimum quality of ser-

vice: think about Netix, but also e-Health, IoT

applications, innovative payment systems and

many others, which denitely cannot work if

they are not aided by some guarantee of ser-

for Sustainable Internet Regulation”, TPRC 41: The 41st Research Con-

ference on Communication, Information and Internet Policy, 15 August

(http://dx.doi.org/10.2139/ssrn.2242600).

22 See Manuel Palacin, Miquel Oliver, Jorge Infante, Simon

Oechsner and Alex Bikfalvi (2013), “The Impact of Content Delivery

Networks on the Internet Ecosystem”, Journal of Information Policy,

Vol. 3, pp. 304-330.

23 A deeper reection reveals that this has always been the

case in the history of the Internet revolution, and more generally in the

history of information technology. In complex systems with multi-lay-

ered value chains, strong indirect network effects and a modular de-

sign,certainmodulesbecomepivotalastheychieyaffectendusers’

preferences. This was initially the case of so-called ‘de-facto industry

standards’, such as the IBM processors, and later Microsoft Windows;

more recently they have taken the form of multi-sided platforms, such

as Google’s home page and Facebook’s social network page. In the fu-

ture, they are likely to become more centred on the Internet of Things

and content delivery: players such as Spotify and Netixare already

eroding the leadership of the GAFA (shorthand for American tech gi-

ants Google, Apple, Facebook and Amazon), triggering reactions such

as Google’s YouTube restructuring and Apple’s new music streaming

service. In all this, the tendency of innovation over the Internet is quite

consistent over time: large-scale innovation moves network effects and

leading platforms at higher layers of the Internet ecosystem, at the same

time commoditising lower layers.

vice quality and latency. Similarly, the lack of in-

centives to invest in new infrastructure can also

jeopardise the emergence of innovative ser-

vices, which critically depend on the availability

of sufcient bandwidth. Furthermore, new tech-

nologies such as 5G mobile broadband systems

will adopt a multi-tier architecture consisting

of macrocells, different types of licensed small

cells, relays, and device-to-device networks

to serve users with different quality-of-service.

Since this clearly entails trafc management

and prioritisation, it is not clear if pro-neutrality

legislation would lead to a major drawback in

the rollout and uptake of 5G networks.

Finally, net neutrality legislation at the infra-

structure layer can, under certain circumstanc-

es, divert investment incentives towards the

creation of private networks for the provision of

enhanced quality services: this might lead to

acceleration in the fragmentation of the Inter-

net, and an even speedier loss of neutrality. A

good example is the creation of large Content

Delivery Networks that interconnect with the

public Internet very close to the end users, such

as those owned by Netix and Akamai. I return

to this point below.

In summary, innovation can emerge both un-

der neutrality and diversity: the more the Web

grows, the more applications diverge in terms

of required latency and capacity; the more

user attention becomes scarce, the more some

degree of trafc optimisation will be needed to

protect the end user experience.

2.4. User choice and democracy

Is net neutrality really so effective in promot-

ing user choice and empowerment? To be sure,

it allows users to access all content of choice,

without undue discrimination. But it does not

protect end users against restrictions to con-

24 See, inter alia, S. Lauson (2015), “Suddenly, net neutrality

doesn't look so great for 5G”, IDG News, 4 March (www.pcworld.com/

article/2893032/5g-net-neutrality-may-be-headed-for-a- showdown.

html).

tent availability and application discrimination

applied by platforms located at higher layers.

A quick observation of current practice on the

Internet reveals that most of the discrimination

takes place at the higher layers, not at the in-

frastructure layer. Large platforms block or

degrade certain applications, and search en-

gines, by denition, have to make a selection in

order to prove useful for their end users. Whether

these practices are good or bad for end users

is a question that still awaits a good, evidence-

based, debate. In principle, behavioural eco-

nomics suggests that on the Internet, “a wealth

of information creates a poverty of attention”.

This, in turn, leads end users to increasingly rely

on any intermediary that is credibly able to se-

lect the most relevant information and offer it to

the end user, thus reducing search costs and,

more generally, transaction costs. At the same

time, it is important to reect on the extent to

which market forces alone could provide the

right incentives for intermediaries to select infor-

mation in a way that offers the best possible ser-

vice to the end user. And most importantly, as I

will argue below, there is reason to doubt that

neutrality in the selection of information would

be necessarily in line with the interest of the end

users.

Also, the jury is out concerning questions of

democracy. It is very important to avoid extrem-

ist stances in the debate: while neutrality can

in many circumstances contribute positively to

democracy, intended as the granting of equal

rights to all users and the absence of censor-

ship (at least at the network level), the com-

plete standardisation of Internet offerings has

very little to do with democracy. The prevailing

rhetoric in Brussels (and now Washington) is as

follows: since Internet should be treated as a

service of general interest, just like water, every-

body should have access at the same (afford-

25 See H.A. Simon (1971), "Designing Organizations for an

Information-Rich World", in Martin Greenberger (ed.), Computers,

Communication, and the Public Interest, Baltimore. MD: The Johns

Hopkins University Press. pp. 40–41.

able) terms and conditions. No one should be

able to access the Internet at a better speed,

or at more favourable conditions than others.

However, while this statement sounds very at-

tractive, in reality it is controversial. Is a world

in which everybody has access to the basic

postal service, but no one can have access to

express courier services more democratic than

the world we live in? Does democracy entail

that only public hospitals exist, and no private

clinics? Is it democratic to have just state high-

ways with no toll lanes, rather than forms of traf-

c optimisation based on users’ preferences?

Rather than mirroring democracy, full-edged,

rigid net neutrality rules are equivalent to what

the Trabant was in Eastern Germany: the only

car that people could have, very neutral, very

bad, very cheap, identical for everybody. It be-

came famous in the Western world when the

Berlin wall fell 25 years ago, and thousands of

East Germans drove their Trabants over the bor-

der: once in the ‘free’ world, they immediately

abandoned their ‘neutral’ cars, and started to

enjoy their new, non-neutral life.

As a result, the openness and democracy an-

gle of the net neutrality debate appears to be

heavily polluted by a layer of rather supercial

ideology. Once again, I do not mean to argue

that neutrality is always bad for democracy. On

the contrary, I believe that censorship should be

avoided at all layers of the Internet architecture.

At the same time, presenting democracy as a

situation in which only one Internet offer exists

for all users does not do justice to the richness of

user preferences and of the Internet itself. Once

again, reality is more complex; we need a more

evidence-based debate before we propose

unacceptably extreme visions of what is good

and bad on the Internet

2.5. Openness:

A means, not an end

On the question of ‘openness’, net neutrality

cannot be a stand-alone, self-sufcient solution.

While it is true that the Internet could develop

initially also thanks to the non-proprietary stan-

dards that govern it, it would be a mistake to

believe that once net neutrality is mandated,

the Internet would become open, let alone

neutral. Hence, the European Commission is

wrong in stating that net neutrality is what keeps

the Internet open.

Following the most widespread denition,

openness means that users can have access to

any content, anytime, anywhere and from any

device. But it is clear that the Internet ecosys-

tem has never been like this, and is increasingly

less so. Since the development of mass personal

computing, all business models on the Internet

have evolved in a way that mixes proprietary

elements with open ones. Microsoft Windows

was an early example of a semi-open archi-

tecture: it brought enormous advantage to its

end users due to enhanced standardisation

and network effects, despite the fact that it was

not interoperable with other operating systems.

Even free and open source software, initially

characterised by full openness (to the extent

that the rst licenses like the GPL were ‘viral’,

i.e. they could not be used in combination with

proprietary software), gradually became part

of largely proprietary business models. And as

a matter of fact, the real champions of open

source software today are companies that pos-

sess huge patent portfolios, and often use open

source software as a ‘Trojan horse’ to conquer

customers (e.g. IBM has become over the past

decade the most powerful sponsor of Linux).

In recent years, mobile access to the Internet

has been dominated by platforms that are not

completely open. iPhone owners cannot use

Android or access Android-specic or Windows-

specic applications, and vice versa. More gen-

erally, many great inventions at the logical and

applications layer of the Internet have initially

entailed a mostly proprietary model, and later

became more open. Forcing openness from the

very beginning might just not be a good idea to

start with.

At the same time, there is a more subtle as-

pect of the openness debate that is worth re-

calling here. As a matter of fact, it is not always

true that more openness is better. For example,

the usage restrictions featured by PDF les are

harming users’ ability to modify the les they re-

ceive: but it is exactly this feature that has made

the fortune of the PDF. Given restrictions on le

manipulation, documents can circulate much

more easily, and trust between senders and re-

ceivers becomes easier to establish. Likewise,

the closed nature of the iTunes-iPod- FairPlay

architecture has made it possible, for Apple, to

create the rst online store for legal music down-

loads with a sustainable business model. Any al-

ternative, including a more open architecture

with no vertical integration between the de-

vice and the format of the downloaded songs,

would have meant the failure of the business

model itself.

Openness of course does not coincide with

neutrality, and is broader since neutrality entails

strict non-discrimination between bits of trafc.

That said, since openness is not entirely a reality

on the Internet, neutrality a fortiori cannot be.

And indeed, it is clear that even with mandatory

network neutrality, trafc on the Internet would

be discriminated and toll lanes would continue

to abound. Just think about the growing role

that Content Delivery Networks (CDNs) play on

the network: players such as Akamai, Limelight

and Level 3 offer services that accelerate trafc

on the Internet, and are used by IT giants such as

Apple to ensure that services such as FaceTime

work better than the average, non-accelerat-

26 See Boston Consulting Group (2011), “The new rules of

openness’’ (www.libertyglobal.com/pdf/new_rules_%20of_open-

ness6-en.pdf).

27 See Martijn Poel, Andrea Renda and Pieter Ballon (2007)

"Business model analysis as a new tool for policy evaluation: Policies

for digital content platforms", info, Vol. 9, No. 5, pp.86–100.

28 In practice, the two concepts are often used interchangeably.

ed application.

In addition, some other play-

ers have built this capacity to exploit in-house

caching rather than buying it from third parties:

this is the case for many players, including Skype

and Google, which invested in a private infra-

structure made of a large networks of servers in

order to be able to offer a better service to their

end users. Netix itself used Limelight until 2012

and later moved to an in-house CDN called

OpenConnect.

Against this background, one would con-

clude that the only way to keep the Internet

fully open and neutral would be to impose neu-

trality obligations at all layers of the Internet ar-

chitecture. Would this be desirable? Not really:

Internet freedom should imply also freedom to

experiment with closed or semi-open architec-

tures, and with vast differentiation of product

offerings to match different user preferences.

Conversely, being forced to accept openness

and neutrality has very little to do with freedom.

2.6. Net neutrality will never be an

answer to media pluralism

Media pluralism is one of the most pressing

challenges of today’s digital policy. While some

commentators originally expected that the In-

ternet would address the issue of pluralism by

exponentially increasing the sources of informa-

tion available to the end users, reality showed

that the provision of information on the Internet

is becoming even more concentrated than in

traditional media. This trend has been captured

well by Columbia University Professor Eli Noam,

who showed in 2011 that media ownership

tends to become more concentrated at every

new generation of communications, from radio

to newspapers, to televisions and the Internet.

29 See D. Rayburn (2014), “Apple Building Out Their Own

CDN To Deliver Content To Consumers” (http://blog.streamingmedia.

com/2014/02/apple-building-cdn-software-video-delivery.html).

30 Seehttps://openconnect.itp.netix.com/

The issue of pluralism was associated even

more strongly with net neutrality as govern-

ments have shown the tendency to violate

neutrality by shutting down social networks and

ltering out forms of communication such as mi-

croblogs, in countries like Egypt, Turkey, China,

Russia, Venezuela and others.

However, there are three main reasons why

net neutrality cannot be the answer to the

thirst for media pluralism evoked by many pol-

icy-makers and scholars.

• Censorship, even if made impossible at

the infrastructure layer, can be exercised

by forcing intermediaries to lter com-

munication, even if no blocking takes

place at the infrastructure layer. One

clear example is the decision by the Turk-

ish government to temporarily shut down

Facebook, YouTube and Twitter until they

removed from their sites the picture of a

prosecutor taken hostage and killed by

militants in Istanbul in early April 2015.

• Relatedly, media pluralism is not guaran-

teed at all by the absence of blocking or

discriminatory behaviour by ISPs. Rather,

it would require a similar approach at all

layers of the value chain, since content

could otherwise be ltered out by large

platforms, news outlets, cloud providers,

etc.

• Even if no blocking, throttling or discrim-

ination takes place at all layers of the

value chain, neutrality tout court would

31 See “Facebook, Twitter to Appeal as Turkey Blocks So-

cial Media”, Bloomberg, 6 April 2015 (www.bloomberg.com/news/

articles/2015-04-06/social-media-blocked-in-turkey-on-prosecutor- or-

der-hurriyet).

32 In order not to increase complexity for the reader, I leave

aside here the rst amendment debate raised by Verizon in the US,

which raises the issue whether prohibiting ISPs from exercising edi-

torial powers on the content they transfer would amount to a violation

of their freedom of speech http://www.globalresearch.ca/when-net-neu-

trality-becomes-programmed-censorship-2/5434400 (last visited on

April 12, 2015).

not be sufcient to guarantee pluralism.

The reason is simple: pluralism requires not

only that a plurality of sources of informa-

tion is present on the Internet; on the con-

trary, it requires that a plurality of sources

of information is exposed to the end user.

Against this background, a neutral plat-

form would inevitably end up selecting in-

formation from the most popular sources:

the polarisation of sources of information

would be exacerbated, rather than re-

duced, by a strict neutrality requirement.

2.7. Summing up: What reasons remain

valid for net neutrality regulation?

The previous sections have shown that, re-

gardless of the intrinsic merit of the word and

the underlying concept, a lot more has to be

proven before net neutrality can be considered

as a universal principle, able to address all the

concerns raised with respect to ‘net diversity’

scenarios. Even the decision by the US Feder-

al Communications Commission (FCC) on net

neutrality of February 2015 is not based on any

in-depth evidence-based analysis, or at least

no such analysis has been published. As things

stand, net neutrality does not appear to be a

stand-alone remedy that would x any of the

identied problems, nor would it achieve any

of the ofcially pursued objectives as set out in

section 1 of this paper. That said, one should not

immediately conclude that net neutrality should

not be mandated at the infrastructure layer.

Simply no one has brought sufcient evidence

that this is the case, and probably the debate

has focused on the wrong motivations. Impos-

ing network neutrality might still be a good idea,

but for other reasons. In this section, I try to imag-

ine such potential motivations.

A rst reason why it might be a good idea to

33 See e.g. C.R. Sunstein (2001), Republic.com, Princeton, NJ:

Princeton University Press and C.R. Sunstein (2009), On Rumors: How

Falsehoods Spread, Why We Believe Them, and What Can Be Done,

Princeton, NJ: Princeton University Press.

impose net neutrality is that the infrastructure

layer is by far the most stable in the Internet

ecosystem. Accordingly, it might be easier to

monitor practices adopted by ISPs compared

to what takes place at higher layers, where the

Schumpeterian gale of ‘creative destruction’

operates at the speed of light, making it impos-

sible to dene markets, detect practices and

administer sanctions. Net neutrality legislation,

in this respect, would mean keeping the Inter-

net open at the lowest level of its architecture,

and then monitoring the market at higher levels

to discourage abusive practices, which would

result in a violation of end-users’ rights to a rea-

sonably open Internet. This regulatory option, of

course, would come at a cost, i.e. the loss of in-

centives to ISPs to roll out new public infrastruc-

ture.

A second, related reason that might favour

pro-net neutrality legislation is that it is a lot eas-

ier and efcient to implement this rule, com-

pared to any of the alternatives

. This would be

due to the fact that, as I have noted in a previ-

ous paper, implementing legislation that implies

difcult judgments such as whether the open in-

ternet is being ‘materially impaired’ might prove

to be impossible.

What is material impairment?

When does an impairment become material?

How do we get to know the counterfactual (i.e.

how much would a given ISP have invested in

broadband infrastructure absent the neutrali-

ty provision)? Where in the network would we

measure the impairment? Would this be in a

sample of locations, or everywhere in the ‘last

mile’, or in each apartment? How would we

allocate responsibility between providers that

manage consecutive trunks of the network,

34 A similar argument, referring to net neutrality as a “bright-

line rule”, is made by B. van Schewick (2015) “Analysis of Proposed

Network Neutrality Rules”,18 February (http://cyberlaw.stanford.edu/

downloads/ vanSchewick2015AnalysisofProposedNetworkNeutrali-

tyRules.pdf)

35 See A. Renda (2013), “Net Neutrality and Mandatory Net-

work-Sharing: How to disconnect the continent”, CEPS Policy Brief

No. 309, CEPS, Brussels, 18 December (on the “rst legislate, then

think” syndrome).

in case congestion is slowing down a service?

Would this imply an ex-ante regulatory remedy

or an ex-post enforcement tool?

Which pa-

rameters will we use to judge whether impair-

ment is below or above the admitted threshold?

Which services will be taken as a benchmark for

QoS parameters? There is enough uncertainty in

these questions to make any legislator want to

run away from the issue.

A third reason why one would want to have

neutrality legislation at the infrastructure layer is

that this is a rst step towards imposing neutrality

at all layers of the Internet. As I argue in the next

section, however, and despite the current em-

phasis being placed on neutrality as a universal

concept, there are reasons to believe that this

would be terrible news for Internet users (see be-

low, section 3).

Finally, regardless of the underlying reason,

several key observations emerge from the pre-

ceding discussion:

• Strict net neutrality regulation would re-

move some of the incentive motivating

ISPs to invest in the open Internet and

might lead to the creation of alternative,

private networks in the attempt to reach

end users with selected content (like

Comcast does in the United States).

• In a related vein, treating the Internet as

a service of general interest, subject to

strict neutrality requirements, may lead to

enhanced public funding of basic Inter-

net access, aimed at avoiding all sorts of

discrimination between end users’ terms

of access and the ability to send and re-

ceive content. This approach might be

pursued at the EU level through a prior-

36 David D. Clark, Steven Bauer, William Lehr, K.C. Claffy,

Amogh D. Dhamdhere, Bradley Huffaker and Matthew Luckie (2014),

“Measurement and Analysis of Internet Interconnection and Conges-

tion”, paper prepared for 43rd Research Conference on Communica-

tions, Information and Internet Policy, Arlington, VA, 9 September

(available at SSRN: http://ssrn.com/abstract=2417573).

itisation of broadband investment in the

so-called ‘Juncker plan’.

• Expecting ISPs to invest in the open Inter-

net without being able to optimise traf-

c or create specialised services, and

without being able to compete with

CDN-enabled content providers, comes

very close to “having your cake and eat-

ing it”, or what I referred to as the ‘Gal-

ileo syndrome’ in a previous paper.

The only remaining incentive for ISPs to

invest in better networks would be com-

petitive pressure exerted by other ISPs.

But either such competitive pressure is

already there (in which case, there is no

specic need to introduce net neutrality

legislation, according to many), or there

is a tiny chance that a less and less prof-

itable market, such as EU broadband ac-

cess, could attract signicant investment

and, accordingly, generate more vibrant

competition in the years to come.

3. Should we expand neutrality to

higher layers of the Internet ar-

chitecture?

While the network neutrality debate still looms,

the rst steps of the Juncker Commission seem to

have led to new efforts to extend the neutrality

principle to Internet ‘platforms’, including, inter

alia, wireless operating systems (Android, iOS,

Windows Phone) and web services managed

by Internet giants (e.g. Amazon cloud services).

These proposals echo the recent positions ad-

opted by the European Parliament (following

mostly the German and French governments)

on the need to adopt structural measures to re-

duce the market power of large Internet players

(notably Google) and to ensure the portability

of data across platforms to (allegedly) stimulate

competition and avoid user lock-in effects. This

37 See Renda, Net neutrality and mandatory network sharing,

op. cit.

trend conrms what some commentators had

envisaged a few years ago: that the neutrality

debate can easily spread to cover all layers of

the Internet ecosystem and be transformed into

a more general call for an all-neutral Internet.

Whether this trend will continue in the next Eu-

ropean Commission’s proposed packages on

the Digital Single Market and on Audiovisual

Services, both expected in the coming months,

is too early to predict. The European Commis-

sion has announced that it intends to launch a

stakeholder consultation on the possible regu-

latory approach to ‘digital platforms’, although

the contours of this initiative are still unknown at

the time of writing.

Below, I briey reect on two possible exten-

sions of neutrality principles that have been con-

sidered in the past months at EU level: search

neutrality and platform neutrality.

3.1. Search neutrality: Heaven or hell?

One of the most famous applications of the

concept of neutrality at higher layers of the

internet architecture is emerging from the still

rather obscure antitrust investigation launched

by the European Commission against Google,

reportedly coming closer to a nal judgment in

mid- 2015. From the few documents that have

left the premises of the European Commission

in the past months, as well as from the ofcial

statement of Commissioner Margrethe Vestag-

er on 15 April 2015 announcing the formalisation

of allegations against Google for abuse of dom-

inance, it seems clear that the most important

part of the investigation is related to Google’s

alleged abuse of dominance, consisting of the

manipulation of search results in favour of ‘pre-

ferred’ (often, Google’s own) content and to

the detriment of other results, demoted for var-

ious reasons.

Such behaviour, according to

38 38 See European Commission Factsheet, Antitrust: Com-

mission opens formal investigation against Google in relation to An-

droid mobile operating system, Brussels, 15 April 2015 http://europa.

eu/rapid/press-release_MEMO-15-4782_en.htm.

the European Commission, has the potential to

foreclose smaller search engines such as those

specialised in specic sectors (‘verticals’), which

end up being disadvantaged vis-à-vis the giant

search engine powered by Google.

The main allegation against Google turns out

being one of ‘non-neutrality’: Google is thought

to unduly discriminate between Internet content

by providing a non-neutral, non-objective view

of the Internet. This implies, inter alia, algorithmic

choices that demote bad-quality services, sites

that only aggregate information without adding

new one, and ltering out of illegal sites, hate

speech and copyright infringing content.

But

the obvious counter-argument is that a search

engine is not supposed to be neutral: in particu-

lar, Google had just completed a transition from

a ‘ten blue links’ model to that of an integrat-

ed search engine, which entails more editorial

responsibility, and at the same time more rele-

vant and satisfactory results for the end users.

Paradoxically, the European Commission’s case

against Google is mostly summarised by this di-

vergence: the Commission accuses Google of

not being neutral, but any search engine, not

39 For example, in both investigations carried out by the US

Federal Trade Commission (FTC) and by the European Commission,

UK search comparison site Foundem was one of the leading claimants,

arguing that Google had anti-competitively degraded its ranking, caus-

ing loss of market share. As explained by Crane (2014), inter alia, the

FTC found that Google’s transformation of its search engine from a

“ten blue links” system to an integrated search portal, in which Goo-

gle itself takes more editorial responsibility, has meant innovation and

enhanced consumer welfare, and positive results also for the types of

content that were considered to be useful for the end users. Foundem

was found to be a lousy service, and as such demoted by Google in

what is denitely, and fortunately, a non- neutral search engine. See

Daniel A. Crane (2014), "After Search Neutrality: Drawing a Line be-

tween Promotion and Demotion", I/S: Journal of Law and Policy for the

Information Society, Vol. 9, No. 3; Marvin Ammori and Luke Pelican

(2012), “Competitors’ Proposed Remedies for Search Bias: ‘Neutrality’

and Other Proposals”, Journal of Internet Law, Vol. 15, No. 11.; Daniel

A. Crane (2012), “Search Neutrality and Referral Dominance”, Journal

of Comparative Law and Economics; Daniel A. Crane (2012), “Search

Neutrality as an Antitrust Principle”, George Mason Law Review, p.

1199; and Geoffrey A. Manne and Joshua D. Wright (2012), “If Search

Neutrality is the Answer, What’s the Question?”, Columbia Business

Law Review, 151.

just Google, would reply “why should I be neu-

tral?”

Should search neutrality be a policy objec-

tive at all? Indeed, there are reasons to believe

that a neutral search engine would be hated by

consumers. As I mentioned in the previous pag-

es, the key role of Internet intermediaries, and

especially search engines, is to eschew neu-

trality by selecting the information that is likely

to prove more useful and relevant for the end

user. It is this reduction of complexity that makes

them so pivotal in their role of gatekeepers of

the Internet: just like our brain simplies reality

to make the abundance of information in the

outside world more manageable and useful,

a search engine has to reduce complexity to

help us nd our way through the Internet. This

requires, inevitably, a ranking (and thus, a dis-

crimination) of results, which can be based on

relevance as well as on any other factor that

is likely to increase customer satisfaction when

using the search engine. As recalled by James

Grimmelman,

users continually return to a spe-

cic search engine because they nd the ‘bi-

ased’ or ‘subjective’ results to t their needs, not

because they nd the results to be objective.

Without entering into the merit of the Google

investigation, which would go beyond the sub-

ject matter of this paper, it is clear that advo-

cating neutrality for search engines is far from

being a straightforward policy stance. Apart

from what has already been explained above

(that users are unlikely to want neutrality in a

search engine), it is clear that implementing

search neutrality would be undesirable in many

respects. First, the polarisation of results induced

by search engines purely based on relevance

might lead to even greater barriers to entry for

new companies that seek to enter the market.

Since they have never been listed on the Inter-

40 J. Grimmelman (2011), “Some Skepticism about Search

Neutrality”, in Berin Szoka and Adam Marcus (eds), The Next Digital

Decade: Essays on the Future of the Internet, TechFreedom, and NYLS

Legal Studies Research Paper No. 10/11 #20 (available at SSRN: http://

ssrn.com/abstract=1742444).

net, any crawling and mapping of the Internet

performed by a search engine would not nd

them. And before they achieve a minimum

scale of popularity, which would enable them

to appear in the rst page of a neutral search

engine’s home page, they might have already

gone bankrupt.

Second, search neutrality would need to be

veried, and this might require that Google, as

well as all competing search engines, disclose

their algorithms in a fully transparent way. How-

ever, apart from the fact that this would chill in-

novation by denying trade secret protection to

the result of massive R&D investment, it would

also expose the algorithm to attacks, as well

as strategic behaviour aimed at exploiting the

weaknesses of the algorithm to rank better in its

results. Even this outcome would not be neutral

in the end

Third, and relatedly, such a remedy is being

proposed as an antitrust remedy, although the

current debate on platform regulation also hints

at search neutrality as a way to promote plu-

ralism (see next section). As an antitrust reme-

dy, however, search neutrality requires a nding

of dominance within a given relevant market,

and this is prohibitively difcult to imagine if one

takes antitrust seriously. More specically, dom-

inance – let alone its abuse – requires a situa-

tion in which an undertaking is able to behave,

to an appreciable extent, independently of its

competitors, customers and consumers. In other

words, a situation in which a company is able to

sit down and relax since no one is able to effec-

tively and seriously challenge its market power.

Notwithstanding the very high share of search

queries that Google holds in Europe, the deni-

41 The Financial Times reported on 16 April 2015 that a pro-

posal currently making its way through the French senate could force

Google to publish the details of how its search rankings are calculated.

According to the newspaper, the proposed bill would allow the coun-

try’s national telecoms regulator to monitor search engines’ algorithms,

with powers to ensure its results are fair and non- discriminatory (see

www.ft.com/intl/cms/s/0/643f49ec-e285-11e4-aa1d- 00144feab7de.ht-

ml#axzz3XGqpfF5O).

tion of dominance as ascribed by the Court of

Justice of the European Union (CJEU) portrays a

different situation compared to the one Google

seems to be experiencing in Europe and global-

ly. A market leader that constantly innovates to

preserve its leadership is not a dominant com-

pany under EU antitrust law, regardless of the

market share.

In summary, search neutrality is a awed rem-

edy, both in antitrust terms and even more as

a general regulatory measure. In terms of anti-

trust, it amounts to throwing the baby out with

the bath water, runs counter to consumer wel-

fare and should be dened at a minimum as a

disproportionate remedy under EU law. In reg-

ulation, it is simply an ill-conceived extension of

the important, but per se controversial, principle

of network neutrality

3.2. Platform neutrality and regulation:

Where all contradictions explode

The intrinsic contradictions of EU digital policy

become fully apparent if one considers the pro-

posals to regulate platforms and impose forms

of neutrality on online intermediaries that have

recently been tabled in Brussels. One of the

rst to use the expression ‘platform neutrality’

was the French National Digital Council (Con-

seil National du Numérique), which published

a detailed report on this same concept in June

2014, following a 2013 request from the Min

is-

try of the Economy and Digital Affairs as well as

the Secretary of State on Digital Affairs.

The

report argued that platforms such as so-called

GAFTAM (Google, Amazon, Facebook, Twitter,

Apple, Microsoft) maintain their dominant po-

sition by three main operations: acquisition, di-

42 See Conseil National du Numérique (2014), “Platform Neu-

trality: Building an open and sustainable digital environment”, Opinion

No. 2014-2, of the French Digital Council, Paris (www.cnnumerique.fr/

wp-content/uploads/2014/06/PlatformNeutrality_VA.pdf).

versication and exclusion,

and that in doing

so, they harm competition to the detriment of

consumers. In the following weeks, the French

and German governments explicitly called on

the Commission to establish regulation for es-

sential platforms, invoking neutrality as one of

the attributes of such platforms’ future con-

duct.

The debate also surfaced in the US this

year, when Blackberry CEO John Chen ofcially

complained that Netix had not made movies

available for Blackberry phones, and invoking

platform neutrality – or ‘app neutrality’ – as a

much-needed remedy.

The echo of these calls is also heard inside

the European Commission. Some of the leaked

documents related to the upcoming policy ini-

tiatives on the Connected Continent and the

Digital Single Market hint at platforms as, very

generically, multi-sided markets where suppli-

ers and consumers of content, goods and ser-

vices meet. In consideration of the fact that

more than one third of Internet trafc goes to

the only 1% of websites which are used in all

member states, these ‘platforms’ are thought to

signicantly alter consumer choice by providing

misleading information. Reference is also made

to the difculty for consumers in distinguishing

between organic and paid-for search results, as

well as the ‘ranking’ (order) of results. Being so

43 Platforms buy innovative start-ups that could threaten their

dominance in the long run and/or that can be fruitfully integrated in

theirexistinginfrastructureinordertoprovideamorediversiedplat-

form. The report lists the acquisitions of the GAFTAM from 2010 to

January 2014, which shows that these platforms have been engaging in

acquisitionanddiversication.Thelastmainmoveoftheplatformsis

exclusion. For instance, the report argues that, when Google introduced

Google Maps and Google Shopping, the trafc of websites offering

similarservicesdroppedsignicantlybecausetheirpageranksuddenly

worsened.

44 See i.a. “Europe's demands on Google mount”, Financial

Times, 26 November 2014 (www.ft.com/intl/cms/s/0/66b5149e-758a-

11e4-b082-00144feabdc0.html#axzz3WpR2sSn7).

45 See Karl Bode, “No, 'App Neutrality' Is Not A Thing”,

13 February 2015 (www.techdirt.com/blog/netneutrality/arti-

cles/20150122/08093329777/no-app-neutrality-is-not- thing.shtml).

powerful, platforms can also impose unfair con-

tractual clauses to SMEs. Moreover, the digital

single market is perceived as negatively affect-

ed by the lack of interoperability between plat-

forms, and in particular by the fact that some

apps only run on specic operating systems,

some e-books are readable only by specic

e-readers, etc.

That is all to say that the European Commis-

sion seems to have been pervaded by a neu-

trality delirium, which so far has not produced

ad-hoc legislation, although this might just be

a matter of weeks away. There are many rea-

sons why platform neutrality is a contradictio

in terminis. First, as already recalled, platforms

are dened very broadly. Some commentators

have observed that based on the proposed

denition, even a mall can be a platform. Any

newspaper is a platform, just like any game con-

sole. Spotify is a platform. Uber is a platform too.

So, maybe the most vocal governments and

the European Commission only wanted to refer

to large, digital platforms. Or maybe to domi-

nant, digital platforms. Or maybe to GAFTAM, or

to GAFA, or perhaps only to Google. But then,

an explanation should be given of why compe-

tition law would not be sufcient to tackle the

problem. Maybe because it is an uphill battle for

these platforms to dene a relevant market and

nd dominance in contrast to what occurs at

the infrastructure level, especially if one follows

an orthodox antitrust approach. As a matter of

fact, the common feature of these platforms is

that they compete for end users, but they do it

with a rather different mix of products and ser-

vices, and very different business models. And

the ght to conquer end users’ attention is a

common feature of all market players, even in

the traditional world. Since users can, and do,

‘multi-home’ by using services provided by vari-

ous platforms at the same time, the existence of

a number of large and heterogeneous players,

which can be dened as large digital platforms,

does not say anything about the existence of

an antitrust problem. And even if one tries to

dene relevant markets for each of those plat-

forms, possibly ending up with a single market

for each of the GAFTAM, then the existence of

high market shares could not be used as a proxy

for dominance, since the economics of network

effects and tipping suggests that in these mar-

kets the winner takes it all, and competition is

normally very aggressively dictated by the need

to secure a paramount role in the next genera-

tion of market products.

Second, platform neutrality is an oxymoron,

since platforms capture the attention of end us-

ers exactly because they violate the neutrality of

the Internet and offer users a selection of Inter-

net content. They normally do not block or hide

content, but they necessarily prioritise content.

App stores present end users with a selection of

the best apps, the newest apps and those that

best t the user’s needs. Search engines are all

about relevance and salience, and compete

on the quality of their selection, as well as on

their ability to capture the attention of end users,

which in turn becomes a target of proled ad-

vertising. It is inherent in the nature of platforms

that these subjects will end up violating the neu-

trality principle. Again, the possibility (effective-

ly achieved in reality) for users to multi-home

is what determines the degree of competition

between various users: in this respect, recent re-

search on the multiplicity of channels that rms

can user to reach end users and evidence of

the increasingly cross-platform nature of apps

suggest that competition between non-neutral

platforms is producing virtuous results for end us-

ers.

And innovation at the app layer, as already

46 See C. Shapiro and H. Varian (1999), Information Rules:

A Strategic Guide to the Network Economy, Cambridge, MA: Harvard

Business School Press and R. Pardolesi and A. Renda (2002), “How

safe is the king’s throne? Network externalities on trial”, Chapter 11

in R. Pardolesi, A. Cucinotta, R. Van den Bergh (eds), Post-Chicago

Developments in Antitrust Law, Cheltenham: Edward Elgar.

47 See P. Nooren, W. Koers, M. Bangma, F. Berkers and M.

Boerkers (2014), “Regulation in the Converged Telecom-Media-Inter-

net Value Web”. TNO Report R11428, October 2014, the Netherlands.

observed, seems to be extremely vital, because

of this hybrid form of competition between dif-

ferentiated platforms. Recent data suggest that

400,000 Europeans are building apps, and that

the broader App Economy supported already

1.8 million European jobs in 2013, with a reve-

nue of €17.5 billion that same year, expected to

grow by 300% (to €63 billion) by 2018.

How can

this evidence be reconciled with the claim that

platforms are choking innovation?

Third, the platform neutrality debate evident-

ly clashes with a simultaneous trend, i.e. the at-

tribution of greater responsibilities to digital plat-

forms for the conduct of their users.

Such trend

is visible in several initiatives adopted at the

EU level, including the European Commission’s

plans to review the 2000 e-commerce Directive

to modify the ‘mere conduit’ principle (Article

12) to introduce a ‘duty of care’ principle, i.e. a

requirement for online intermediaries to act pro-

actively and remove illegal content hosted on

their platforms,

the attribution of growing liabil-

ity to online intermediaries for copyright protec-

tion, enforcement of privacy laws (including the

‘right to be forgotten’), defamation, spam lter-

ing, notication of security breaches, the ght

against terrorism and other monitoring activi-

ties. The contradiction lies in the fact that some

parts of EU law seem headed in the direction

of imposing neutrality obligations on online in-

termediaries; whereas on the other hand, other

legislation is requiring intermediaries to be more

proactive in managing, prioritising and editing

the content they pass on to the end users. How

would this work? Can, for example, search en-

gines be forced to operate ‘neutrally’ and at

48 See “How Europe can win in the global app economy”, Eu-

ractiv, 3 February 2015 (www.euractiv.com/sections/innovation-enter-

prise/how-europe-can-win-global-app-economy- 311788).

49 “Europe enlists Internet giants in ght against on-

line extremism”, by C. Spillman, 9 October 2014 (http://phys.org/

news/2014-10-eu-internet-giants-online-extremism_1.html#inlRlv).

50  See inter alia www.internetsociety.org/sites/default/les/

ISOC%20EU%20Newsletter%2027%20March%20201 5%20FINAL.

PDF.

the same time be attributed editorial responsi-

bility and related liability? This would amount to

saying “you can’t control or lter your results, but

if anything unlawful comes out of user queries,

you’ll be responsible”. It would also be a new

generation (if not an aberration) of the mere

conduit principle, in which a company is forced

to act as a mere conduit, but is also considered

liable for whatever happens in the conduit.

Which company would accept to continue op-

erations under these rather tricky terms?

Finally, the platform neutrality principle is in

stark contradiction with the objective of media

pluralism, which is also being pursued by EU law

and is currently subject to a ‘tness check’ (or

‘REFIT’ exercise) in the European Commission,

in view of reforms to be adopted in 2016. The

problem is similar to the one already outlined

for net neutrality in section 1 of this paper, but

exacerbated by the scarcity of attention and

trust that characterises the provision and con-

sumption of media content. In short, platforms

need to select content, and in selecting con-

tent polarise the attention of end users on a sub-

set of available information. A neutral search

engine would not address media pluralism,

since it would simply convey the most popular

and relevant results to the end user, and would

leave aside the long tail content that otherwise

adds to the plurality of voices we would want to

see on the Internet. Again, this does not mean

that nothing can be done to pursue media plu-

ralism; however, that ‘something’ that can be

done has nothing to do with neutrality. Sever-

al scholars, including Gillespie (2010), Helberg-

er (2012), Crawford (2013), Latzer et al. (2014),

Sunstein (2009), Zittrain (2014) and Goodman

(2014) have fuelled the debate

on how to

51 See Tarleton L. Gillespie (2010), “The Politics of 'Plat-

forms'”, New Media and Society, Vol. 12, No. 3 (available at SSRN:

http://ssrn.com/abstract=1601487); N. Helberger (2012), “Exposure

diversity as a policy goal”, Journal of Media Law, 4 (1), 65-92. doi:

10.5235/175776312802483880; Latzer et al. (2015), “The Economics

of Algorithmic Selection on the Internet”, forthcoming in Johannes

Bauer and Michael Latzer (eds), Handbook on the Economics of the

Internet, Cheltenham: Edward Elgar; Sunstein (2009), On Rumors,

design a proactive media policy in the age of

online intermediaries: this debate is inspired by

an understandable sense of urgency as regards

the need to address the prominent role played

today by platforms in conveying news and con-

tent to end users. But at the same time, this de-

bate has nothing to do with extremist neutrality

positions and rightly recognises that the way to

ensure plurality of content exposure (not merely

presence) in the age of algorithms is much more

complex than simply dictating the neutrality of

platforms. The debate is in its early stage, how-

ever. Authors like Zukerman (2013) even pro-

pose a serendipity engine that brings in random

content that might be relevant for the end user,

while Grimmelman (2011) convincingly demon-

strates that any lter or algorithmic rule entails

an editorial choice.

Very recently Grötker

(2015) proposed the creation of a public search

engine that acts as a benchmark for the results

of private engines, but the proposal stops short

of explaining how such public engines could be

made at least as attractive as the commercial

ones.

Summing up, platform neutrality seems to

represent a awed response to a badly dened

problem. This does not mean of course that

there are no problems to solve: monitoring the

way in which platforms make use of their editori-

al power is the biggest challenge for media pol-

icy in the years to come. What I am arguing in

this paper is simply that such a challenge will not

be addressed by imposing neutrality obligations

on platforms, but rather by seeking cooperation

op. cit.; Jonathan Zittrain (2014), “Engineering an Election”, Harvard

Law Review Forum, Vol. 127, and Harvard Public Law Working Paper

No. 14-28 (available at SSRN: http://ssrn.com/abstract=2457502); and

E.P. Goodman (2014), “Informational justice as the new media plural-

ism”, LSE blog, 19 November (http://blogs.lse.ac.uk/mediapolicyproj-

ect/2014/11/19/informational-justice-as-the-new-media-pluralism/).

52 See Grimmelman, op. cit.

53 See R. Grötker (2015), “The Citizens’ Internet: The Many

Threats to Neutrality”, Netopia, Brussels, March (www.netopia.eu/

wp-content/uploads/2015/04/Netopia-Report-The-Citizens- Internet.

pdf).

with platforms to ensure that the design of their

algorithms and their editorial choices are com-

patible with media pluralism objectives that are

considered to be in the public interest.

4. Conclusion

This short paper digs into the roots of the neu-

trality debate and comes to a number of con-

clusions, which I hope will be of interest to the

reader. They are summarised below.

First, while there might be reasons to impose

neutrality at the infrastructure layer of the Inter-

net, these reasons have little to do with the idea

that the network should be fully neutral. Rather,

they have to do with the ease of implementa-

tion, the stability of the infrastructure layer, the

possibility of identifying dominant positions and

the difculty of establishing a threshold for the

‘material impairment’ of the best-effort Internet.

At the same time, this policy choice comes with

a trade-off: policymakers should recognise that

this means diverting incentives to invest in the

best-effort Internet towards private networks;

and that this might also imply a greater involve-

ment of public funding to secure investment in

a high-capacity network. Policymakers should

also be open to the possibility that this policy

option brings more fragmentation of public

and private networks, and the ow of many val-

ue-added services into private networks.

Second, whatever outcome the net neutrality

debate produces, this will not make the Internet

neutral. The juxtaposition of various multi-sided

platforms with varying degrees of openness and

the use of various forms of trafc acceleration

make the Internet non-neutral, inevitably and

fortunately. Whether keeping the infrastructure

layer neutral would be a way to ensure that the

rest of the ecosystem evolves towards open

and innovative platforms, rather than to deprive

the trafc acceleration market of one category

of players, is a matter worthy of further research.

Third, the neutrality debate should not be ap-

plied to the higher layers of the Internet. Doing

this would fundamentally contradict the eco-

nomics of the Internet and the evolution of the

Internet itself. The fact that the Internet is no

longer a place where “nobody knows you’re a

dog” is an acknowledged fact, and allows no

turning back.

A fortiori, imposing both neutral-

ity and liability all at once on online intermedi-

aries would undesirably place them between

a rock and a hard place. Rather, the extent to

which online intermediaries can cooperate with

public authorities to protect and empower on-

line users is the key research and policy question

of the future. Both the liability of intermediaries

in e-commerce, copyright, data protection and

the protection of fundamental rights, and the

design of a smarter and proactive media policy

represent key challenges and opportunities for

EU policymakers in view of a ourishing and cre-

ative digital single market.

54 I refer here to a cartoon authored by Peter Steiner and pub-

lishedin1993ontheNewYorker.Thecartoonshowedapetdogsurng

the Internet and enthusiastically telling a fellow dog: “On the Internet,

nobody knows you’re a dog!”

Multisided

Platforms, Dynamic

Competition and

the Assessment

of Market Power

for Internet-Based

Firms

David S. Evans

10 March 2016

Abstract

Market power on each side of a multisided

platform, whether in the form of increasing pric-

es or decreasing quality, is constrained by the

risk of losing sales on the other sides. That tends

to weaken market power on each side and en-

courages platforms to keep prices lower and

quality higher than they would absent these

feedback effects. In some cases the nature of

the business model, and competition, result in

the platform allowing one type of customers to

participate in the platform for free or even to

subsidize their participation. Non-price methods

of attracting customers are especially import-

ant in this case, particularly when the business

model adopted by the industry makes it dif-

cult for platforms to move from free participa-

tion. To provide a reliable assessment of com-

petitive constraints, market power analysis must

consider the interdependencies in demand by

the participants on the platform as well as have

heightened focus on non-price competition

when the participation for one group is free.

Market shares should be used cautiously in as-

sessing market power for multi-sided platforms,

especially when they reect only one side of the

platform, and therefore do not account for the

interdependent customer groups, or concern a

free platform side where there is no monetary

measure of value. Finally, dynamic competition

makes the analysis of market power complex

because it results in feature competition, and

potentially drastic innovation, on one side of a

platform that has feedback effects on the other

side of the platform. The courts and authorities

have recognized these points in Qihoo 360 v.

Tencent, Cartes Bancaires v. European Com-

mission, the Facebook/WhatsApp merger, and

the Microsoft/Skype merger. These principals

should become part of the standard analysis of

multi-sided platforms by courts and competition

authorities globally. These concerns are illustrat-

ed in the context of multi-sided platforms that

offer online services where free services and dy-

namic competition are especially important.

I. Introduction and Summary

Many online businesses operate multi-sid-

ed platforms that help different types of par-

ticipants get together and enter into value-in-

creasing exchanges. Facebook, for example,

makes it possible for friends, businesses, adver-

tisers, and developers to interact with each other.

This business model has ancient roots going back

at least as far as the village matchmaker. Many

traditional businesses, such as newspapers and

shopping malls, use this model. New technologies,

particularly mobile and the cloud, however, have

* Evans is the Executive Director, Jevons Institute for Com-

petition Law and Economics and Visiting Professor, University College

London; Lecturer, University of Chicago Law School; and Chairman,

Global Economics Group. The authors gratefully acknowledge funding

from Google and excellent research support from Clara Campbell and

Nicholas Giancarlo at Global Economics Group.

turbocharged the multi-sided platform business

model. Online platform businesses are forming at

a rapid clip and disrupting not only traditional in-

dustries but relatively new ones as well.

Online multi-sided platforms pose a challenge

for competition policy analysis. Some have be-

come large national or global enterprises quick-

ly. Competition authorities are, quite properly,

vigilant about making sure that these successful

rms adhere to sound competition-law princi-

ples. In making economically reliable assess-

ments, however, competition authorities, as

well as courts, should account for three features

of these online platforms set them apart from

many other businesses in evaluating the market

power held by these platforms.

First, the demands by the different groups

of participants served by multi-sided platforms

are interdependent. As a simple mathematical

matter, that interdependency renders stan-

dard formulas wrong at least without signicant

modications.

In particular, a price increase, or

quality decrease, to one group of participants

reduces the demand not only by that group

but also by the other groups who then have

fewer participants with which to interact. That

does not mean that an online platform could

not have market power, only that the analy-

sis needs to consider these interdependencies

and the resulting feedback effects.

Second, many online businesses make the

platform “free” to one group of participants,

or even subsidize those participants, and earn

prots from the other groups of participants who

they do charge.

Although the basic concepts

1 See David S. Evans and Richard Schmalensee, Matchmak-

ers: The New Economics of Multisided Platforms (Boston, MA: Har-

vard Business School Press, 2016) Available at Matchmakers.

2 See David S. Evans, “The Consensus among Economists on

Multisided Platforms and Its Implications for Excluding Evidence That

Ignores It,” Competition Policy International, April 13, 2013. Avail-

able at http://ssrn.com/abstract=2249817 or http://dx.doi.org/10.2139/

ssrn.2249817

3 See Evans and Schmalensee, Matchmakers, Table 2.1, and

of competition policy analysis apply to free pric-

es, many of the traditional tools used for com-

petition policy analysis, such as the SSNIP test,

do not work, without signicant modication, as

a straightforward mathematical matter. Most

importantly, though, the existence of a group of

customers who are served for free highlights the

importance of considering the other interde-

pendent sides in assessing market power. The

platform is ordinarily making participation “free”

for a group because that group is very import-

ant for attracting paid participants. Anything

that deters “free” users from participating—such

as a decrease in quality—also reduces the in-

centives for the paid users, who generate all the

prots, from participating as well.

Third, online platforms often engage in con-

stant incremental innovation as they seek to

obtain advantages over rivals to attract par-

ticipants on multiple sides and are subject to

episodic, but increasingly frequent, disruptive

innovation in which new, or seemingly different,

rms attract their customers away. This dynam-

ic competition is particularly important for “at-

tention” platforms for which competition is de-

signed to attract the attention of users, which is

then resold to marketers, including advertisers,

who want to persuade those users to buy things.

An attention seeker is under constant threat that

someone will come up with an entirely clever

new way to grab people’s attention. For com-

petition policy analysis, this means that market

power analysis needs to consider the constraints

imposed by dynamic competition and in new

products and services that may appear very dif-

ferent than the rm under investigation.

Courts and competition authorities have

come to recognize these points as they have

had the chance to analyze online platforms

and absorb the teachings of the new econom-

ic literature on multi-sided platforms. Although

it did not involve online businesses, the Europe-

the detailed discussion in Chapter 7.

an Court of Justice recognized that the analysis

of competitive effects, and therefore implicitly

the exercise of market power, needed to con-

sider the linkages between the separate sides

of multi-sided platforms.

The Chinese Supreme

People’s Court concluded that dynamic com-

petition among platform businesses, including

one seeking and selling attention, limited mar-

ket power.

Antitrust regulators, including those

in the European Union and United States, ap-

proved Microsoft’s acquisition of Skype and

Facebook’s acquisition of WhatsApp in be-

cause they recognized how uid market bound-

aries and dynamic competition would discipline

the market power of the merged entities.

None of these judgments or decisions in

any way suggests that competition authorities

should let their guard down when it comes to

online platforms. Taken together, however,

with the new economics of multi-sided plat-

forms, and the growing body of evidence on

the dynamics of online competition over the

last two-decades, these judgments and deci-

sions do indicate that courts and competition

authorities should exercise caution, and adjust

their tools, in analyzing market power for online

platforms.

4 Groupement des cartes bancaires v European Commission,

Judgement of the Court, September 11, 2014, available at: http://curia.

europa.eu/juris/document/document.jsf;jsessionid=9ea7d0f130d57c-

17cb5e4cdc4d5f8196c74dd814db12.e34KaxiLc3eQc40LaxqMbN4O-

c3iSe0?text=&docid=157516&pageIndex=0&doclang=EN&mode=l-

st&dir=&occ=rst&part=1&cid=293160; Federic Pradelles and

Andreas Scordamaglia-Tousis, “The Two Sides of the Cartes Bancaires

Ruling: Assessment of the Two-Sided Nature of Card Payment Systems

Under Article 101(1) TFEU and Full Judicial Scrutiny of Underlying

Economic Analysis,” Competition Policy International Journal, Autumn

2014, Volume 10 Number 2.

5 David Evans and Vanessa Zhang, “Quhoo 360 v Tencent:

First Antitrust Decision by the Supreme Court,” October 21, 2014,

https://www.competitionpolicyinternational.com/qihoo-360-v-tencent-

rst-antitrust-decision-by-the-supreme-court.

6  Case No Comp/M.6281 - Microsoft/Skype, Ofce of the

Publications of the European Union, July 20, 2011, http://ec.europa.eu/

competition/mergers/cases/decisions/m6281_924_2.pdf and Case No

COMP/M.7217 – Facebook/WhatsApp, Ofce of the Publications of

the European Union, March 10, 2014, http://ec.europa.eu/competition/

mergers/cases/decisions/m7217_20141003_20310_3962132_EN.pdf.

This paper describes the new economics of

multi-sided platforms in Section II. Then it shows

in Section III how new technologies have turbo-

charged this business model and led to online

mobile platforms anchored by websites and

mobile apps. Section IV examines the implica-

tions of the online multi-sided platform business

model for the analysis of market power for at-

tention seekers. Section V offers some conclud-

ing observations.

II. The New Economics of Multi-

Sided Platforms

Although multi-sided platforms have ancient

roots economists came to understand them as

an important, and distinct type of businesses in

2000 when a now classic paper by Rochet and

Tirole began circulating.

Soon after, economists

began exploring the implications of the new

economics of multi-sided platforms for antitrust

issues.

As this work has become mainstream,

courts and competition authorities have grad-

ually absorbed the new learning and applied it

to cases.

A. Fundamentals of Multi-Sided Platforms

A multi-sided platform is called multi because

it provides a way for two, or more, types of par-

ticipants to get together. It is called a platform

because it typically operates a physical or vir-

tual place that enables these different types of

agents to interact. Each side of the platform

consists of the participants who have the option

of using the platform to connect. A shopping

mall is a physical platform. It provides a place

where shoppers and stores—the participants

7 Jean-Charles Rochet & Jean Tirole (2001) "Platform Com-

petition in Two Sided Markets," Working Paper, November 26, 2001.

An earlier version was in circulation in 2000.

8 David Evans, “The Antitrust Economics of Two-Sided

Markets,” Yale Journal of Regulation, Summer 2003, http://ssrn.com/

abstract=363160.

on the two sides—can connect. A ride-sharing

app is a virtual platform. It uses cloud-based

software, accessed through Internet-connect-

ed mobile phones, to match up drivers and

passengers who are the participants on the two

sides.

Multi-sided platforms typically reduce frictions

that get in the way of economic agents nding

each other, interacting, and exchanging value

on their own. Buyers and sellers, for example,

could nd each other in a variety of ways. A

marketplace, such as Flipkart in India, makes it

easier for them to nd each other through, for

example, posting tools for sellers and search

tools for buyers. It also makes it easier for them

to engage in a transaction through the use of

electronic payment methods and with con-

dence through Flipkart’s Replacement Guar-

antees and Seller Protection Fund.9 Multi-sided

platforms also create value by increasing the

odds that participants will nd counterparties

that generate value for value. An online dating

site, such as eHarmony, secures many women

and men thereby increasing the likelihood that

people will nd someone they would like to date

and perhaps even marry.

Multi-sided platforms face a chicken-and-

egg problem when they start as a result of what

they are trying to accomplish. Consider a plat-

form that is in the business of getting Type As to-

gether with Type Bs. Type As may not want to

consider the platform unless they know it has

attracted Type Bs, but Type Bs may not want to

consider the platform unless they know it has at-

tracted Type As. The platform has to gure out a

way to get both types of participants on board,

in sufcient numbers, to provide value to either.

When YouTube started, for example, it had trou-

ble persuading people to upload videos since

no one was coming to the site to watch them

9 Flipkart, “Returns and Cancellations” available at http://

www.ipkart.com/s/help/cancellation-returns; Flipkart, “Seller Hub:

GettingStarted”availableathttps://seller.ipkart.com/slp/faqs.

and trouble persuading people to come to the

site to view videos since there were few videos

to watch.

Typically, Type As value a platform if it has

more Type Bs and vice versa.

There are, in

economic terminology, positive indirect net-

work effects and positive feedback effects. A

platform that gets more Type As becomes more

attractive to more Type Bs, which in turn makes

it more attractive to more Type As, and so forth.

These positive feedback effects drive platform

growth. YouTube, for example, persuaded more

people to upload videos, more people came

to watch those videos, that got people more

interested in uploading videos, and that in turn

attracted more trafc to the site.

Positive indirect network effects can give big-

ger platforms economic advantages. These are

often limited in practice, however, by platform

congestion, or other diseconomies of scale, and

by platforms differentiating themselves on one

or more sides. In most countries, for example,

there are several competing payment card net-

works despite the positive feedback effects be-

tween cardholders accepting merchants and

despite scale economies in operating the net-

work. Mobile money platforms—where mobile

phones are used to send and receive money

and provide other nancial services—are evolv-

ing in the same way. More than 20 mobile wal-

let providers have started in India.

Based on

the experience of countries in Africa, where the

mobile money markets are more mature, we

10 For a detailed discussion of how they solved this problem

see Evans and Schmalensee, Matchmakers, Chapter 5.

11 As we discuss below ad-supported platforms may have pos-

itive externalities in one direction—advertising value more viewers but

viewers may not value more advertising.

12 Importantly, positive feedback effects work in reverse as we

discuss below. The loss of users on one side leads to losses of users on

the other side and so on. Positive feedback effects in reverse can result

in a death spiral.

13 See, http://letstalkpayments.com/wallet-wars-in-india-in-

tensies-with-uber-and-others-being-the-battleeld/

would expect the in the long run the market will

have several competing providers.

Multi-sided platforms differ fundamen-

tally from the traditional rms described in eco-

nomic textbooks and business school courses.

Traditional rms typically buy inputs, they make

products, and they sell those products to cus-

tomers. They operate along linear supply chain.

And since they do not have customers with in-

terdependent demands they are single-sided.

Multisided platforms sell participants in each

group access to the participants in each other

group. As a result, the customers are the main

inputs into providing the platform service. A typ-

ical retail store, which is a single-sided rm, buys

products from wholesale distributors or manu-

facturers and then sells them to customers. A

shopping mall, which is a two-sided rm, recruits

stores for its mall, and recruits shoppers to come

to its mall, and provides a platform where the

stores get access to the shoppers and the shop-

pers get access to the stores.

B. Pricing Structures and Strategies

The fact that the demand for one group de-

pends on the demand by the other group has

interesting implications for how multisided plat-

forms price their services. Platforms have to

choose prices that balance these demands.

Higher prices for Type As would discourage them

from participating in the platform. That would

deter Type Bs from participating in the platform

since they would have access to fewer Type A

participants. In fact, it may make sense to price

very low to one group of participants because

the other group will pay a high price for access

to them. That, in fact, is the secret behind ad-

vertising-supported media as we show below.

It could even make sense to subsidize one

group by charging them a price less than the

incremental cost of serving them, including let-

14 See GSMA, State of the Industry: Mobile Financial Ser-

vices for the Unbanked: 2014. Available at http://www.gsma.com/mo-

bilefordevelopment/wp-content/uploads/2015/03/SOTIR_2014.pdf

ting them use the platform for free, or even giv-

ing them rewards for participating. Economists

have shown that, as a matter of theory, plat-

forms may be able to maximize prots by sub-

sidizing one side of the platform in this way and

that, as matter of fact, many platforms have do

just that.

A popular restaurant reservation site

in the U.S., OpenTable, for example does not

charge people to make reservations with its site

and it gives them rewards that they apply to re-

duce the cost of their meals. Although “free”

is popular for online platforms it is by no means

universal. Dating sites, such as Trulymadly in

India and FarmersOnly.com in the US, charge

men and women the same. They contrast with

nightclubs which, in the US, have “Ladies Night

Free” pricing.

C. Advertising-Supported Platforms

Some multi-sided platforms connect consum-

ers and advertisers. This might seem odd since

in many cases consumers do not like advertis-

ing. They even spend money to avoid it by, for

example, buying DVRs that make it easy to skip

over ads and paying for alternative sources of

media, such as Pay TV, or ad-free versions of ser-

vices, such as Spotify Premium.

These platforms, however, have gured out

ways to connect consumers and advertisers in

ways that make both groups better off. They

typically offer valuable content to persuade

people consumer to come of their platforms

where these people are exposed to advertising

messages. Meanwhile they persuade advertis-

ers to pay for reaching these people. The view-

ers are the subsidy side of the platform and the

advertisers are the money side. So long as the

advertisers are willing to pay more for deliver-

ing messages to these consumers than the plat-

form spends on content the advertisers benet,

the consumers benet, and the platform makes

money.

15 See Evans and Schmalensee, Matchmakers, Table 2.1.

16 In fact this advertising supported media is a clever way of

One can think of ad-supported platforms as

buying eyeballs—usually by paying with valu-

able content—and selling those eyeballs to ad-

vertisers. The Internet has made that far easier

as we see next.

III. Online Multi-Sided Platforms

Online platforms have become more com-

mon and prominent participants in domestic

economies and some have rapidly become

global players. Many of these online platforms

provide free content or services to people to

attract their “attention” and then charge ad-

vertisers for delivering messages to these peo-

ple. These attention seekers engage in dynamic

competition in which they are constantly intro-

ducing new ways of attracting attention, and

copying methods used by others, to persuade

people to come to their platforms. Smart mo-

bile phones have accelerated the pace of dy-

namic competition, the frequency of disruptive

innovation, for online platforms.

A. The Technology Revolutions Behind On-

line Platforms

Several mutually reinforcing technologies,

and the businesses the make those technolo-

gies available, have made multi-sided platforms

increasingly powerful methods for reducing fric-

tions, and creating valuable new services, on a

global basis.

1. The PC-Web-Browser Revolution

The rst wave of innovation launched the

web-economy in the mid 1990s. The Internet

provided a physical network and standards for

solving the following exchange problem. Rahul would pay $20 to meet

Aditya. Aditya doesn’t like Rahul and would pay $5 to avoid him. Still

there is room for trade and an intermediary can make Aditya and Rahul

both better off. The intermediary pays Aditya $12 to meet Rahul and

charges Jose $14 for the introduction. Aditya is ahead $7 (-$5+$12),

Rahul is ahead $6 ($20-$14), and intermediary earns a prot of $2

(-$12+$14). In the case of advertising, instead of paying $14, the media

property provides entertainment or other content that Aditya values at

$14.

connecting computers around the world, the

Web provided a framework and software tech-

nologies for creating and linking content on

those computers, and the web browser provid-

ed an application for personal computers that

enabled people to consume Web content.

Businesses could use these technologies to

provide content and services on websites. The

cost of doing so was relatively low since it in-

volved writing software, using server computers,

and the small fees for connecting to the Inter-

net. And the company could reach an entire

country immediately and, in fact, much of he

world. Almost all the content, data, and pro-

cessing work resided in the cloud and consum-

ers accessed it through using a browser on their

Internet-connected personal computers.

The number of web-based businesses and In-

ternet trafc exploded following the launch of

the commercial Internet in the 1990s. A num-

ber of global online platforms emerged such as

Amazon, eBay, Facebook, Google, PayPal, and

Yahoo. This growth was made possible by the

development and expansion of increasingly fast

broadband delivered over xed wires such coax-

ial cable, ber optic line, or even a copper wire.

2. The Mobile-App Revolution

Mobile phones were in widespread use in the

U.S. and other countries by the late 1990s. Cel-

lular networks, however, were not able to carry

enough data fast enough for people to use the

Internet from their mobile phones. Innovations

in cellular technology starting in the mid 1990s

increased the potential capacity and speed of

cellular networks and mobile devices for mak-

ing better use of these faster more capacious

broadband technologies. Anticipating the roll

out of mobile broadband a number of com-

panies started investing in developing various

components of smart phones, including mo-

dem and processing chips, operating systems,

and handsets in the early to mid 2000s.

Innovations by Apple and Google, in particu-

lar, have led to spread of smart mobile phones

around the world, enabling billions of people

to consumer Internet-based services and mil-

lions of businesses to provide mobile-app based

services to them. Apple introduced the iPhone,

which consisted of a powerful computer, a mo-

bile operating system, and a standard set of

applications including a mobile browser in June

2007. Google invested in developing a mobile

operating system, Android, which it ran as an

open-source project, and developing and or-

ganizing an ecosystem of handset makers, mo-

bile network operators, and other technology

partners. It introduced the rst Android phone in

October 2008.

Apple and Google also stim-

ulated the production of mobile apps by pro-

viding software tools for developing apps for

their operating systems, creating a quality cer-

tication process for these apps, and creating

“app stores” that provided centralized places

for developers to distribute apps and for users to

download them on their mobile devices.

Smart mobile phones changed the online

game in a number of ways as they became

widely adopted, millions of apps became avail-

able for them, and faster and more capacious

mobile broadband networks were rolled out

around the world. People could access the

Internet anywhere and anytime using smart-

phones running on mobile broadband networks.

More people could do that because mobile

phones and data plans were much cheaper

than buying PCs and xed broadband connec-

tions. Businesses could reach billions of people

by developing mobile apps and distributing

them in apps stores. Apps could exploit the GPS

capabilities of phones, which make it possible

to know where individuals are in physical space.

This, together with the related development of

the “Internet of Things” is leading to the deep

integration of the online and physical worlds.

17 Kent German, “A Brief History of Android phones,” August

2, 2011, http://www.cnet.com/news/a-brief-history-of-android-phones/.

3. The Movement from PCs/Browsers to Mo-

bile/Apps

Businesses that want to provide online ser-

vices, and consumers who want to consume

online services now have several choices. App

developers can develop websites that people

can visit from browsers on their PCs or from their

mobile devices. They can develop mobile apps

that people use on their mobile phones or mo-

bile browser-apps that try to mimic these apps.

Different businesses have adopted different ap-

proaches depending on the content and ser-

vices they are providing. Consumers have, how-

ever, shifted their use dramatically from PCs to

mobile devices and from using websites to using

apps.

Consider the US. Between 2008 and 2015 the

proportion of time spent online using mobile de-

vices increased from 12.7 percent to 54.6 per-

cent. Commerce has moved dramatically from

PCs to mobile. Americans made 57 percent of

their online purchases from mobile devices in

2014 compared with likely none before 2010.

On Thanksgiving Day, November 26, 2015,

around 60 percent of US website visits were

made from mobile devices in the US.

Advertis-

ing has moved to mobile in response. Facebook

earned 78 percent of its global advertising reve-

nue from mobile in 2015Q3

compared with 14

percent in 2012Q3.

These trends are expected

to continue.

18 David Murphy, “IBM: Christmas Day Sales Up 8.3 Percent,

Mobile Purchases up 20.4 Percent,” PC Magazine, December 26, 2014,

http://www.pcmag.com/article2/0,2817,2474217,00.asp.

19 Hiroko Tabuchi, “Black Friday Shopping Shifts Online as

StoresSeeLessFootTrafc,”NewYorkTimes,November27,2015,

http://www.nytimes.com/2015/11/28/business/black-friday-shopping-

shifts-online-as-stores-see-less-foot-trafc.html?_r=0.

20 Facebook Inc., “10-Q for Period Ending September 30,

2015,” p. 40.

21 Facebook Inc., “10-Q for Period Ending September 30,

2012,” p. 27.

22 Chantal Tode, “M-Commerce Sales to Reach $142B in

2016: Forrester,” Mobile Commerce Daily, October 8, 2015, http://

On mobile devices people typically access

Internet-based services using mobile apps rath-

er than using websites with their mobile browser.

Mobile apps accounted for nearly 90 percent

of the time Americans spend using mobile apps

or browsers on their mobile devices.23 As a re-

sult the proportion of time people spend online

using mobile apps has increased from what was

likely a very low level in 2008 to 54 percent in

2015.24 This share is likely to increase further as

the shift from PCs to mobile continues and as

the shift from browser-based to mobile app-

based delivery continues.

Many countries have had low penetration of

PCs and xed broadband because of their ear-

ly stages of economic development. The adop-

tion of smart mobile phone and mobile broad-

band are increasing rapidly in those countries

because it is cheaper and even more rapidly in

the faster growing ones. More than 90 percent

of Facebook's Indian users

and 60 percent of

Amazon’s Indian users

access it through mobile

www.mobilecommercedaily.com/mcommerce-sales-to-reach-142b-in-

2016-forrester; Matthew Hobbs, “Internet Advertising,” 2015, http://

www.pwc.com/gx/en/industries/entertainment-media/outlook/seg-

ment-insights/internet-advertising.html.

23 Simon Khalaf, “Seven Years into the Mobile Revolution:

Content is King … Again,” Flurry Insights, August 26, 2015, http://

urrymobile.tumblr.com/post/127638842745/seven-years-into-the-mo-

bile-revolution-content-is; https://www.comscore.com/Insights/Presen-

tations-and-Whitepapers/2015/The-2015-US-Mobile-App-Report.

24 comScore, “The 2015 U.S. Mobile App Report,” Sep-

tember 22, 2015, https://www.comscore.com/Insights/Presenta-

tions-and-Whitepapers/2015/The-2015-US-Mobile-App-Report.

25 Total time spent on digital media using mobile apps in-

creased at a compound annual growth rate of 38 percent per year be-

tween 2013 and 2015, compared to 7 percent for desktops and 24 per-

cent for mobile browsing. The share for mobile apps increased from

43 percent to 54 percent over this period, an increase of 11 percentage

points, or a compound annual growth rate of 12 percent. Data are not

available back to 2008.

26 BGR, “90% of Facebook’s 132 million users from India

come from mobile phones,” September 27, 2015, available at http://

www.bgr.in/news/90-of-facebooks-132-million-users-from-india-

come-from-mobile-phones/

27 AshwiniGangal,“’Over60percentofourtrafccomes

through mobile’: Manish Kalra, Amazon India,” August 28, 2015,

http://www.afaqs.com/interviews/index.html?id=469_Over-60-per-

cent-of-our-trafc-comes-through-mobile-Manish-Kalra-Amazon-In-

dia

devices. In 2014, leading Indian e-commerce

companies, including Flipkart and Snapdeal,

derived the majority of their gross merchandise

value from mobile devices.

B. Overview of Online Multi-Sided

Platforms

The development of online technologies has

made it cheaper and easier to reduce frictions

through multi-sided platforms and to do so over

large geographic areas. The Internet makes

it possible to connect participants over wide

geographic areas and in principle from around

the world. Software programs running on high-

speed computers in the cloud provide power-

ful technologies for nding good matches and

consummating exchanges. Mobile has extend-

ed these capabilities throughout the day and

throughout physical space.

Almost immediately after web commerce

became viable in the mid 1990s entrepreneurs

started using the new technologies to start

multi-sided platforms. Not everyone chose a

multi-sided model. Amazon, for example, start-

ed with a typical retail model in which it bought

products, initially books, wholesale and sold

them to people through its online store. Many,

though, used a multi-sided approach often be-

cause it was the only way to provide the prod-

uct or service. eBay started an online market-

place for buyers and sellers, match.com started

an online matchmaker for men and women,

and Yahoo started a online portal that used

content to attract viewers and then attracted

advertisers who wanted to reach those views.

Many of the established platforms followed

the shift from the PC-browser-centric model to

the mobile-app centric model. Entrepreneurs,

however, discovered that the mobile-app cen-

28 BGR, “Smartphone shopping to contribute up to 70 per-

cent of total revenue in online shopping: Experts,” November 30, 2014,

available at http://www.bgr.in/news/smartphone-shopping-to-contrib-

ute-up-to-70-percent-of-total-revenue-in-online-shopping-experts/.

tric model provided new opportunities. Uber,

for example, has built a business that connects

drivers and riders in real-time and in physical

space using mobile apps.

Tables 1 and 2 provide an overview of online

multi-sided platforms based on their presence

in the US, which reects global platforms, and

India, which reects domestic platforms and

global ones. In each country we have select-

ed 20 platforms. We include the largest ones

based on the number of times over the space

of a month people clicked on pages on those

sites (“pageviews”). That is a particularly useful

measure for content-oriented sites. We have

erred on the side of showing diversity of online

platforms and the table is not intended to be an

accurate summary of the economically most

important online platforms. In each case we

summarize the multi-sided business model and

the extent to which one side receives service for

free.

As these tables show online platforms are

highly diverse. However, they often have sever-

al of the following features that are relevant for

antitrust analysis. First, they are all based on soft-

ware. They can add new features, and intro-

duce new products and services, by modifying

or adding software code and related databas-

es. That is much different than physical plat-

forms. Second, the marginal cost of participants

to software-based platforms running in the cloud

is virtually zero. That increases the normal ten-

dency of multi-sided platforms to allow a group

of participants to use the platform for free. Third,

dynamic competition is more intense for online

platforms because technological change has

reduced the capital cost of starting a platform

and the software-based nature of these plat-

forms makes it easier for platforms to offer new

products and services in competition with oth-

er platforms.29 Fourth, dynamic competition is

29  Case No Comp/M.6281 - Microsoft/Skype, Ofce of the

Publications of the European Union, July 20, 2011, http://ec.europa.

also more intense for online platforms because

the participants have lower switching costs,

and face less lock-in, than on physical platforms

where they often have to make costly sunk-cost

commitments to the platform. Fifth, online plat-

forms are in the midst of a massive technolog-

ical shift resulting from the move of consumers

from the PC-browser to the mobile-app centric

way of using online services.30 These points are

especially true one of the largest categories on

online platforms.

C. Online Attention Seekers

t is has turned out many online platforms

make money primarily by helping businesses

sell things to consumers through advertising and

marketing.

As we discussed above the way

they do this is simple but clever. They provide

reasons to consumers to come visit them by of-

fering engaging content or services valued by

consumers. Consumers typically do not pay for

obtaining the content or services. They are free

in that sense. But consumers are receiving val-

ue by coming to these platforms. In that sense

the real price of participating in the platform is

even better than free, it is negative, so that plat-

form is paying consumers to come visit. Once

they have gotten consumers to spend time of

the platform they allow businesses to present

advertising or other marketing messages to con-

sumers. They charge businesses for this and that

eu/competition/mergers/cases/decisions/m6281_924_2.pdf ; “FTC

NotiesFacebook,WhatsAppofPrivacyObligationsinLightofPro-

posed Acquisition,” Federal Trade Commission, April 10, 2014, https://

www.ftc.gov/news-events/press-releases/2014/04/ftc-notifies-face-

book-whatsapp-privacy-obligations-light-proposed; and Case No

COMP/M.7217 – Facebook/WhatsApp, Ofce of the Publications of

the European Union, March 10, 2014, http://ec.europa.eu/competition/

mergers/cases/decisions/m7217_20141003_20310_3962132_EN.pdf.

30 See Hemant Bhargava, David S. Evans, and Deepa Mani,

“The Move to Smart Mobile and its Implications for Antitrust Analysis

of Online Market In Developed and Developing Countries,” Forthcom-

ing.

31 David Evans, “Attention Rivalry Among Online Platforms,”

Journal of Competition Law & Economics, May 14, 2013, Volume 9 Is-

sue 2:313-357, http://jcle.oxfordjournals.org/content/9/2/313.abstract.

is how they cover their costs and make prots.

Online attention seekers compete to get the

attention of consumers and then sell portions of

that attention to businesses that aren’t able to

get it easily on their own. They seldom make any

money directly from providing content or ser-

vices to consumers. Recognizing this is important

for understanding the dynamics of competition.

Entrepreneurs compete to come up with clever

ideas for attracting eyeballs—say by inventing

tweeting or pinning—not so they can charge

people for clever content or services they are

providing but so they can sell access to those

eyeballs to advertisers. Attention seekers may

come up with ways to differentiate themselves

from the standpoint of attracting consumer at-

tention and selling advertising. But overall they

are competing to attract a limited pool of at-

tention and advertising and marketing budgets

to reach those consumers. Now consider the

ve features that we highlighted above.

Attention seekers are all built on software

platforms. They do not have printing presses, ca-

ble networks, or radio towers. When they want

to add features to the platforms they hire soft-

ware engineers to write code. They can often

make changes quickly and roll those changes

out globally. It took about 5 months, for exam-

ple, for Facebook to develop Facebook Mes-

senger which is one of the leading apps for

smartphones.

The marginal cost of another participant on

an attention seeker is essentially zero. Google

does not incur any signicant out of pocket

cost when a person conducts another search

or when it puts another search ad on a search

results page. That is true for virtually all attention

seekers with the exception of some, such as

Pandora, that have to pay for the content they

deliver.

The capital cost of starting an attention seek-

er is low and that has intensied dynamic com-

petition. That is more so true now as a result of

mobile apps. The founders of WhatsApp had

to write software code so that messaging app

would work for Apple and Android phones and

for the cloud-based service those apps were

connected with.

Once they did that they

had a platform that could provide messaging

services globally to unlimited number of users

with the addition of some cheap server capac-

ity. Many other mobile messaging apps have

started. They compete with older messaging

PC-based messaging apps as well as the new

mobile-based ones.

It is easy for consumers to reduce the amount

of attention they provide one platform, or drop

it altogether, and increase the amount of at-

tention they provide another platform. Since

the platforms are free they can use as many as

they want and switch their attention depend-

ing upon the relative attractiveness to spending

time on one or the other. The consumer bears

no cost from shifting time from looking at Yahoo

to looking at Flipboard. While some online plat-

32 Facebook, “Building Facebook Messenger,” August 12,

2011, available at https://www.facebook.com/notes/facebook-engineer-

ing/building-facebook-messenger/10150259350998920/.

33 One estimate is that it would cost about $250,000 and take

about nine months to build a robust version of an app like WhatsApp.

See Courtney Boyd Myers, “How much does it cost to build the world’s

hottest startups?” TNW News, December 2, 2013. Available at http://

thenextweb.com/dd/2013/12/02/much-cost-build-worlds-hottest-start-

ups/#gref

forms involve some cost of switching in practice

it does not limit people from doing so. In the

case of social networks, Americans switched

from Friendster to MySpace and then from MyS-

pace to Facebook.

People in other countries,

such as Brazil and India, switched from Orkut to

Facebook.

Finally, the shift of consumers from looking

at websites with their browsers to using apps on

their mobile phones has resulted in dramatic

changes in attention seeking platforms. There

has been a dramatic increase in the amount

of online attention available as a result of peo-

ple being able to go online with their mobile

devices for much more of the day. The oppor-

tunities for connecting businesses with consum-

ers have also changed now that people carry

mobile phones all the time and in particular

when they go shopping. Search is one of the

attention-seeking businesses that is undergo-

ing disruption as a result of this.

Search en-

gines index websites and allow people to nd

things on those websites. But now an enormous

amount of online activity is happening with

mobile apps. At this point it is unclear how

people will be able to nd app-based content

and what companies will ultimately succeed in

doing so. Apple, Facebook, and Google are

among the companies that are trying to gure

this out.

What’s should be clear from the discussion so

far is that multi-sided platforms are governed by

different rules than traditional linear businesses

and that competition among online platforms

is often more intense and more dynamic than

among physical platforms. Both point have im-

34 Evans and Schmalensee, Matchmakers, Chapter 9.

35 Elena Trost, Social Media Marketing in BRIC Countries

(Zurich, Lit Verlag GmbH & Co., 2013), Chapter 3.

36 ErinGrifth,“Facebook,Googleandthebattleformobile

intent,” September 8, 2015, available at http://fortune.com/2015/09/08/

facebook-google-mobile-search-advertising/

37 ErinGrifth,“Facebook,Googleandthebattleformobile

intent,” September 8, 2015, available at http://fortune.com/2015/09/08/

facebook-google-mobile-search-advertising/

portant implications for antirust analysis.

VI. Market Power Analysis of Online

Attention Seekers

Economists typically assume that the de-

mand for a product depends on the price of

that product, the price of substitute products,

and the price of complementary products. The

demand for a particular brand of beer, for ex-

ample, depends on the price of that brand, the

prices of other kinds of beer and other alcoholic

beverages, and perhaps the demand for nuts,

chips, and other things that people eat with

beer. Most economic theories relied on in anti-

trust analysis, such as those involving predatory

pricing, and economic tools, such as SSNIP tests,

are based on this model of product demand.

All of those factors are relevant for consider-

ing the demand for product and services pro-

vided by multi-sided platforms. But those stan-

dard factors do not include the most critical

factor that drives the demand for platforms. The

demand by members of one group of custom-

ers, say Type A, depends, roughly speaking, on

the participation of the other group of custom-

ers, say Type B, in the platform.

To avoid be-

ing mathematically wrong and unreliable, eco-

nomic models and tools must account for the

interdependent demand and consider all sides

of the platforms. The fact that the demands by

the various groups of platform participants are

interdependent also means that analyses that

focus on one group of participants in isolation

are not correct as a straightforward mathemat-

ical matter.

38 More precisely, platform customers care about the likeli-

hood that they will be able to enter into valuable exchange on the plat-

form; we are using the number of potential trading partners as a short-

hand for describing all of the characteristics of one side of the platform

that affects the demand by the other side.

39 David Evans and Richard Schmalensee, “The Antitrust

Analysis of Multi-Sided Platform Businesses,” January 30, 2013, http://

papers.ssrn.com/sol3/papers.cfm?abstract_id=2185373; Roger Blair

and Daniel Sokol, eds., Oxford Handbook on International Antitrust

Economics, Oxford University Press, 2015; University of Chicago

Antitrust analysis needs to examine the plat-

form overall taking these interdependencies

into account.

Generally, that requires treating

the platform as a whole, rather than focusing on

one group of customers or another, or at least

carefully considering the inter-linkages between

these groups. Platform competition tends to

force overall prices down and reduces the prof-

its the platform can earn. Typically, though, it

does not force prices down to incremental costs

for all, or even any, sides of the platform. Even

with competition platforms may choose to sub-

sidize one side of the platform and make prots

for other sides of the platform.

The magazine business, for example, is high-

ly competitive yet most magazines subsidize

readers; the cover price for the magazine often

does not cover printing and distribution costs let

along the cost of the content that attracts read-

ers. In fact, competition to attract participants

to the platform can result in greater subsidies

to one side. For example, in the U.S., compe-

tition among payment card networks appar-

ently resulted in bidding up payments (called

interchange fees) to banks that issue cards to

consumers.

As a result, evidence that price

Institute for Law & Economics Olin Research Paper No. 623. Avail-

able at SSRN: http://ssrn.com/abstract=2185373; David Evans, “The

Consensus Among Economists on Multisided Platforms and its Impli-

cations for Excluding Evidence that Ignores It,” Competition Policy

International, (April 13, 2013). Available at SSRN: http://ssrn.com/ab-

stract=2249817 or http://dx.doi.org/10.2139/ssrn.2249817.

40 In Cartes Bancaires v. European Commission, the European

Court of Justice concluded that to analyze competitive effects it was

necessary to consider the two interlinked sides of the platform. See

Groupement des cartes bancaires v European Commission, Judgement

of the Court, September 11, 2014. In, Qihoo 360 v. Tencent, the Chinese

Supreme People’s Court found that it was necessary to consider plat-

form competition in evaluating market power. See, David Evans and

Vanessa Zhang, “Quhoo 360 v Tencent: First Antitrust Decision by the

Supreme Court,” October 21, 2014, https://www.competitionpolicyin-

ternational.com/qihoo-360-v-tencent-rst-antitrust-decision-by-the-su-

preme-court; Charles Rivers Associates, “Qihoo v. Tencent: economic

analysisoftherstChineseSupreme

Court decision under Anti-Monopoly Law” February 2015, available at

http://www.crai.com/sites/default/les/publications/China-Highlights-

Qihoo-360-v-Tencent-0215_0.pdf

41 OECD, “Competition and Payment Systems,” June 28,

2013, http://www.oecd.org/competition/PaymentSystems2012.pdf.

is great than incremental cost on one side pro-

vides no meaningful evidence that the platform

has market power and evidence that the plat-

form charges a price less than marginal cost on

another side provides no meaningful evidence

that the platform is engaging in predatory pric-

ing. The analyst needs to look at the platform

overall to assess market power and predation.

In practice, it often makes sense to look at pric-

ing and competition on both sides but then ac-

counting for the interdependencies.

This section applies these general principles

to the analysis of market power for online atten-

tion seekers which is one of the most important

categories of online platforms.

A. Free and Feature Competition

Traditional antitrust analysis assesses market

power by considering whether the rm can in-

crease price protably. That approach does

not make any economic or business sense for

online attention seekers. The business is based

on paying consumers to use the platform and

charging advertisers for access to those con-

sumers. An exercise of market power over

consumers would could involve increasing the

price to them but, more likely, would involve re-

ducing the quality of the content and services

the platform is providing to attract their atten-

tion.

Whether that reduction in quality is prof-

itable depends on the extent to which it would

decrease the attractiveness of the platform to

advertisers. A platform could consider reduc-

ing its expenditures on quality improvements

by $1 million. Whether this is protable depends

on whether the lower quality would reduce the

amount of advertising, given the lower atten-

tion it attracts, by less than $1 million.

This highlights the importance of feature and

42 The decision by online attention seekers to charge fees is

quite rare even for ones that are highly successful. Some online news-

papers have tried paywalls with mixed success.

quality competition. Online attention seekers do

not compete based on price. Therefore, to assess

market power, one needs to assess the extent to

which a lower provision of quality would divert at-

tention to other online platforms. In considering

that diversion there is no business or economic

reason to limit the inquiry to online platforms that

provide the same service. It is an empirical ques-

tion whether consumers would turn their atten-

tion to completely different services.

In practice market power analysis for online

attention seekers can consider substitution pos-

sibilities by considering a small but signicant

increase in price or a small but signicant de-

crease in quality. Either one reduces the value of

the platform for users and could induce switch-

ing. The SSNIP, however, must consider small ab-

solute increase in price since a percentage in-

crease is undened when the initial price is zero.

The Chinese Supreme People’s Court, in Qihoo

360 v. Tencent, decided that the SSNIP evidence

was not relevant and considered informally how

consumers would react to small but signicant

decreases in quality (SSNDQ) of the instant mes-

sage products under consideration.

Since attention makers make virtually all of

the revenue and prot from advertisers the other

issue concerning market power is whether they

can take actions that increase the price of ad-

vertising above competitive level. The analysis

of that question needs to consider the extent to

which advertisers can get the attention of con-

sumers in other ways and the extent to which

the online platform offers some consumer atten-

tion, perhaps based on demographic proles or

the context in which they’ve captured that at-

tention, for which there are limited substitutes.

Free pricing, however, shouldn’t be ana-

lyzed in isolation. In fact, the existence of con-

sumers being offered something for nothing is

43 See, David Evans and Vanessa Zhang, “Quhoo 360 v Ten-

cent: First Antitrust Decision by the Supreme Court,” October 21, 2014,

https://www.competitionpolicyinternational.com/qihoo-360-v-tencent-

rst-antitrust-decision-by-the-supreme-court.

almost always an indication that the business is

a multi-sided platform. That means that the de-

mand by consumers on the “paid side” is linked

to the demand by consumers on the “free side”

to the demand. The SSNIP and SSDNQ analy-

ses should account for the interdependencies

of demand for taking a holistic approach, and

considering the platform overall, or by carefully

considering the linkages in demand and their

implications for competitive constraints.

B. New Entry, Cross-Category Entry, and Fea-

ture Competition

Market power analysis needs to consider the

ease of entry and of feature competition for on-

line attention seekers. As discussed above the

capital cost of entry for online attention seekers

is low. The main difculty is attracting consumers

to the platform with persuasive content and ser-

vices. Importantly, though, the analysis needs

to at least consider the impact on the platform

of entry by completely different services. For

example, suppose Facebook reduced its in-

vestment in the quality of its social networking

platform. It could lose advertising revenue in

part because that increases the likelihood that

consumers will more likely to shift attention to

“the next new thing”—not necessarily to a so-

cial network—and that will cost the company

advertising revenues. In addition, market power

analysis needs to consider entry from other cat-

egories. Because it is easy to change features

through software online attention seekers can

add features that mimic those of other very dif-

ferent attention seekers. Twitter and Pinterest,

for example, have both recently introduced

“buy buttons” that help businesses make sales

on their platforms, like Amazon Marketplace, in

addition to just advertising to those consumers.

That feature competition is an example of dy-

namic competition which we turn to next

B. Dynamic Competition

Dynamic competition has characterized on-

line attention seekers for the last twenty years

and shows no signs of abating. Attention seek-

ers have no guarantee that they can hold onto

consumers without engaging in persistent incre-

mental feature and disruptive innovation. We

see this in a variety of ways.

First, the relative importance of attention

seekers changes dramatically over time.

Table

3 shows the 20 largest advertising-supported at-

tention seekers by time spent on the webpage

in 2002, 2007, and 2012. Pinterest (8) is a US ad-

vertising supported webpages that users spent

the most time visiting during September 2012 did

not exist in September 2007, while several web-

pages were in the early stages of development

including Facebook (1), Youtube (2), The Huff-

ington Post (9), and Tumblr (10). This illustrates

how quickly and dramatically the landscape for

online advertising can change.

Second, successful attention seekers have

declined and in some cases failed when they

have not kept up, while new ones have risen

quickly. Orkut was the dominant social network-

ing site between 2005 and 2010 in India.

Face-

book overtook it in July 2010.

MySpace had a

similar experience in the US where it was the larg-

est between 2005 and 2009 and also displaced

by Facebook.

Yahoo was a highly successful

attention seeker for many years. While it still at-

tracts a large number of pageviews the market

44 See David Evans, “Attention Rivalry Among Online Plat-

forms,” Journal of Competition Law & Economics, May 14, 2013, Vol-

ume 9 Issue 2:313-357.

45 Sahil Shah, “Social Networking War in India: Facebook

vs Orkut,” January 25, 2011, https://www.techinasia.com/indian-so-

cial-networking-wars-facebook-vs-orkut-2

46 comScore, “Facebook and Orkut Growth in India,” Novem-

ber 4, 2010. http://www.comscore.com/Insights/Data-Mine/Facebook-

and-Orkut-Growth-in-India

47 Pete Cashmore, “MySpace, America's Number One,” July

11, 2006, http://mashable.com/2006/07/11/myspace-americas-num-

ber-one/#tqA37Md.SgqA; Choloe Albanesius “Home/News & Analy-

sis/More Americans Go To Facebook Than MySpace

More Americans Go To Facebook Than MySpace,” June 16, 2009,

http://www.pcmag.com/article2/0,2817,2348822,00.asp.

value of the portion of advertising-supported

portion of the business is negligible according to

various reports

Third, mobile apps have provided opportu-

nities for the creation of new attention seekers

and have reduced the relative importance of

incumbent attention seekers. Facebook, for ex-

ample, has become one of the largest online

advertising platforms in the world through its

success in attracting attention of mobile device

users and selling that attention to advertisers. It

now provides three of the ten mobile apps that

attract the largest number of page views.

Tra-

ditional search advertising, while still important

on mobile, is much less signicant than it is on

the web.

C. Market Shares as Indicia of Market

Power

A number of commentators have pointed

out that market shares must be used with care in

assessing market power.

This advice is particu-

larly sound when it comes to measuring market

power on the consumer side of online attention

platforms. In traditional markets sound practice

involves measuring market shares based on val-

ue to account for quality differences between

products. It also makes sense to focus on price

because it is an important dimension of com-

petition. Most online attention seekers do not

charge consumers for using the platform. Price

48 Steven Levy, “Yahoo and Alibaba: Joined at the Balance

Sheet,” March 3, 2015, https://medium.com/backchannel/yahoo-

and-alibaba-joined-at-the-balance-sheet-94b459233894#.cklylx3x3;

Lawrence Meyers, “Yahoo Stock: Is YHOO Worth Nothing Without

BABA?,” September 21, 2015, http://investorplace.com/2015/09/ya-

hoo-stock-yhoo-baba-alibaba/#.VnNaiPkrKM8.

49 comScore, “comScore Reports July 2015 U.S. Smartphone

Subscriber Market Share,” September 3, 2015

50  Louis Kaplow, “Market Denition, Market Power,” May

2015, http://papers.ssrn.com/sol3/papers.cfm?abstract_id=2605179##

; Jonathan B. Baker and Timothy F. Bresnahan, “Economic Evidence

in Antitrust: Dening Markets and Measuring Market Power,” Eco-

nomic Evidence in Antitrust, http://web.stanford.edu/~tbres/research/

buccirossi_01_ch01_001-042.pdf; Howard H. Change, David S. Ev-

ans, and Richard Schmalensee, “Assessment of the Relevant Market in

Competition Matters,” March 30, 2011.

is therefore not available as a measure of qual-

ity differences and for that matter is not an im-

portant element of competition relative to the

content and service subsidies.

Market shares are poor indicia of market

power for online attention seekers in part be-

cause precise market boundaries are more dif-

cult to establish. Narrow market denitions, con-

ned to functional substitutes for the content or

services provided by the platform, seldom make

sense because consumers shift their attention

uidly among different platforms. That is not to

say that a broad denition is appropriate either

since many platforms have some source of dif-

ferentiation that makes consumers more likely to

give them their attention. To the extent market

shares are used they should be calculated using

different plausible denitions of the relevant set

of substitutes.

V. Conclusion

Multi-sided platforms comprise an increas-

ingly large portion of the economy, in part as a

result of the technological changes described

above. Online multi-sided platforms are now

behind waves of creative destruction. Protect-

ing competition in this part of the economy is

important and competition authorities should

be commended for being vigilant in making

sure that dominant platforms do not violate the

competition rules and that rent-seeking incum-

bents do not stand in the way of innovative new

platforms.

Antitrust analysis, however, needs to adjust

the standard tools for assessing market power

so that they are accurate, as a matter of eco-

nomics and mathematics, for multi-sided plat-

forms. That includes recognizing that important

implications of interdependent demand, and

interlinked sides, for platforms. Particular care

is needed to online platforms, and especially

online attention seekers, because of the impor-

tance of non-price competition, the pervasive

use for zero prices, and the role, at least for now,

of intense dynamic competition

The double duality

of two-sided

markets

Alfonso Lamadrid de Pablo

Introduction

The increasing relevance of multi-sided mar-

kets

and business models in the economy has

over the past few years been mirrored in aca-

demic writings, mostly in economic literature,

and increasingly in competition law enforce-

ment.

The intention of this brief intervention is not to

incorporate novel theories into the discussion of

multi-sided platforms nor to summarise the main

ndings of the literature that is currently avail-

able. As an avid reader of academic works

on the subject, and although I much appreci-

ate their lessons, when I read them I realise that

the vast majority of papers have been authored

by economists, mostly academics, and only in

very rare cases by lawyers in private practice.

This – like other features I will comment on lat-

er – has dual implications: on the one hand it

means that practitioners haven’t (yet) muddied

the discussion by writing one-sidedly in defence

of the positions they are hired to represent;

the other hand, it also means that certain prac-

tical legal issues may not have received the at-

tention they perhaps should.

When legal scholars have touched upon the

application of competition law in two-sided

platforms they have moreover done so for the

most part in relation to specic markets, notably

payments, media and search engines. There

is nothing to criticise this focus, but while

specicity has advantages, it also has down-

sides. Indeed, in my view, complex problems

are better assessed with perspective; it is only

with a wider approach that patterns become

clear and that conclusions intended to be of

general application can be adopted without

inuence or prejudice derived from fact, case,

or market-specic elements.

The relative lack of attention on the part of

legal scholars has not been compensated by

any clarication by competition authorities. In-

deed, the majoritarian position of competition

authorities has been one that at rst sight may

appear as prudent, but that on closer inspec-

tion may not be proving the wisest: to argue that

Garrigues, Brussels. The following pages are an edited tran-

script of the authors’ intervention at the Swedish Competition Authori-

ty’s 2014 Pros and Cons Conference in Stockholm. The presentation that

accompanied the oral intervention is available at: https://antitrustlair.

les.wordpress.com/2014/11/lamadrid_the-double-duality-of-two-sid-

ed-markets.pdf. I am most grateful to Pablo Ibañez Colomo and Kevin

Coates for their comments on a previous version of this paper, and to

Sam Villiers and Miguel Ángel Bolsa for their help with the editing

work.

2 Whereas the title of my intervention at the Pros and Cons

conference referred to ‘two sided markets’, I will hereinafter refer to

‘multi-sided platforms’ in order to avoid misunderstandings with the

competition law notion of ‘market’ as well as to acknowledge that plat-

forms may have more than only two sides.

3 For a survey of economic literature on the topic, see D Ev-

ans and R Schmalensee, ‘The Antitrust Analysis of Multi-Sided Plat-

form Businesses’, in R Blair and D Sokol (eds), Oxford Handbook on

International Antitrust Economics (Oxford University Press, 2015)

or University of Chicago Institute for Law / Economics Olin Research

Paper No 623, available at:http://chicagounbound.uchicago.edu/cgi/

viewcontent.cgi?article=1482&context=law_and_economics.

4 Some of the articles written by practicing lawyers (and

practicing economists) in this regard are indeed so one-sided that it

is surprising to see them written on both sides of the paper. Contrary

to this tradition, this article does not intend to defend the particular

position of a given client; I have, rather, chosen to adopt a different

forward-looking approach and present both the ‘pros and the cons’ of

market structures and business practices in multi-sided settings. This

balancing exercise is not only in line with the theme of the Pros and

Cons conference, it also should also have the positive externality of

lowering my switching costs should that be necessary.

the economic literature is still at an early stage,

that there is little empirical work from which to

draw lessons and, in sum, that more economic

research is needed prior to advancing changes

in the way the law is applied.

Against this background, the pages that

follow seek to provide the personal views of a

practitioner on how to deal with a subject that

has become increasingly relevant to the prac-

tice of competition law and that lies at the core

of some of the most prominent cases in recent

times.

I essentially intend to submit that – con-

trary to the most widely held stance – perhaps

we know all we need to know about two-sided

platforms in order to rene our legal approach

tothem.

Indeed, ‘unlike, say, macroeconomics or

behavioral economics, there is no serious con-

troversy among economists’ on this topic and

therefore it seems fair to claim that ‘the multisid-

ed platform analysis is well within the economic

mainstream’;

over the past few years thanks to

the work of many economists we have robust

theoretical and empirical grounds on which to

build, these theories already have their Nobel

prize,

and perhaps the time is ripe for the law

to take the driver’s seat in these discussions.

My concern, however, is that we, as lawyers

and jurists, seem not to know what to do with

it. Indeed, authorities and lawyers are used to

(let us not change metaphors)driving in autopi-

lot, repeatedly resorting to the same tools, tests

and rules, and feel uncomfortable in multisid-

ed platforms because the setting forces us to

go back to basics and to interrogate ourselves

about where we really want the application of

competition law to take us.

In other words, by breaking the inertia of busi-

ness as usual, multi-sided platforms place us out

of our comfort zone, expose our contradictions

and insecurities and oblige us to think. This may,

on the one side, be most uncomfortable but, on

the other side, presents us with a most interest-

ing opportunity to go back to the basics of our

discipline, perhaps too often forgotten.

In sum, I will argue that what is missing is not

empirical work but a wider reection on the

goals of competition law and on how they are

to be attained.

The complexity and duality of

multi-sided scenarios

On the need to rene traditional

tools and rules

It already has become commonplace to say

that multi-sided platforms pose particular chal-

lenges to competition law enforcement, and it

is true in many ways that the logic, the rules and

the tools we are accustomed to are not valid in

these settings, at least not without importantre-

nements.

Such claims are not unusual. As a lawyer, I

5 See for example European Commission, ‘Note to the

OECD’s Roundtable on Two-Sided Markets’, 28 May 2009, p 5, stat-

ing that ‘empirical research is lacking’ and is ‘indispensable’ and that ‘it

isstillearlyforacompetitionauthoritytoadoptanydenitiveviews,

let alone concrete policies or assessment methodologies, concerning the

application of competition policy un cases involving two-sided plat-

forms’.

6 These include various investigations in Google’s search and

mobile OS products, several investigations into payment networks as

well as on Most Favoured Nation clauses in online websites.

7 D Evans, ‘The Consensus among Economists on Multi-

sided Platforms and Its Implications for Excluding Evidence that Ig-

nores It’ (2013), p 3, available at http://papers.ssrn.com/sol3/papers.

cfm?abstract_id=2249817.At p 11 he states that ‘[w]hile the result that

traditional models may not be applicable to multisided platforms is in-

convenient in practice, it is not controversial among professional econ-

omists’.

8 At roughly the same time both the Swedish Competition Au-

thority and the Swedish Academy decided to honour the developments

intheeldoftwo-sidedmarkets,albeitsomehowasymmetrically:the

latter by granting a Nobel Prize to Jean Tirole, one of the pioneers of this

literature and the former inviting me, among others, to participate at the

Pros and Cons conference.

9 European Commission note to the OECD’s Roundtable, n 5,

above, p 4. For a ‘not necessarily complete compendium of known and

well-documented problems with applying results based on a single-sid-

ed analysis to multi-sided platforms’, see Evans, n 7, above, p 9. For a

list of eight one-sided fallacies in which may occur when dealing with

multi-sided platforms,seeJWright,‘One-sidedLogicinTwo-sidedMar-

kets’(2004)3(1)ReviewofNetworkEconomics44–64.

do not recall having ever worked on a case in

which someone did not claim that the sector at

issue deserved special antitrust scrutiny; all sec-

tors claim to be special and, in a sense, they all

are. Admittedly, however, multi-sided scenarios

(which might arise in many markets, both tech-

nological and not) do seem to pose, or rather

exacerbate, practical problems that take com-

petition law out of its comfort zone.

Most of the theoretical models on which com-

petition law typically relies assume one-sided-

ness, in that they consider one single set of cus-

tomers and their reaction to changes in supply,

as well as the response of suppliers to changes

in that demand. In multi-sided platforms howev-

er, the assessment becomes multi-dimensional.

In these settings one needs to factor in the ex-

istence of multiple customer groups with inter-

dependent demand and analyse how each

side will react to a given move on the part of

the platform; (ii) how will the platform react to

moves on the different sides; and (iii) how each

side will react to each other.

The complexity of these exercises is further

enhanced by another important dimension to

consider: time. One of the crucial features of

these markets – particularly technology markets

– is the speed at which they progress; business

practices are not only complex, but also highly

dynamic; the ability of these platforms to grow,

and the speed at which they scale, is unprec-

edented in any other business. Accordingly,

these platforms are constantly increasing their

depth and reach, constantly redening their

boundaries as well as those of entire industries.

In case things were not difcult enough, com-

petition authorities are asked to react swiftly to

rapidly evolving situations. Moreover, and aside

from substantive questions, the time dimension

also raises enforcement issues: when should

competition authorities intervene? Is it prefera-

ble to prevent or to cure?

Interdependency, the pattern of cross-re-

sponses and speed are, in sum, what makes ev-

erything a bit more complicated, in life, in eco-

nomics and also in multi-sided platforms. And

by ‘everything’ I mean, literally, everything. As

acknowledged by the European Commission,

‘[t]his pattern of cross responses will generally

affect each step of standard antitrust analysis,

from product market denition, the competitive

assessment, entry, efciencies, etc’.

In light of the above, it is unquestionable that

having to apply competition law to multi-sided

platforms breaks the inertia and forces us not to

do things like we used to, thereby obliging us to-

think.

Against this background, I submit that the

thinking has been asymmetrical on the part of

economists, on the one side, and lawyers on the

other.

Much attention on the part of economists

and scholars has lately centred on how to adapt

and make practicable the tools we are most

accustomed to (such as the small but signicant

and non-transitory increase in price (SSNIP) test

or the Areeda-Turner/AKZO test), and progress

has certainly be made in this regard.

Whereas the renements and adaptations

to our toolkit proposed by economists are most

valuable and welcome, my contention is that

they may be of little use if jurists continue not to

address other questions raised by these markets

which go more profoundly both to the rootoft-

hedisciplineandtothewayinwhichtherulesareen-

forcedinpractice.

As I will submit in the discussion which follows,

the duality or ambiguity for competition purpos-

es of practices carried out in multi-sided plat-

forms has not been properly accounted for in

the law. As will be explained below, this duality

raises substantive and practical questions that

expose the inconsistencies and insecurities of

competition law and oblige us to question the

very values we purport to defend and the ob-

jectives we intend to pursue.

On the double duality of business

practices in multi-sided platforms

The platforms discussed in this contribution

typically receive special attention because of

their already explained duality; that is, they are

said to be peculiar because they involve two (or

more) sets of users that interact with each other

through the platform which, in turn, means that

business practices will be felt on multiple sides of

the market.

But in my view there is a second element of

duality of two-sided platforms that has not re-

ceived equal attention and that relates to the

competitive ambiguity of the practices carried

out in these settings.

Indeed, the circumstances in which practices

in multi-sided platforms may lead to foreclosure

are precisely the same ones in which they may

yield benets for consumers.

The dening characteristic of a multi-sided

platform is that it solves a transaction prob-

lem and creates value by bringing together

– physically or virtually – different groups/sides

that need each other but that cannot get to-

gether easily on their own. The platform makes

users better off by harnessing indirect network

effects by ensuring that there are enough play-

ers on both sides. This means that advantages

arise when a platform or intermediary manages

to attain a critical mass of users, and balances

and optimises the network effects (often by re-

sorting to asymmetric pricing, exclusivity and/or

tying, among other possible strategies).

On the other side, however, attaining the

necessary scale may very well imply depriving

competing platforms of the critical mass they

need, thus leading to their exclusion from the

market. Such exclusion may occur as a result

either of the natural tipping of the market to-

wards the most valuable platform

or of exclu-

sionary strategies which in other contexts would

be deemed irrational (in these settings each

time a competing platform is deprived of a giv-

en customer it loses not only the potential reve-

nues from that customer but also suffers a loss

in the overall value of the platform). One illus-

trative example is that of interoperability deni-

als. Although lack of interoperability diminishes

the value of a given network, it may appear as

a rational strategy given that it may particular-

ly damage small networks by denying them a

minimum viable scale.

This second dimension of the duality of

two-sided markets (ie their competitive ambi-

guity) is indeed not exclusive to this context but

rather derives from the existence of network

externalities which – although present by de-

nition in multi-sided settings – may well exist in

one-sided ones. It implies, in sum, that moves

to increase the scale of the side of the market

generating those externalities might result both

in greater scale and concentration (typically as-

sumed to be detrimental to consumer welfare)

as well as in increased platform value (which is

welfare enhancing for its members).

10 Whereas it is by now acknowledged that in most instanc-

es multi-sided platforms (or more generally speaking markets charac-

terised by network externalities) do not tend to monopolise given the

prevalence of product differentiation on attributes or quality and of the

possibilities for multi-homing, switching and, in many cases, interoper-

ability, the fact is that many multi-sided platforms operate in highly con-

centrated environments. This is not necessarily good (except for com-

petition lawyers) or bad, it simply is part of the background in which

competition rules are to be applied.

The problem that competition law faces

in these settings is similar to the one faced by

multi-sided platforms when conducting busi-

ness – in essence, how to strike a balance be-

tween the two sides, in this case the offensive/

anti-competitive and the defensive/pro-com-

petitive.

In what follows, I will contend that whereas

it is clear in economics that business practices

in multi-sided platforms setting have both an

offensive and defensive potential, and where-

as this seems to have been acknowledged on

a theoretical basis by competition authorities

and courts, the practical application of the

competition rules results in an imbalance that

overplays the offensive, or anti-competitive, po-

tential of such practices and makes defences

effectivelyunavailable.

Multi-sided market features

as a sword

Competition authorities have been aware

since the early 1990s of the offensive potential

of network effects, including in multi-sided plat-

form settings.

Interestingly, both the doctrine and the appli-

cation of the law in the face of network effects

have tended to focus on their anti-competitive

potential (which is somehow paradoxical for a

positive, theoretically desirable, externality). In-

deed, most of the attention paid to network ef-

fects by antitrust enforcers and scholars – later

consolidated in precedents and guidelines

–

eminently relates to their characteristic as a bar-

rier to entry. As a result, network effects have

proved to be, in practice, a most effective basis

for legal arguments challenging allegedly an-

ti-competitive conduct.

This has been evident with regard to a wide

array of practices, both price and non-price re-

lated, as well as in relation to merger control.

11 HJ Hovenkamp, The Antitrust Enterprise, Principle and

Execution (Harvard University Press, 2005), at 281 (‘These same

features that make networks attractive also create the opportunity for

anticompetitive practices’); M Schanzenbach, ‘Network Effects and

AntitrustLaw:Predation,AfrmativeDefenses,andtheCase ofU.S.

v. Microsoft’ (2002)4 Stanford Technology Law Rev 3 (asserting that

‘network competition provides unique opportunities for anti-competi-

tive strategies’, but emphasises that ‘network competition also provides

someuniquepro-competitivejusticationsforpracticesthathavetra-

ditionally received antitrust scrutiny, such as tying, exclusive dealing,

and low-pricing strategies’, concluding that ‘network effects can be

a double-edged sword’); GL Priest, ‘Rethinking Antitrust Law in

an Age of Network Industries’ (2007) 4 Yale Law & Economics Re-

search Paper No 352, at 4 (‘[M]any practices in the context of networks

that may seem puzzling become understood when the need to correct

for positive network externalities is taken into account’); DJ Gifford,

‘The European Union, the United States, and Microsoft: A Comparative

Review of Antitrust Doctrine’, CLEA 2009 Annual Meeting Paper,

available at SSRN: http://ssrn.com/abstract=1434089or http://dx.doi.

org/10.2139/ssrn.1434089,pp 19–20: ‘Network effects carry a double

edge’; SF Ross, ‘Network Economic Effects and the Limits of GTE

Sylvania’sEfciencyAnalysis’(2001) 68AntitrustLaw Journal951:

‘Firms that produce goods with network effects can engage in con-

ductthatpromotesefciency,inthesensethattheresultingproductis

cheaper, intrinsically superior in quality, or that the product’s greater

use by others increases each consumer’s utility. The same conduct can

simultaneouslyhave signicant exclusionaryeffectsbecause thecon-

ductmakesitevenmoredifcultfornewentrantstoovercomethefact

that so many consumers now use the dominant rm’sproduct’; WH

Page, ‘Microsoft and the Limits of Antitrust’ (2009) Journal of Com-

petition Law & Economics (forthcoming), University of Florida

Levin College of Law Research Paper No 2009–40, available at SSRN:

http://ssrn.com/abstract=1501079,at 9: ‘The very existence of network

effects makes certain practices that resemble antitrust violations social-

lybenecial…’;WJKolasky,‘NetworkEffects:AContrarianView’

(1999) 7 George Mason Law Review 578: ‘Network effects may

well exhibit unique characteristics, but these characteristics do not all

point in one direction. Network effects will as often provide a valid

precompetitivebusinessjusticationforconductastheywillareason

for holding otherwise lawful conduct unlawfully’.

12 European Commission, Guidance on the Commission’s en-

forcement priorities in applying Article 102 TFEU to abusive exclu-

sionary conduct by dominant undertakings, paras 17–20 (‘The Com-

mission will normally intervene under Article 82 where, on the basis of

cogent and convincing evidence, the allegedly abusive conduct is likely

to lead to anti-competitive foreclosure. The Commission considers the

following factors to be generally relevant to such an assessment: (…)

the existence of economies of scale and/or scope and network effects.

Economies of scale mean that competitors are less likely to enter or

stayinthemarketifthedominantundertakingforeclosesasignicant

part of the relevantmarket. Similarly, the conduct may allow the dom-

inant undertaking to ‘tip’ a market characterised by network effects in

its favor or to further entrench its position on such a market. Likewise,

ifentrybarriersintheupstreamand/ordownstreammarketaresigni-

cant, this means that it may be costly for competitors to overcome pos-

sible foreclosure through vertical integration’), 24; Guidelines on the

assessment of horizontal mergers under the Council Regulation on the

control of concentrations between undertakings (2004) OJ C 31/5, para

72; Guidelines on the assessment of non-horizontal mergers under the

Council Regulation on the control of concentrations between under tak-

ings (2008) OJC265/6,paras62,101.

100

The approach of competition authorities to

pricing practices is perhaps most striking, at least

at rst sight.

As observed by Rochet and Tirole,

‘theoretical models predict that skewed pricing

is more likely to be the norm than the exception

for MSPs [multi-sided platforms]’ but ‘surprising-

ly, skewed pricing has sometimes been used by

competition authorities in completely opposed

ways’.

Indeed, proving true a well-known

quote that is often used to ridicule competi-

tion law enforcement,

competition authorities

and courts in the EU have taken action against

prices that were too high (notably concerning

the multilateral interchange fee (MIFs) applied

by card payment systems), against prices that

were allegedly too low (see eg the allegations

on predatory pricing on the part of Google re-

garding maps and mobile operating systems)

as well as against prices that were considered

to be too stiff (eg the recent investigations

into MFNs/best price guarantees applicable

to online resellers in hotel reservation systems,

e-books or Amazon’s marketplace). In relation

to non-pricing practices attention has tended

to focus on exclusivity arrangements,

as well

as on tying/bundling

and on alleged ac-

cess and discrimination issues against so-called

‘gatekeepers’ – a fashionable term nowadays

– or ‘competitive bottlenecks’ (which arise

when a given platform is an unavoidable trad-

ing partner for agents on one side of the market

to reach the single-homing agents on the other

side).

The same vigilant approach is visible in the pre-

ventive eld of merger control. The predominant

tendency on the part of antitrust agencies has typ-

ically

been to assume that network effects may

increase barriers to entry as well as incentives to

act anti-competitively following a change of mar-

ket structure pursuant to a merger.

Multi-sided market features

as a shield

Whereas discussions on network effects have

typically focused on their offensive potential,

discussions on multi-sided platforms (which, as

explained, deal in reality with the same root

13 Perhaps with the exception of the Microsoft / Skype case

(in which the Commission’s unconditional authorisation in Phase I was

validated by the judgment of the General Court of 11 December 2013

in Cisco Systems, Inc and Messagenet SpA v European Commission

(CaseT-79) EU:T:2013:635, but for reasons not attributable to (and

still not well understood by) the lawyers representing the applicants,

including myself. For my comments on this case, see http://chilling-

competition.com/2014/05/12/a-comment-on-case-t-7912-cisco-sys-

tems-and-messagenet-v-european-commission-microsoftskype/.

14 In reality, the apparent contradictions observed in this regard

may not be such, given that, in practice, pricing practices can cer-

tainlybeusedtoexploit,excludeandreinforcearm’sposition.That

there may not be a real inconsistency in the approach of competition

authorities does not mean that the approach in all individual cases has

been the right one.

15 J-C Rochet, J Tirole, ‘CPI Introduction to the Symposium’

(2007) 3(1) 148, available at: http://papers.ssrn.com/sol3/papers.

cfm?abstract_id=987339

16 ‘Ronald [Coase] said he had gotten tired of antitrust because

when the prices went up the judges said it was monopoly, when the pric-

es went down they said it was predatory pricing, and when they stayed

the same they said it was tacit collusion’, W Landes, ‘The Fire of Truth:

A Remembrance of Law and Econ at Chicago’ (1981) Journal of Law

and Economics 193.

17 The prevalent thinking among competition authorities with

regard to exclusivity on one side of multi-sided platforms is that it

wouldarticiallyincreaseswitchingcosts,therebyhinderingcompet-

ing platforms’ ability to obtain the necessary critical mass with which

to gain a foothold on the market. In these circumstances it is generally

assumed that network effects exacerbate the collective choice problem,

since consumers will be aware of the disincentives created by exclu-

sivityforotherconsumerstoshiftnetwork.Consequently,arivalrm,

even one which could offer a superior product or service, would not

have any opportunities unless users have the ability to act coordinately,

whichmayberare.

18 Tying and bundling are looked at more suspiciously in in-

dustries with network effects, economies of scale and high barriers

to entry (see, eg, the Microsoft cases, concerning the tying of Windows

Media Player and Internet Explorer; see also the European Commis-

sion’s Staff Discussion Paper on the application of Article 82 EC to

Exclusionary Abuses, para 180 (2005), available at http://ec.europa.

eu/competition/antitrust/art82/discpaper2005.pdf).

19 Theories of harm alleging the existence of competitive bot-

tlenecks have been brought in relation to search engines, computerised

reservation systems, mobile communication networks, Internet Service

Providers, credit card networks and supermarkets, amongothers.

20 Perhaps with the exception cited at n.13(apologies for the

one-sidedness on this one).

21 Accordingly, network effects are generally seen as factors

with the potential to complicate the anti-competitive effects of a merger.

101

phenomenon) rather tend to highlight their the-

oretical defensive potential.

The key idea I want to convey here is that the

economic literature shows that demand-side

efciencies achieved by multi-sided platforms

may turn typically condemned practices into

welfare enhancing ones. This is the case, for ex-

ample, with horizontal cooperation agreements

within the network aimed at capturing external-

ities or expanding the network (think of the MIFs

in the payments industry, of standard setting

agreements, of collecting management societ-

ies or of airline alliances), of what could prima

facie be regarded as predatory/excessive pric-

ing,

as well as with other unilateral practices

such as exclusivity arrangements

and tying/

bundling,

all of which might be used to harness

network effects.

In spite of these conceivable defences, it is

still most rare to see demand-side efciencies

being effectively acknowledged as a valid de-

fence in real cases.

My contention is that this lack of consider-

ation for possible redeeming virtues arising from

demand-side efciencies stems both from the

inability of economics to quantify the external-

ity, as well as from the inability of the law to ac-

count for this gap and adapt to it.

I attribute this problem to three main causes:

First, competition authorities are out of their

comfort zone when asked to assess defences

based on the internalisation of network exter-

nalities (ie the increase in the value of the plat-

form to make it viable or more effective) in the

absence of quantication. The lack of a reliable

method to quantify the advantages derived

from the externality means that, in practice, at-

tempts to bring up defences related to the ef-

ciencies arising from a larger or more balanced

platform are typically doomed.

The requirement of objective quantication

contrasts with the much lighter burden imposed

Perhapstherstexampleoftheapplicationofnetworktheorytotheas-

sessment of merges lies on the MCI/WorldCom case, which was cleared

both by the European Commission and by the DOJ upon the condition

that WorldCom would divest MCI’s internet business. Since then, the

European Commission has approached mergers in multi-sidedplat-

formswithparticularcare;think,forinstanceofMicrosoft/Yahoo,Travel-

port/Worldspanand Google/DoubleClick, all of which were centered on

theories of harm which were explicitly based on cross-market effects.

Once again, the outlier here is Microsoft/Skype, a case in which a the-

oretically straightforward cross-market effects theory of harm was put

forward but was labelled as ‘conglomerate’ and rapidly dismissed too

‘complex’and‘uncertain’.

22 Contrary to the Areeda/Turner and AKZO assumption that

underlying a price below marginal cost there is generally an anti-com-

petitive purpose, exclusionary intent is not necessarily present in rela-

tion to penetration pricing aimed at providing incentive for one side to

join, thus making the platform viable or expanding its reach and, conse-

quently, its value. In fact, low pricing is the most obvious way in which

a network owner can internalise the consumption externality by setting

the price charged for joining at a price below the costs that the addition

impliesforthenetworkrmwithaviewtobuildingcriticalmass.Ina

way,therm usingsuch strategy isinvesting inthe networkthrough

the purchase of its most valuable input: the customer. Low prices on

one side of the platform may very well be accompanied by what could

be regarded as excessive prices on the other side thereof. Admittedly,

the increase in the value of the network could also be coupled by the

exclusion of competitors, there by making it necessary to balance pro

andanti-competitive effects.

23 As observed by Shapiro, whereas it is widely assumed that

in network settings pro-competitive features will be outweighed by

greater competitive harm, exclusivity can also serve to differentiate

products and networks, to encourage investment in these networks, and

to overcome free riding. Exclusivity obligations may also act to the det-

rimentofanincumbentrmfacingaparticularlystrongentrantgiven

that it may ‘[induce] customers who would otherwise be a member of

both networks to join only the new network’. In addition, multi-sided

platforms may possibly enhance some of the pro-competitive effects of

exclusivity. It is commonly admitted that exclusivity might facilitate

long-termplanning, thus reducing the risk of incurring xed costsin

production. This contribution to the elimination of uncertainties is par-

ticularly useful in multi-sided markets, characterised by the necessity

of incurring large sunk costs in unpredictable contexts. C Shapiro,

‘Exclusivity in Network Industries’ (1999) 7 George Mason Law

Review675.

24 In certain multi-sided settings tying may contribute to pre-

serving or expanding the positive network externality by adding new

functionality to network platforms, by helping entrants overcome barri-

ers to entry. See, eg Priest, n 11, above, p 8 andPage,n11,above,p9.In a

multi-sided market tying / bundling canal so be used as a monetisation

strategy.

25 This has lead even Nobel-prize winners to underline the ‘in-

sufcientattentionpaidtoefciencyconsiderationsrelatedwithusage

externalities’, Rochet and Tirole, n. 15, above, p. 148.

102

on the authorities, which are not obliged to ‘ob-

jectively quantify’ restrictions; they can meet

their burden of proof on the basis of qualitative

factors, but private undertakings having to de-

fend themselves cannot. This unevenness in the

applicable standard of proof risks – as will be

discussed in detail later – an effective shift of the

burden of proof from the Commission to the

undertaking. Consequently, in doubtful cases

authorities and courts may risk following the

reex of condemning complex practices de-

spite, or precisely because of, the impossibility

to adequately assess their effects.

Secondly, the legal principles typically used

to assess redeeming virtues have not always

been interpreted in a manner well-suited to ac-

count for cross-market efciency assessments,

which will obviously be necessary when more

than one side of a platform is affected by a giv-

en practice.

This is in contrast with the lessons

derived from economics, which tell us that the

ideal solution here would be to strike a balance

between all interests at play (balancing favour-

able and detrimental effects of the agreement

across markets and across customer groups).

According to the Guidelines on the applica-

tion of Art 101(3) efciency gains are in principle

assessed ‘within the connes of each relevant

market to which the agreement relates’.

Whereas the guidelines envisage that where

two markets are related one can take into ac-

count the efciencies in the other, they never-

theless require that the group of consumers af-

fected by the restriction and benetting from it

be substantially the same.

The Guidelines on 101(3) not only reected,

or rather set, the creative – and in these settings

problematic – approach adopted by the Com-

mission in this regard, but they also somehow

captured the General Court,

which validated

them on this point its Mastercard judgment.

appeal, however, despite upholding the Gener-

al Court’s Mastercard ruling, the Court of Jus-

tice of the European Union (ECJ) made it clear

that when assessing compliance with 101(3) ‘it

is necessary to take into account the system of

which that measure forms part, including, where

27 The two dualities that we have referred to can furthermore

be apparent at the same time whenever a reduction of competition on

one side is coupled by welfare enhancing effects on another side of the

platform. This, moreover, will frequently be the case for affecting the

cross-group externality; those practices might both enhance users’ wel-

fare and exclude third parties, often at the same time.

28 Commission Guidelines on the application of Article 81(3)

oftheTreaty,OfcialJournalNoC101of27.4.2004,at43.

29 For more developed comments on this issue, see

http://chillingcompetition.com/2015/01/20/on-the-misapplica-

tion-of-article-1013-of-judicial-capture-and-cross-market-assess-

ments/.

30 In its judgment of 24 May 2012 in MasterCard, Inc and

Others v European Commission (Case T-111/08) EU:T:2012:260,

the General Court acknowledged at para 228 that ‘the appreciable ob-

jectiveadvantagestowhichtherstconditionofArticle[101(3)]EC

relates may arise not only for the relevant market but also for every

othermarketonwhichtheagreementinquestionmighthavebenecial

effects’, but nevertheless ruled that ‘as merchants constitute one of the

26 An illustration can be found in the Mastercard Interchange

Fees case. In a nutshell, a 2007 Commission decision concluded that

MasterCard’s intra-EEA cross-border multilateral interchange fees for

credit and debit cards were contrary to Art 101(1) TFEU in as much as

theyrestrictedcompetitionbetweenacquiringbanksbyarticiallyin-

creasing the basis on which these banks set their charges to merchants.

MasterCard argued that the anti-competitive effect outlined above

couldbeoutweighedbyefcienciesstemmingfromMIFintheformof

lower cardholder fees on the opposite side of the market. In other words,

MasterCard’s reasoning was that in light of the ‘two-sided’ nature of the

payment card industry, MIFs were ‘set to balance issuing and acquiring

demands, so as to ‘get both sides on board’, thereby internalising net-

work externalities and maximising output and consumer welfare. The

CommissionobservedthatMIFswerealsoabletogeneratesignicant

efcienciesinlightofthe‘two-sided’natureofthemarket.Nonetheless,

it rejected MasterCard’s allegations considering that even though MIFs

couldbeapotentialsourceofefciencies,‘MasterCardfailedtosubmit

the required empirical evidence to demonstrate any positive effects on

innovation and efciency which would allow passing on a fair share

oftheMIFbenetstoconsumers’.Thedecisionwasappealedbefore

the General Court and subsequently before the ECJ, both of which

upheldtheCommission’sdecision,thusconrmingthedifculties

incumbentupon any partywishing to claim the benets arising from

network effects in multi-sided markets. The General Court ruled that

Mastercard had ‘failed to submit empirical evidence on the positive ef-

fect of MIFs on system output’ and that since these had not been ‘objec-

tivelyquantied’,theycouldnotbetakenintoaccount.TheECJdidnot

disputethisnding.

103

appropriate, all the objective advantages ow-

ing from that measure not only on the market in

respect of whichthe restriction has been estab-

lished, but also on the market which includes the

other group of consumers associated with that

system, in particular where, as in this instance, it

is undisputed that there is interaction between

the two sides of the system in question’.

The

same idea, in relation to Art 101(1) underscores

the judgment in Groupement des Cartes ban-

caires rendered by the ECJ on the very same

day.

In these judgments the ECJ showed that it

had become aware of this aw in the traditional

analytical framework, and appears to have set

the law on a new course, at least with regard

to multi-sidedplatforms.

Thirdly, we have so far proved unable to

trade off the benets and the perils of having

one large scaled platform as well as the circum-

stances in which one platform is preferable to

having several.

The Mastercard case once again provides a

useful example of a circumstance in which, in

the face of doubt or ambiguity, the offensive

theory is favoured.

A similar illustration can be

found in the CFI’s judgment inMicrosoft.

Addressing a practical imbalance – back to

basics

We have seen that while business practices

in multi-sided platforms are often pro-competi-

tive, or at least ambiguous from a competitive

standpoint, the practical application of the law

reects an imbalance in which offensive argu-

ments are favoured and conceivable defences

are most often effectively ignored.

This results partly from the inability to quanti-

fy the value of the externality, partly from the

wording and the ‘funnel structure’ of compe-

tition law provisions, partly from the difculties

inherent in cross-group assessments and partly

from our natural inclination to favour narrow

prisms and one-sidedness in the face of com-

plexity.

In my view, this imbalance results in a prob-

lematic imbalance or, putdifferently, in enforce-

menthemiplegia.

Against this background, what this contri-

two groups of users affected by payment cards, the very existence of the

second condition of Article [101(3) TFEU] necessarily means that the

existence of appreciable objective advantages attributable to the MIF

must also be established in regard to them’. In other words, since no

quantiable advantages benetted merchants, there was no need   to

verifywhetheranysuchadvantagesbenettedcardholders.

31 MasterCard Inc and Others v European Commission (Case

C-382/12 P) EU:C:2014:2201, para 237. In spite of this sensible state-

ment, para 248 of the same judgment implicitly validates the contrarian

approachadoptedinrstinstancebytheGeneralCourt.

32 Groupement des cartes bancaires (CB) v European Commis-

sion (Case C-67/13 P) EU:C:2014:2204, paras 76–79.

33 Paragraph 222 of the General Court’s judgment states that

‘anincreaseintheplatform’soutputcanbethesourceofefciencies,so

inadditiontogivingrisetoefciencies,itcouldalsoenableMastercard

to extract rents’. In the case at issue, the Court understood that the fact

thatMastercardcouldextractrentswasautomaticallysufcienttonulli-

fytheadvantagesowingfromtheoutput/networkexpansionsoughtby

MIFs(whichtheoreticallyshouldbenetallmembers)withoutconsid-

ering it necessary to undertake any balancing exercise.

34 In this case the Courtrejected Microsoft’s arguments on the

existence of an objective justication for its conduct. Microsoft had

contended that integrating Windows Media Player in Windows pro-

videdsoftwaredeveloperswithastableandwell-denedplatformfor

software development that could facilitate their tasks. In response to

this claim, the CFI stated that ‘[a]lthough, generally, standardization

may effectively present certain advantages, it cannot be allowed to be

imposed unilaterally by an undertaking in a dominant position by means

of tying (…) [I]t cannot be ruled out that third parties will not want the

de facto standardization advocated by Microsoft but will prefer it

if different platforms continue to compete, on the ground that that will

stimulate innovation between the various platforms’. (See Microsoft

Corp v European Commission (Case T-201/04) EU:T:2007:289,

paras 1152–1153.) As pointed out by Larouche, the Court’s argu-

ment that some third parties would rather prefer competition between

platforms is little more than a mere unsupported conjecture. See in this

regard Larouche arguing that the CFI’s reasoning in this regard calls

for ‘further research on the link between competition policy, innova-

tion policy, and standardization’, P Larouche, ‘The European Microsoft

Case at the Crossroads of Competition Policy and Innovation’

(2008) 22 TILEC Discussion Paper No 2008–021,available at http://

ssrn.com/abstract=1140165.

104

bution posits is that if there is a problem at the

legal level, it is there that we need to act; it is

therefore not only economic tools that need to

be rened in the presence of multi-sided plat-

forms, but also the law, or rather the application

thereof.

These renements I am referring to do not re-

quire a policy revolution, but rather increased

analytical vigilance, mainly concerning (i) the

assessment of welfare enhancing features;

and (i) the upholding of certain basic limiting

principles that should not be forgotten.

Assessment of welfare enhancing

features/efciencies

To start with, in my view many of the identied

problems in two-sided markets would be avoid-

ed if we were to adopt a more reasonable in-

terpretation of Art 101(3) TFEU and of the ‘ef-

ciency defences’ as part of the assessment of

practices and mergers under Art 102 TFEU and

the MergerRegulation.

In what follows I will mainly refer in this regard

to Art 101, given that it incorporates an explicit

and specic sub-provision laying down the an-

alytical principles governing the assessment of

welfare enhancing features that also inspire – in

practice – the operation of Art 102 TFEU and of

the Merger Regulation,

and that make more

evident some of the issues that I intend to ex-

plain.

From an orthodox perspective, welfare en-

hancing features pertain to an Art 101(3) anal-

ysis, but nevertheless assessments carried out at

this stage very rarely prosper in individual cases.

I would argue that the overly restrictive interpre-

tation of Art 101(3) endorsed by competition

authorities and exposed in soft law instruments

is at the root of many contemporary controver-

sies and contortions in EU competition law, also

regarding multi-sided platforms.

In my view, reasonable competition law en-

forcement should be characterised by fewer

object cases, more effects cases, and many

more Art 101(3) cases, and it is therefore here

that we should rst take action. Among others,

and in line with what has been explained

above, I submit that a proper interpretation of

this sub-provision (and of analogous analytical

steps in other areas of competition law) should

not require ‘objective quantication’ of de-

mand-side efciencies, and that it should allow

for cross-market assessments, also regarding dis-

tinct groups of customers, contrary to what had

been done in the recent past.

The EU Courts more recently seem to have

acted at this level, but with a twist that is, in my

view, a second best, but welcome, solution.

In the light of the truncated analysis carried

out under competition law and the circum-

stances already explained, what we see now-

adays is that in any given case the accusing

party is able to more easily discharge its burden

of proof in a rst step of the analysis, whereas

the defendant will very rarely be able to com-

ply with the burden of showing the existence of

sufcient, objective, in-market welfare enhanc-

ing features.

Perhaps the acknowledgement of this situa-

tion may have led the EU Courts increasingly to

tolerate and encourage that redeeming con-

35 As José Ortega y Gasset said in one of my favourite quotes,

included in Toward a Philosophy of History (WW Norton & Company

Inc, 2002): ‘Aligning oneself fully with the left, as with the right, is only

one of the numberless ways open to man of being an imbecile: both are

forms of moral hemiplegia.’

36 See eg para 30, Communication from the Commission –

Guidance on the Commission’s enforcement priorities in applying Ar-

ticle 82 of the EC Treaty to abusive exclusionary conduct by dominant

undertakings (2009) OJ C 45; Art 2(1) of Council Regulation (EC) No

139/2004 of 20 January 2004 on the control of concentrations between

undertakings (the EC Merger Regulation) (2004) OJ L 24; Section VII

of Guidelines on the assessment of horizontal mergers under the Coun-

cil Regulation on the control of concentrations between undertakings

(2004) OJC31.

105

siderations be looked at within a rst stage in

which the burden of proof remains on the ac-

cusing party.

This, once again, is particularly evident in the

Mastercard and Cartes bancaires judgments of

11 September 2015, in which the ECJ arguably

conveyed the message that welfare enhanc-

ing features derived from two-sidedness can be

better assessed as part of Art 101(1), instead of

within Art101(3).

In my view, the analytical framework called

for in these two recent judgments implies that,

in the future, if a competition authority or com-

plainant were to suspect a prima facie Art 101

or 102 infringement in a case, it would then be

up to the defendant to bring a prima facie

– even if abstract – claim that the practice is

necessary to create, maintain, balance or ex-

pand the platform at issue. Should the defen-

dant then be able to make such claim, it would

be up to the party claiming the existence of

an infringement to motivate why such conten-

tions are not valid and/or to conduct itself the

balancing of pro- and anti-competitive effects,

dispensing the defendant of that burden.

This is also in many ways the message that, in

the US the DC Circuit Court sent in its Opinion

in Microsoft II with regard to similar issues: that

one should not hurry to condemn practices for

which a prima facie justication could be put

forward.

Given that, in practice, the operation of the

burden of proof often determines the outcome

of cases, acting at this level – by alleviating the

burden of proof incumbent upon the defendant

– is probably the most simple and effective

way of addressing these difculties

Do not forget the value of limiting

principles – back to basics

In addition to the above, and in order to

achieve consistency in the application of

competition rules in these markets, I believe

that we also need to revisit some basic tenets of

competitionlaw.

The ‘double duality’ of two-sided markets

raises substantive and practical questions that

expose the inconsistencies and insecurities of

competition law and oblige us to interrogate

ourselves about the very valued we purport to

defend and of the goals we intend to pursue.

In my view, the main challenge posed by

these markets/platforms lies not in the novelty

of the issues they raise, but on the intensity with

which those issues – notably related to the ‘old’

phenomenon of ‘scale’ – arise in these settings.

This means that the questions we are facing

now are to a great extent ones to which anti-

trust already replied in earlier days; the differ-

ence is mainly one of degree.

Indeed, in any given case involving multi-sid-

ed platforms competition enforcers will invari-

ably face certain empirical questions

but,

ultimately there will remain other more ‘philo-

sophical’ ones that go to the heart of the disci-

pline and that risk being answered on casuistic,

inconsistent and almost reexive or ideological

– and, as such, unsupported – grounds.

Against this background, I submit that it is nec-

essary to recall clear principles, lters or bright

37 See notably Groupement des cartes bancaires (CB) v Euro-

pean Commission (Case C-67/13 P) EU:C:2014:2204, paras 72–79 and

MasterCard Inc and Others v European Commission (Case C-382/12 P)

EU:C:2014:2201, paras 170–180.

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

at 58–59. (‘If a plaintiff successfully establishes a prima facie case un-

der s.2 by demonstrating anticompetitive effect, then the monopolist

mayproffera“pro-competitivejustication”foritsconduct.Ifthemo-

nopolistassertsa precompetitivejustication–a nonpretextualclaim

that its conduct is indeed a form of competition on the merits . . . then

the burden shifts back to the plaintiff to rebut the claim . . . [I]f the

monopolist’s pro-competitive justication stands unrebutted, then the

plaintiff must demonstrate that the anticompetitive harm of the conduct

outweighstheprecompetitivebenet.’)

106

lines capable of adding some predictability to

the law or, in other words, to go back to basics.

I would propose to go back to basics and to

reinstate some key principles that competition

law has learnt over the years, but that are worth

recalling now that some of the earlier questions

antitrust faces are resurfacing with increased in-

tensity.

In my view, we should hold the following to be

self-evident, also, and particularly, in multi-sided

settings:

(1)

Absence of rivalry does not equal in-

fringement (protecting competition vs

protecting competitors).

Economics teaches us – and compe-

tition authorities have accepted in some

settings

– that, at the extreme, a mo-

nopolistic structure could in some scenar-

ios (natural monopoly, no diseconomies

of scale on the cost side, no congestion

effects on demand, homogeneous con-

sumers on both sides) be the most efcient

market structure. In this regard, it is per-

fectly conceivable, at least in theory, that

the benets of a larger platform outweigh

other possible downsides of market power

such as higher prices.

(2)

Remedies and objective justications

follow the establishment of an infringe-

ment by the

authority,not the other

way around.

As competition law has become increasingly

more regulatory

this – I would say obvious –

principle seems to have been partly forgotten.

Competition law, which is repeatedly held to

be quasi-criminal in nature, is not supposed to

kick-in in the face of a sub-optimal functioning

of markets, but only when an infringement has

been established by the authority or plaintiff.

(3)

Companies must be free to choose

their business model.

It is companies and not competition enforcers

which will strive or fail in the adoption of their

business models, and it is therefore companies

and not competition enforcers who are to de-

cide on what business models to use. Some will

prove successful and others will not; some com-

panies will thrive and some will disappear, but

with experimentation with businessmodels,suc-

cessandfailureareandhavealwaysbeenpartoft-

hegame.

In other words, we should not forget that com-

petition law is, or should be, business-model

agnostic, and that regulators are – like anyone

else – far from omniscient.

(4)

Competition law is about protecting

the process of competition from un-

due restraints.

It is not about shaping the process (see above),

and it is not about creating or preserving com-

petition in the face of the natural evolution of

markets.

41 Kolasky, n 11, above, at 585. (‘[S]ince positive network ef-

fectsgiverisetoefciencieswhichrmsmaycaptureandpassonto

consumers, it is important that we not interfere with the natural oper-

ation of the market, making the old mistake of protecting competitors,

rather than competition.’)

42 P Ibañez Colomo, ‘On the application of competition law as

regulation: elements for a theory’ (2010) 29 Yearbook of European Law

261–306.

39 Among others, how strong is the interdependency/network

affects across the different sides? What is the relative strength of differ-

entiation versus network effects? How easy is it for users to switch plat-

form? Is multi-homing possible and/or prevalent? What is the optimal

scale and what is the minimum critical mass for others to compete?

40 See the European Commission note to the OECD’s Round-

table on Two-Sided Markets, n 5, above, p 7: ‘… as in all markets

with network externalities, there is often the possibility that one plat-

form will corner (both sides of) the market if the inter-group external-

ities are powerful. It can be very hard for an entrant in such markets to

get started. However, this outcome is not necessarily bad from asocietal

point of view when externalities are strong’.

107

(5)

Competition law should be applied con-

sistently and there is no reason to favour

one parameter of competition over oth-

ers.

There will be situations in which our natural re-

exes will lead us to think – in the abstract – that

an apparent reduction in static competition

might possibly reduce

innovation, choice or

quality even if (or especially when) the analyt-

ical framework centred on price does not en-

able us to nd an infringement.

However, since we are not yet capable of ade-

quately balancing the benets of differentiation

and possible innovation against the increases in

value of a multi-sided platform these decisions

may be adopted on the basis of ideological

considerations,

which are, in my view, ill-suited

to be the basis of a sanctioning regime.

In sum, the fact that it is harder to measure pa-

rameters other than price and output does not

mean that these should be privileged over oth-

ers (rather the contrary) or that it is justied to

depart from well-established principles when

intervention is a response to the alleged impact

of a practice on a parameter other, and more

abstract, than price.

(6)

In dubio pro reo (or, when in doubt,

don’t chill competition).

In many ways, the above can be summed up

in one simple idea, that competition law should

explicitly acknowledge its limitations and not

condemn what it does not fully understand.

Indeed, being aware of the fact that many

practices carried out by or within multi-sided

platforms may be efciency enhancing and

that the prohibition of such arrangements may

greatly damage consumer welfare is useful and

necessary, and pleads in favour of the inade-

quacy of outdated and simplistic per se rules

to these settings. In other words, traditional as-

sumptions and inclinations should be relaxed,

and particular caution is needed to approach

multi-sided platform issues with more humility.

Established economics tells us that any wel-

fare enhancing policy should encourage, or

at least tolerate, internalisation strategies. On

the contrary, failing to identify and protect net-

work efciencies in multi-sided platforms will be

to the detriment of societal welfare. The main

competition law issue must therefore be to sort

the practices that effectively contribute to bal-

ancing the externalities and contribute to the

optimal size of the platform from those that

do not.

To be sure, this is not to say that competition

law does not have a role to play in multi-sided

platforms markets. The fact that authorities are

to try harder does not mean that they may not

be able to bring solid theories of harm; it only

means that they cannot do this in a simplistic

manner or in the abstract. In particular, I believe

that the proposed lter does not in any way hin-

der authorities’ ability to purse cases concern-

ing what should perhaps be their main target in

these markets: ‘cheap exclusion’.

Finally, I submit that it would be most useful for

these principles to be reected in some informal

guidance, deserving specic treatment within

the main soft law instruments issued by competi-

tion authorities. Competition law is often seen as

too special an animal by companies and judg-

43 See FA Hayek, ‘Competition as a Discovery Procedure’,

in New Studies in Philosophy, Politics, Economics and the History of

Ideas (University of Chicago Press, 1978), at 179.

44 Interestingly, the use of an ideological approach to con-

demning conduct in network markets has been explicitly advocated

by some commentators. See Ross, n 11, above, at 947 (proposing that

‘wheremonopolisticconductsignicantlyinhibitstheabilityofrivals

to engage in fair competition by means that to some extent frustrate con-

sumer preferences, and network effects suggest that courts cannot prac-

ticallydetermineifclaimedefciencybenetsoutweightheseharms,

courts should employ a “Jacksonian” value of equal economic oppor-

tunity to proscribe the conduct and give others a meaningful chance to

competewiththedominantrm’).

45 On a personal level I could even agree with the contention

thatinsomesettingsthemosteconomicallyefcientoutcomemightnot

be the most convenient for societal welfare. That, however, is a problem

that can, if needed, be addressed via regulation, but not through the use

of competition law.

46 See in this regard SA Creighton, DB Hoffman, TG Kratten-

maker and EA Nagata, Cheap Exclusion (2005) 72 Antitrust Law Jour-

nal 975.

108

es, and all of them would benet from having

an established analytical framework in writing,

which would moreover contribute to minimising

the risk of divergences in the resolution of cases.

Conclusions – a reminder of the

fallibility of competition law

The preceding pages submit, rst, that a cru-

cial peculiarity of multi-sided settings derives

from the fact that when there are various sides

to one platform, there are often two sides for ev-

ery story or theory of harm; that everything has

pros and cons.

Economic lessons have served us well in this

regard by providing us with a balanced view

of the ambiguity of business practices in these

settings and conrming that some problems are

so complex that one needs to be very well in-

formed just to be undecided about them.

In spite of economic consensus on the duality

or ambiguity of practices carried out in multi-sid-

ed settings (as a subset of the situations in which

network effects are key), I have attempted to

show that there is an imbalance in the practi-

cal application of the law that favours offensive

theories to the detriment of equally plausible

defensive ones.

Against this background, what my contribu-

tion posits is that we need to be aware of that

imbalance with a view to correcting it at the

level of the application of the law, and that

what is needed is not a policy revolution, but

analytical prudence.

In many ways, however, the ideas that I have

tried to develop in this contribution do not relate

to multi-sided platforms alone. In reality, they

are pertinent to the application of competition

law in general.

Indeed, and as already noted, the main spec-

icity of multi-sided markets is that they pose the

very same issues that have troubled antitrust law

since its inception; the difference is that those

same issues arise now with renewed strength,

particularly in technology enabled markets in

which the phenomenon of scale has reached

new heights.

By presenting us with extreme cases and

questions, multi-sided platforms not only reveal

the inadequateness of traditional tools and

proxies to these specic settings, they also ex-

pose the insecurities and inconsistencies of this

discipline by reminding us that, in those tools

and proxies are, in reality, and irrespective of

multi-sidedness, never accurate.

In this sense, the challenges raised by multi-sid-

ed platforms are a useful reminder of the fallibil-

ity of competition law, and of the need for hu-

mility, prudence and clear limiting principles.

109

Should Uber

be allowed to

Compete

in Europe?

And if so How?

Damien Geradin

The disruptive market entry of Uber has gener-

ated lively debate, several taxi drivers’ demon-

strations and a number of court decisions all over

the world. The question of whether such entry

should be accommodated or banned by regu-

lators has been the object of thoughtful analysis

by policy-makers, judges and scholars. Damien

Geradin goes a step forward and, recognizing

that sooner or later innovative services will nd

their way to the market, explores different alter-

natives of making such transition smoothly and

compatible with EU and domestic regulations.

Abstract

Uber’s arrival in Europe has generated mas-

sive demonstrations by taxi drivers and a num-

ber of court judgments banning or restricting

Uber’s services on the ground that the company

engaged in “unfair competition”. Uber and oth-

er online-enabled car transportation services to

connect passengers with drivers offer an attrac-

tive alternative to regular taxi services.

The difculty is that these services are pro-

tected by regulatory measures that create sig-

nicant barriers to entry. Uber’s business mod-

el presents many efciencies and there is little

doubt that it will prevail over time. Regulatory

authorities thus face two options. One option

is to resist the market entry of Uber and other

similar companies. This approach would deprive

users of attractive services and trigger many

years of litigation. The other option is to em-

brace technological change and allow Uber

to compete on a level playing eld with taxi

companies. The regulatory changes that will

be needed raise complex questions, but these

questions are unavoidable and it is important to

tackle them early. Taxi companies can also em-

brace technologies and rely on the competing

online-enabled car transportation services that

are already available to them.

I. Introduction

While many industries are characterized by

constant innovation, the development of the

peer-to-peer economy has injected dynamism

in industries, which for a long time operated un-

der static business models. That is, for instance,

the case of the taxi industry. For almost a centu-

ry, taxi companies in all parts of the world relied

on a similar business model. Passengers can hire

taxis by queuing at a cab stand, by hailing them

in the street or by making a telephone reserva-

tion. Historically, technology played little role

in the industry, which is not surprising since taxi

services are subject to barriers to entry created

by regulatory intervention. Taxi regulations, for

instance, limit the number of vehicles authorized

to provide taxi services in a given locality. This

would not matter so much if the industry was

characterized by high levels of performance.

However, taxi fares remain often expensive

while the quality of the service can be mixed.

At certain periods of the day, taxis tend to be

scarce. Users may thus experience long waiting

times and, in some cases, taxis do not show up

Founding partner, EDGE Legal, a boutique competition and

litgationlawrmbasedinBrussels.ProfessorofCompetitionLaw&

Economics, Tilburg University; Professor of Law at George Mason Uni-

versity; visiting Professor, University College London. Email: dgera-

[email protected]

110

at all. This led some countries to engage dereg-

ulatory initiatives to improve the performance of

the taxi sector, but the often reverted to regula-

tion given the mixed results of these initiatives.

Until recently, it seemed that this sector was not

called to evolve and that users would have to

put up with the service as it is.

This situation changed with the arrival of

Uber and other providers of what I will refer to

as online-enabled car transportation services

to connect passengers with drivers.

While Uber

likes to call itself as a ride- or car-sharing service,

the ride or the car are not truly shared. What

characterizes Uber compared to regular taxi

services is that it is a marketplace where inde-

pendent drivers are connected to passengers

through an online platform. Uber’s mobile app

is user-friendly and its rates are generally attrac-

tive compared to the rates charged by regular

taxis.

That made the service popular in many

cities. While Uber has acionados among users

and investors,

it has however a large number of

enemies in the taxi industry. The arrival of Uber in

Europe has triggered massive protest from taxi

drivers and companies on the ground that Uber

does not comply with taxi regulations and there-

fore engages in “unfair competition”.

This led

taxi companies to seize the courts and Uber ac-

1 See, e.g., Ambrosius Baanders and Marcel Canoy, “Ten

Years of Taxi Deregulation in the Netherlands – The case for Re-reg-

ulation and Decentralisation”, Association for European Transport and

contributors, 2010, available at http://abstracts.aetransport.org/paper/

index/id/3411/cond/16

2 The notion of online-enabled platform has, for instance, has

been used by the California Public Utilities Commission (CPUC). See

infra footnote 45. While Uber is the best known online-enabled plat-

form, other companies such as Lyft or Sidecar provide comparable ser-

vices.

3 See, e.g., Sara Silverstein, “These Animated Charts Tell

You Everything About Uber Prices In 21 Cities”, Business Insider, 16

October 2014 at www.businessinsider.com/uber-vs-taxi-pricing-by-

city-2014-10

4 Sarah Mishkin, “Uber raises $1.6bn from Goldman clients”,

Financial Times, 21 January 2015 (indicating that Uber had a $40bn

valuation”)

5 See Matthew Field, “’Uber protest’ by black cab drivers

bringstrafcchaostoWestminster”,The Guardian, 26 May 2015.

tivities have been banned or subject to serious

restrictions in Member States, such as Belgium,

Germany,

Italy

and Spain.

Although some

public authorities are considering changes in

the applicable regulatory framework in order to

accommodate Uber and similar companies,

the situation remains chaotic.

Against this background, this short essay ar-

gues that the restrictions that have been placed

on Uber’s activities are undesirable as they de-

prive users of an attractive and innovative al-

ternative to regular taxi services. While some of

these restrictions can possibly be challenged

under EU law, this does not mean that Uber

should be allowed to operate in a legal void.

Innovation does not alter the need for measures

designed to ensure public safety, as well as to

protect users from various categories of risks. This

means that the regulatory framework should be

adapted to allow Uber to operate legally, as

well as to compete on a level playing eld with

taxi services. The legalization of Uber and similar

services raises, however, a number of complex

issues that will only be briey touched upon in

this essay. A complex question is, for instance,

whether online-enabled car transportation ser-

vices and taxi services should be subject to the

same regulatory regime or to separate regimes

adapted to their characteristics. Another ques-

tion is whether taxi companies and/or drivers

should be compensated for the loss of the in-

6 JamesFontanella-Khan,“€10,000nesthreatforUbertaxis

in Brussels”, Financial Times, 15 April 2015.

7 Eric Auchard and Christoph Steitz, “German court bans

Uber’s unlicensed taxi services”, Reuters, 18 March 2015, available at

www.reuters.com/article/2015/03/18/us-uber-germany-ban-idUSKBN-

0ME1L820150318

8 “Italian court bans unlicensed taxi services like Uber”, Re-

uters, 26 May 2015, available at www.reuters.com/article/2015/05/26/

us-italy-uber-idUSKBN0OB1FQ20150526

9 Harriett Alexander, “Judge in Spain bans Uber taxis”, The

Telegraph, 9 December 2014.

10 Frances Robinson, “Brussels to Propose New Laws Gov-

erning Uber”, 24 November 2014, available at blogs.wsj.com/dig-

its/2014/11/24/brussels-to-propose-new-laws-governing-uber/

111

vestment that they may have made in, for in-

stance, acquiring a license, the value of which

will considerably decrease following Uber’s

market entry. These questions will be briey ex-

amined in this essay and looked at in greater

detail in a subsequent paper.

This paper is divided in VI sections. Section II

provides a brief history of the regulation of taxi

services, which in some cities is almost one cen-

tury old. Section III describes Uber’s business

model and how it contrasts with the services

provided by traditional taxi companies. Section

IV discusses the EU law provisions that could be

used by Uber and other similar companies to

challenge the regulatory restrictions that pre-

vent them from offering their services in many

parts of the EU. Section V argues that the way

forward is for the relevant public authorities to

revisit the regulatory framework applied to taxi

services in order to allow Uber to compete legal-

ly against taxi companies. Section VI concludes.

II. A brief history of taxi regulation

Although taxi services are fairly basic in na-

ture (transporting passengers from point A to

point B) and do not require much capital or

skill (a car and a driver), they have been for a

long time subject to fairly intrusive regulation

with variations across countries and localities.

Among the reasons evoked for regulating taxi

services gure, for instance, the fact that in the

absence of control on entry there would be too

many taxis in the streets and this would create

congestion.

There has also been a fear, par-

ticularly during the great depression, that if taxis

were in excessive numbers, they would engage

in ruinous competition, which would in turn lead

11 For an historical perspective, see Paul Stephen Dempsey,

“Taxi Industry Regulation, Deregulation & Reregulation: The Paradox

of Market Failure”, (1996) 24 Transportation Law Journal 73.

12 See, e.g., House of Commons - Transport Committee, The

Regulation of Taxis and Private Hire Vehicle Services in the UK, Third

Report of Session 2003–04 Volume I, at p. 15, available at www.publi-

cations.parliament.uk/pa/cm200304/cmselect/cmtran/251/251.pdf

to low quality of service.

Regulation has also

be seen as necessary to correct information

asymmetries as, in the absence of rate control,

consumers would have no guarantee that the

fares they pay are fair and reasonable.

Simi-

larly, besides having a supercial look at the as-

pect of the car, users have no means to know

whether they will be driven in a safe vehicle.

Hence, regulatory requirements are needed to

ensure the safety of passengers.

As a result, with some degree of variation,

regulation of taxi services typically involves: (i)

control of entry (with local authorities, for in-

stance, setting the maximum number of vehi-

cles that can be used to provide taxi services);

(ii) licensing and performance requirements (for

the drivers and the taxi companies) designed,

for instance, to ensure safety standards for both

drivers (who need to receive proper training)

and vehicles (which must be inspected on a

regular basis); (iii) nancial responsibility stan-

dards (such as compulsory insurance); and (iv)

the setting of maximum rates based on various

methodologies.

The regulation of taxi services created, how-

ever, a series of problems, such as for instance

the insufcient availability of cars during peak

hours or in certain areas (seen as less protable

by drivers). In many instances, efforts to pre-

vent the oversupply of taxi services effectively

led to an undersupply of such services leading

to user discontent.

With the prices and quality

standards set by public authorities, taxi regula-

tions also did not incentivize taxi companies to

innovate or provide superior quality of service.

This led some countries or local authorities to de-

13 See Dempsey, supra note 11, at p. 77.

14 See “The Taxi Market in Ireland: To Regulate or Deregu-

late?”, Public Policy.IE, 23 October 2014, available at http://www.pub-

licpolicy.ie/the-taxi-market-in-ireland-to-regulate-or-deregulate/

15 Dempsey, supra note 11, at 75.

16 See “The Taxi Market in Ireland”, supra note 14.

112

regulate taxi services.

While in most cases, the

number of taxis increased, this did not necessar-

ily translate into lower waiting time and cheaper

services. In fact, some studies report that service

performance often decreased following dereg-

ulation,

which led authorities to re-regulate the

sector.

A critically important aspect is that these de-

regulatory efforts did not lead to major innova-

tion as new entrants essentially used the same

business model as incumbents. Of course, these

efforts occurred for the most part before the

development of the peer-to-peer platforms,

which, as observed in several industries (air trav-

el, hospitality, etc.), are true game changers in

that they are remarkably effective at matching

the demand with the supply of services without

the need for costly intermediaries. Thus, the fact

that deregulation did not bring innovation in the

past does not mean that it will not happen in

the future.

III. Uber’s disruptive business model

Uber is a marketplace connecting drivers

offering rides and passengers seeking them

through its mobile application.

A prospective

passenger who has downloaded the software

on his smartphone and set up a user account

can, when clicking on the application, see Uber

drivers near his location and on that basis sub-

mit a trip request which is then routed to the

drivers. The passenger is given an estimation on

17 See supra note 14. Note that in France, the report of the so-

called Commission Attali, which aimed to stimulate growth in France,

also suggested that taxi regulations should be relaxed in order to fa-

cilitate entry. See Rapport de la Commission pour la libération de la

croissance française, 2008, available at http://www.ladocumentation-

francaise.fr/var/storage/rapports-publics/084000041.pd, at 161.

18 See Dempsey, supra note 11, at 102 et seq.; Baanders and

Canoy, supra note 1; Roger F. Teal and Mary Berglund, “The Impacts of

Taxicab Deregulation in the USA”, Journal of Transportation Econom-

ics and Policy, January 1987, 37.

19 Id.

20 See Bill Gurley, “A Deeper Look at Uber’s Dynamic Pric-

ing Model”, Above the Crowd, available at http://abovethecrowd.

com/2014/03/11/a-deeper-look-at-ubers-dynamic-pricing-model/

how long his car will take to show up at his loca-

tion. Uber charges are based on a combination

of time and distance parameters and all pay-

ments are handled automatically by the Uber

service, which will charge the passenger’s busi-

ness card on le. Once destination is reached, a

receipt is sent automatically to the passenger’s

email address. On average 80% of the fares will

go to the driver, the rest being kept by Uber.

The strength of the Uber model is that it con-

siderably reduces search costs for users.

Rath-

er that calling a dispatcher and waiting without

knowing for sure whether and when the taxi will

show up, users can hail a car through Uber’s on-

line platform and watch it progress toward their

location. During periods when available cars

are scarce (e.g., Friday and Saturday nights),

Uber can incentivize drivers to take the road

by increasing their fees (a process referred to

as “dynamic” or “surge” pricing).

Dynamic or

surge pricing changes are “driven algorithmi-

cally when wait times are increasing dramat-

ically, and ‘unfullled requests” start to rise.”

Prices increase will at same time increase supply

as drivers will be incentivized to take the road to

earn higher fees, but also reduce demand as

price-sensitive users are incentivized to consid-

er alternatives, such as take their car or public

means of transport.

In sum, Uber’s business model offers several

advantages to users.

First, the software is ex-

tremely easy to use and it gives users a clear in-

dication of where the car they have just hired is

located, as well as the ability monitor its prog-

ress on the screen of its mobile devices. This re-

duces the anxiety associated with desperately

21 Id.

22 See “Pricing the Surge”, The Economist, 29 March 2014.

23 See Gurley, supra note 20.

24 Id.

25 It is important to note that some of these features are not

unique to Uber. Other companies like for instance Hailo also offer soft-

ware that allow users to hail cars through an online platform. See infra

footnote 41.

113

waiting for a taxi. Second, surveys suggest that

compared to regular taxi services, Uber prices

tend to be attractive.

Third, there is no need for

users to carry cash or cards as all transactions

are performed electronically. Finally, users can

rate their driver and thus incentivize the driver to

provide good service in order to boost his or her

reputational score.

Because of its attractive features, Uber’s en-

try on a given market is usually bad news for taxi

companies and their drivers. For instance, the

chart below indicates that during the period

between January 2012 and August 2014, cab

use in San Francisco declined 65%.

Figure : Average monthly number of trips per

cab (San Francisco)

This has led taxi drivers to vigorously protest

against Uber’s effort to penetrate their market,

as well as taxi companies to challenge the le-

gality of Uber’s activities before the courts. In re-

cent months, several national courts declared

that Uber services are illegal on various grounds

26 See supra note 3.

27 See Sergiy Golovin, “The Economics of Uber”, Bruegel,

30 September 2014, available at http://www.bruegel.org/nc/

blog/detail/article/1445-the-economics-of-uber/

(such as, for instance, the fact that Uber drivers

operate without a license and that Uber engag-

es in “unfair competition”).

As a result, Uber is

no longer able to provide services in some EU

Member States and operates in a “grey zone”

in many others. This is far from ideal for Uber and

the passengers who would like to use its services.

The question for Uber is of course to nd out

what it can do to restore its ability to provide its

service unimpeded by regulatory restrictions. As

will be seen in the next section, EU law may pro-

vide some solutions.

IV. Could public restrictions

preventing Uber to compete be

challenged under EU law?

Although Uber has announced that it has

led a complaint to the Commission against the

German and Spanish bans of its services,

it has

not revealed the legal arguments on which it

relies in its complaint. Prima facie, the EU trea-

ties contained several provisions that can be

invoked by companies whose

activities are impeded by pub-

lic restrictions of competition.

Whether these provisions can

be relied upon to challenge

these restrictions, however, de-

pends on the circumstances of

each case since the regulatory

frameworks applicable to taxi

services can vary considerably.

A rst possibility for Uber

would be to invoke Articles

101 and 102 TFEU with Article

4(3) TEU,

which provides for a

28 See supra notes 6 to 9.

29  SeeJulia Fioretti, “Uber les complaintsagainst German

and Spanish bans”, Reuters, 1 April 2015, available at www.reuters.com/

article/2015/04/01/uber-eu-complaint-idUSL6N0WY2TP20150401

30 The duty of “sincere cooperation” set out in Article 4(3)

114

duty of loyal cooperation between the EU and

the Member States.

In its case-law, the Court

of Justice of the European Union (“CJEU”) has

found that a Member State could breach its ob-

ligations under these provisions by adopting or

maintaining legislation that could deprive the

competition rules of their effectiveness.

The

application of this case-law, however, requires

the existence of an agreement contrary to Ar-

ticle 101(1) TFEU, which will be strengthened

or encouraged by the legislation in the Mem-

ber State.

In some cases, the CJEU also found

that Article 101 TFEU and Article 4(3) TEU were

breached when the Member State had del-

egated the power to x prices to operators.

In practice, this means that a pure regulatory

measure adopted by a public authority, i.e. a

decree regulating taxi services, cannot be chal-

lenged under these provisions unless this decree

transforms an anti-competitive agreement ad-

opted by taxi operators into binding law or,

alternatively, delegates to taxi operators the

power to set taxi fares or impose other regulato-

ry requirements.

Another possibility consists in invoking Articles

101 and 102 TFEU in conjunction with Article 106

TFEU. Article 106(1) TFEU provides that

“In the case of public undertakings and un-

dertakings to which Member States grant spe-

cial or exclusive rights, Member States shall nei-

TEU requires Member States to take appropriate measures to “ensure

fulllmentoftheobligationsarisingoutoftheTreatiesorresultingfrom

the acts of the institutions of the Union” as well as to “refrain from any

measure which could jeopardise the attainment of the Union’s objec-

tives”.

31 In theory nothing prevents to combine Article 102 TFEU

with Article 4(3) TEU, but this provision has been essentially applied in

the context of Article 101 TFEU. As will be seen below, State measures

raising issues in relation to abuses of a dominant position have usually

arisen in the context of Article 106(1) TFEU.

32 See Case 13/77, Inno v. ATAB, [1977] ECR. 2115.

33 See, e.g., Case 231/83, Cullet v. Leclerc, [1985] ECR. 305

(challengetoaxedminimumpricefailedbecausetheminimumprice

was a pure State measure unrelated to any agreement between competi-

tors).

34 See, e.g., Case 136/86, BNIC v. Yves Aubert, [1987] ECR.

4789; Case 66/86, Ahmed Saeed, [1989] E.C.R. 803.

ther enact nor maintain in force any measure

contrary to the rules contained in the Treaties, in

particular to those rules provided for in Article 18

and Articles 101 to 109.”

While there is an abundant case-law of the

CJEU in which Article 102 TFEU is combined with

Article 106(1) TFEU, the difculty in this case is that

the application of Article 106(1) requires that the

State measure in question, e.g. a decree regu-

lating taxi services, should benet companies

which have been granted exclusive or special

rights. While it cannot be excluded that a taxi

company could have been granted an exclu-

sive right to provide the service in a given lo-

cation, in the majority of the cases, the right to

provide such services is granted to the limited

number of companies or drivers that are allowed

to acquire a license. The question is thus whether

the licenses granted would amount to “special

rights” as understood in EU law.

While there is no

clear denition of the notion of special rights (as

the case law typically lumps together this notion

with the notion exclusive rights by referring to “ex-

clusive or special rights”), it can be inferred from

EU legislation that “special rights” concerns rights

that are granted by a Member State to two or

more undertakings within a given geographical

area.

Moreover, given the combination with Ar-

ticle 102 TFEU, the State measure in question must

maintain or strengthen a dominant position. Thus,

even if the taxi companies can be considered as

enjoying exclusive or special rights, they still need

to exercise a “single” or “collective” dominant

position, which is by no means a given.

A further possibility is to argue that the taxi

legislation in question breaches the free move-

ment provisions of the TFEU, such as Articles 49

(right to establishment) or 56 (freedom to pro-

vide services). As the CJEU observed “Articles

[49 and 56 TFEU] preclude any national measure

which, even though it is applicable without dis-

35 Article 1(4) of Directive 2008/63

36 See, e.g., Article 1.4 of Directive 2008/63 on competition

in the markets in telecommunications terminal equipment, O.J. 2008, L

162/20.

115

crimination on grounds of nationality, is liable to

prohibit, impede or render less attractive the ex-

ercise by Community nationals of the freedom

of establishment and the freedom to provide

services guaranteed by those provisions of the

Treaty.”

It should thus be possible to argue that

taxi regulations making very hard for companies

based in other Member States to provide their

services could fall foul with Articles 49 and/or 56

TFEU. Restrictions to the free movement princi-

ples contained in the TFEU are, however, per-

mitted when “those provisions are necessary to

meet overriding requirements of general public

importance […], whether they are proportion-

ate for that purpose and whether the aims or

overriding requirements could have been met

by less restrictive means.”

The question thus

becomes whether the restrictions that are con-

tained in taxi regulations in question are neces-

sary to meet overriding requirements of general

public importance and whether they propor-

tionate to achieve the objectives they seek to

protect. This is of course an intensely factual as-

sessment.

While the above approaches may help Uber

and other online-enabled car transportation

services to remove specic obstacles to the

provision of their services, they do not create

a framework allowing regular taxi services and

online-enabled car transportation services to

compete on a level playing eld. In my view,

allowing competition between regular taxi ser-

vices and online-enabled services requires an

overhauling of the various lawyers of taxi legisla-

tion in place in the Member States.

V. The need for a regulatory solution

From a high-level standpoint, the most effec-

tive way to set up a pro-competition regulatory

37 Case C-376/08, Serrantoni Srl, [2009] ECR. I-12169, at §

41.

38 Joined cases C-34/95, 35/95 and 36/95, De Agostini, [1997]

ECR I-3843, at § 52.

framework might be for the EU to adopt a Direc-

tive setting up the principles that should govern

the regulation of taxi services and online-en-

abled car transportation services, while leaving

the implementation of such principles to the

Member States. Such an approach would, how-

ever, be resisted on subsidiarity grounds as such

services are likely to be seen as a local matter.

Moreover, the elaboration of a proposal by the

Commission and its adoption by the Council of

Ministers and the European Parliament would

likely take several years to complete during

which Uber and other similar services would

continue to operate in a grey zone to the det-

riment of users. The better approach is thus for

the national authorities in charge of regulating

taxi services in the Member States to take the

initiative and elaborate regulatory frameworks

allowing taxi and online-enabled car transpor-

tation services to compete on a level playing

eld.

Conceptually, there are several alternative

ways to create such a pro-competitive frame-

work. First, regulatory authorities could create a

single framework applying to both taxi services

and online-enabled car transportation services.

The advantage of such an approach is that it

would ensure a high degree of convergence

in the ways in which these services are regulat-

ed. Yet, this approach would face several dif-

culties. Because the services proposed by taxi

companies and online-enabled car transporta-

tion services are currently so far apart, it may be

difcult to nd a regulatory regime suiting them

both. While incumbent taxi companies may

wish to ensure that Uber is forced to comply with

the same regulatory requirements as applying

to them, such an approach is a non-starter for

Uber and other online-enabled car transporta-

tion services as it would eviscerate their business

39 “Giventheessentiallylocalsignicanceoftaxiservicesand

in line with the principle of subsidiarity, existing Community legislation

intheeldoftransportdoesnotcovertaxiservices”,Answergivenby

Commissioner Tajani on behalf of the Commission to a Parliamentary

Question, 22 June 2009, E-3230/2009

116

model. Ideally, taxi companies should realize

that in the medium-term Uber’s business model

is likely to prevail and that it is therefore a mat-

ter of time before they will have to revisit their

modus operandi. In the short-term, such an

approach is, however, likely to be resisted be-

cause it would lead to job losses, as well as an

acceptance by taxi companies that the busi-

ness model they have so much decried is the

right way to go.

The commercial triumph of online-enabled

car transportation services is, however, only a

matter of time because of its inherent efcien-

cies. This is also what investors seem to think.

Thus, unless local authorities decide to protect

taxi companies through anti-competitive regu-

latory requirements, it will not take long before

all market actors realize it is in their own ben-

et to start relying on online platforms. Some

taxi companies already do so as alternatives

to Uber’s platform exist.

There might still be a

role for traditional taxis waiting for passengers at

cab stands, but traditional reservation models

will likely go away. This type of evolution is by no

means unique to the taxi industry as the power

of the Internet and online reservation systems

have already allowed consumers to largely do

away with travel agents.

There is no reason why

you would want to pay a fee to an intermediary

when you can book a ight, hotel accommo-

dation and a car as effectively. The difference,

however, between many industries affected

by the online platforms and the taxi industry is

that the latter is protected by regulation, hence

making the transition slow and difcult.

40 See supra note 4.

41 Some taxi companies are, for instance, relying on Hailo,

technology platform that matches taxi drivers and passengers through

its mobile phone application. See https://www.hailoapp.com/

42  See Danny King, “Report nds agents losing ground to

online”, mobile bookings, 23 December 2014, available at www.trav-

elweekly.com/Travel-News/Travel-Agent-Issues/Report-nds-agents-

losing-ground-to-online-and-mobile-bookings/

43 Although companies, such as Airbnb are also facing regula-

tory changes, see Roberta A. Kaplan and Michael L. Nadler, “Airbnb:

In the meantime, the better approach is prob-

ably to set up a new regulatory regime speci-

cally designed for online-enabled car transpor-

tation services.

The challenge for this regime

will be to allow Uber and other similar compa-

nies to compete on the merits with regular taxi

services. This means that this regime should be

no less favorable than that the regime being ap-

plied to regular taxi services. After all, in all reg-

ulated industries, new entrants are subject to a

lighter regular burden than incumbents.

While

there is perhaps no reason why Uber should

benet from a more favorable regime than taxi

services, there is no reason either why it should

penalized for offering an attractive alternative

to these services and create competition in a

rather amorphous sector. What equal treatment

means is difcult to determine in the abstract as

it largely depends of the local circumstances,

but it should be one of the guiding principles of

the regime that applies to online-enabled car

transportation services.

A Case Study in Occupancy Regulation and Taxation”, 82 (2015) U

Chi L Rev Dialogue 103, https://lawreview.uchicago.edu/page/airb-

nb-case-study-occupancy-regulation-and-taxation

44 This is, for instance, what has been done in California where

theCaliforniaPublicUtilitiesCommission(CPUC)createdaspecic

regime to apply to “companies that provide prearranged transportation

services for compensation using an online-enabled application (app)

or platform to connect passengers with drivers using their personal

vehicles”. See CPUC Establishes Rules for Transportation Network

Companies, Press Release, 19 September 2013, available at http://docs.

cpuc.ca.gov/PublishedDocs/Published/G000/M077/K132/77132276.

PDF This regime establishes 28 rules and regulations for Transporta-

tion Network Companies whereby they must obtain a license from the

CPUC to operate in California; require each driver to undergo a crimi-

nal background check; establish a driver training program; implement a

zero-tolerance policy on drugs and alcohol; hold a commercial liability

insurance policy that is more stringent than the CPUC’s current require-

ment for limousines, requiring a minimum of $1 million per-incident

coverage for incidents involving TNC vehicles and drivers in transit to

or during a TNC trip, regardless of whether personal insurance allows

for coverage; and conduct a 19-point car inspection. Id.

45 That is, for instance, the case in the electronic communica-

tionseldwhereonlycompanieswithsignicantmarketpower(typi-

cally the incumbent telecommunications operator) are subject to regu-

latory remedies. See European Commission, Regulatory framework for

electronic communications in the European Union, available at https://

ec.europa.eu/digital-agenda/sites/digital-agenda/les/Copy%20of%20

Regulatory%20Framework%20for%20Electonic%20Communica-

tions%202013%20NO%20CROPS.pdf

117

Prima facie, some regulatory requirements

should be equally imposed to taxi drivers and

operators, and to Uber and its drivers. That is,

for instance, the case of safety standards. There

is no reason why Uber cars should escape the

safety controls that apply to taxi vehicles. Sim-

ilarly, Uber drivers should not be less qualied

or trained than taxi drivers, and they should be

subject to background checks. There may also

be some areas where online-enabled car trans-

portation services should be subject to regulato-

ry requirements that do not necessarily apply to

taxi services. One example relate to the usage

of personal data. Uber is able to collect sensi-

tive information about its passengers, such as

their locations at various moments in time. They

also collect nancial information, such as credit

card details, etc. It is, however, not clear that

Uber and other online-enabled car transporta-

tion services should be subject to specic reg-

ulation regarding the storage and treatment of

their passengers’ personal and nancial data as

“horizontal” legislation typically exists prevent-

ing holders of such data to misuse them.

Now, if a specic regime is created for on-

line-enabled car transportation services, it is

subject to question whether taxi regulations

should also be modied. Does it, for instance,

make sense to continue to control taxi fares

when they are subject to competition from

Uber and other similar companies? This is a par-

ticularly difcult question, although one answer

may be to maintain price regulation until such

time online-enabled car transportation services

have captured a certain size of the market (as-

suming that there are in the same market as taxi

services, which is a complex question I do not

address here) and are thus able to exercise suf-

cient pressure to keep taxi rates at bay.

Another complex question in this respect is

46 See, e.g., Directive 95/46/EC of the European Parliament

and of the Council of 24 October 1995 on the protection of individuals

with regard to the processing of personal data and on the free movement

of such data, 1995 OJ L281/31.

whether incumbent taxi operators and/or driv-

ers should be compensated for the investment

they may have made in acquiring a license to

operate taxi services. This question may be par-

ticularly acute in cities where such licenses (or

medallions as they are called in some places)

trade at very high prices.

Giving the right an-

swer to this question is not necessarily easy, al-

though pioneering work has been done on this

type of issue.

On the one hand, offering com-

pensation may facilitate regulatory change as

it will make change easier to accept for the

likely losers. On the other hand, granting com-

pensation to taxi drivers or operators that have

invested in acquiring a license may create a

host of problems. First, compensation may not

be deserved when the investment has been ful-

ly amortized. Second, compensation creates a

problem of valuation.

Calculating the amount

to which a driver should be allowed will not be

simple and alleged calculation errors will lead

to litigation. Third, the prospect of obtaining

compensation in case of change of regulatory

regime may incentivize operators in a variety of

elds to try to obtain exclusive or special rights

from public authorities, hence reducing compe-

tition.

Finally, as the taxi industry is not the only

sector that is sheltered from competition by reg-

ulation, liberalizing the economy may become

a very expensive proposition that may induce

local authorities to not engage in desirable re-

forms. It thus seems on balance that there are

more reasons not to grant compensation to taxi

drivers or operators than to grant them com-

pensation for the investment losses that may

47 See Josh Barro, New York City Taxi Medallion Prices Keep

Falling, Now Down About 25 Percent, The Upshot, New York Times, 7

January 2015.

48 For an excellent discussion of the question of compensation,

see Michael J. Trebilcock, Dealing with Losers – The Political Econo-

my of Policy Transitions, Oxford University Press, 2014.

49 See Edmund W. Kitch, “Can we Buy our Way out of Harm-

ful Regulation” in Donald L. Martin and Warren F. Schwartz, Deregu-

lating American Industry: Legal and Economic Problems, Lexington

Books, 1976, 51, 54.

50 Id. at 52.

118

incur when Uber and other companies are al-

lowed to operate legally.

Another possible approach to address the

investment issue is to open the market to on-

line-enabled car transportation services only

gradually, hence giving taxi companies the

time to adapt to the arrival of Uber and other

companies providing similar services. That is the

approach that has been taken by the EU when

it decided to liberalize network industries, such

as telecommunications, energy and posts. One

of the reasons for this gradual approach is that it

was a political compromise between pro-liber-

alization Member States and anti-liberalization

ones. The need to broker such a compromise

would of course not be needed if the decision

to open the market to online-enabled car trans-

portation services is taken at a local level. More-

over, the problem with gradual liberalization

in this case is that it would unavoidably delay

the benets of competition by several years at

the expense of consumers. In any event, open-

ing the market in this case would be nowhere

as near complicated in legal and institutional

terms than opening the market in network in-

dustries, which for instance required the set-

ting-up of access to the network regimes and

the adoption of measures designed to protect

universal service.

These are some of the difcult questions that

will face local authorities seeking to develop a

regulatory regime allowing online-enabled car

transportation services to operate legally.

VI. Conclusion

While Uber has been subject to a great deal

of criticism, there is no doubt that it offers an

attractive alternative to regular taxi services.

There is therefore no reason why Uber and oth-

er online-enabled car transportation services

to connect passengers with drivers should not

be allowed to compete on a level playing eld

with taxi companies. Because taxi companies

are protected by regulation, it is for public au-

thorities to take the initiative. These authorities

have two options. One option is to resist Uber’s

market entry and face many years of litigation,

which will eventually result in Uber being able to

operate legally. The other, preferable, option is

to embrace technological change and adopt

a regulatory framework allowing Uber and oth-

er similar companies to compete. This does not

mean that Uber should be allowed to operate

free of regulation. For instance, passenger safe-

ty should remain a priority. As to the taxi com-

panies, they do not need to remain passive by-

standers waiting for their market share to be lost

to Uber and other similar companies. They can

also embrace change by, for instance, relying

on other existing online platforms.

119

Online Intermediation

Platforms and Free

Trade Principles –

Some Reections on

the Uber Preliminary

Ruling Case

Damien Geradin

Abstract

Commercial Court No 3 of Barcelona sent

a request for a preliminary ruling to the CJEU

regarding the extent to which Uber which op-

erated its uberPOP service in Spain without an

authorization from the Spanish authorities should

be protected by EU law provisions designed to

ensure the free movement of services in the

European Union. The paper demonstrates that

uberPOP is not a “transport service” falling under

under Title VI TFEU, but an “information society

service” within the meaning of the E-commerce

Directive. Therefore, uberPop benets from the

protection against undue trade restrictions pro-

vided by this directive, as well as by Article 56

TFEU. This implies that regulatory requirements

that do not protect a public interest objective

in a proportionate and non-discriminatory man-

ner are incompatible with EU law. The judgment

of the CJEU will have signicant implications on

the way EU Member States are able to regulate

Uber services, but also the services provided by

other intermediation platforms in the future.

I. Introduction

On 7 August 2015, the Commercial Court No

3 of Barcelona (the “referring Court”) sent a re-

quest for a preliminary ruling to the Court of Jus-

tice of the European Union (the “CJEU”) regard-

ing the extent to which Uber Systems Spain, S.L.,

which operated its services in Barcelona, as well

as in Madrid and Valencia, without authoriza-

tion from the Spanish authorities, should be pro-

tected by EU law provisions designed to ensure

the free movement of services in the European

Union (“EU”).

While the CJEU is unlikely to render

its judgment before the end of 2017, the ques-

tions raised by the Barcelona Court are of great

practical importance as the answers that will

be given to them are likely to shape the ways in

which Uber, but also other intermediation plat-

forms, will be regulated in the European Union.

Uber is a marketplace connecting drivers

offering rides and passengers seeking them

through its mobile software application.

A pro-

spective passenger who has downloaded the

Uber software on his smartphone and set up a

user account can, when clicking on the appli-

cation, see Uber partner-drivers near his location

and on that basis submit a trip request which is

then routed to the partner-drivers. The passen-

ger is given an estimate of how long the driv-

er who has picked up the ride will take to show

up at his location. Uber’s charges are based on

a combination of time and distance parame-

ters and all payments are facilitated by Uber.

Once destination is reached, a receipt is sent

automatically to the passenger’s email address.

On average 80% of the fares will go to the driv-

er, the balance representing the commission

1 Case C-434/15: Request for a preliminary ruling from the

Juzgado Mercantil No 3 de Barcelona (Spain) lodged on 7 August 2015

— Asociación Profesional Élite Taxi v Uber Systems Spain, S.L., OJ

2015, C 363/21.

2 See generally, Damien Geradin, “Should Uber be Allowed

to Compete in Europe? And if so How?”, Competition Policy Interna-

tional, 18 June 2015, available at https://www.competitionpolicyinter-

national.com/should-uber-be-allowed-to-compete-in-europe-and-if-so-

how/

Professor of Competition Law & Economics at the Tilburg

Law & Economics Center (TILEC), Tilburg University; Professor of

Law, George Mason University School of Law; and visiting Professor

of Law, University College London. Founding partner, EDGE Legal,

Brussels ([email protected]). The author thanks his colleagues

Panagiotis Delimatsis for his helpful observations.

120

charged by Uber. Uber offers various categories

of services, such as uberPOP (relying on drivers

not professionally licensed) and UberX (relying

on professionally licensed drivers).

Although Uber’s business model is distinct

from traditional taxi companies, these compa-

nies see Uber as a major threat, and individu-

al taxi companies and taxi associations have

led lawsuits against Uber in several EU Member

States on the ground that, because it does not

comply with the license obligations imposed on

taxi companies, Uber would engage in “unfair

competition”.

This is one of such cases, where a

trade association called Asociación Profesional

Élite sued Uber, which led to the preliminary rul-

ing to the CJEU. This case opposes two distinct

logics. One the one hand, taxi companies see

no reason why Uber should be exempted from

the licensing requirements imposed on them

by public authorities. On the other hand, Uber

considers that unlike taxi companies it does not

provide transport services, but online intermedi-

ation services connecting prospective passen-

gers with drivers.

Uber’s view is that because of

the nature of its services, it should not have to

comply with the same requirements as taxi com-

panies as these requirements go beyond what

is needed to correct market failures, and do not

support innovation and consumer choice.

Using the questions asked by the Barcelona

Court to the CJEU as a departing point, the

objective of this paper is to help clarifying the

nature of the services provided by Uber, i.e.

whether Uber’s services are transport services or

intermediation services, as well as the regulato-

3 With variation from country to country, Uber offers a va-

riety of other services, such as for instance uberPOOL (a car pooling

service), uberRUSH (a courier package delivering service) or even

uberBOAT (a boat service across the Bosphorus).

4 See, e.g., Alissa J. Rubin and Mark Scott, “Clashes Erupt

Across France as Taxi Drivers Protest Uber”, New York Times, 25 June

2015, available at www.nytimes.com/2015/06/26/business/internation-

al/uber-protests-france.html?_r=0

5 See, e.g., Christine Lagorio-Chafkin, “Uber Advisor David

Plouffe: We’re ‘Not Driven by Greed’”, Inc., 3 November 2015, avail-

able at www.inc.com/christine-lagorio/uber-david-plouffe-gig-econo-

my-greed.html

6 Id.

ry implications of this distinction. An issue at the

core of the questions asked by the Barcelona

Court is the extent to which Uber should benet

from the free movement provisions contained

in the so-called “Services” and/or the “E-Com-

merce” directives, as well as the free movement

provisions of the Treaty on the Functioning of the

European Union (“TFEU”). In this paper, I volun-

teer some observations that should ideally assist

the CJEU in its analysis of these questions, as well

as provide a broader perspective on the regu-

latory challenges created by online intermedia-

tion platforms.

The issues discussed in this paper are not only

relevant to Uber, but also to other online inter-

mediation platforms, such as Airbnb. It is, for in-

stance, hotly debated whether Airbnb should

be allowed to escape the regulatory obliga-

tions imposed on hotels, such as for instance

safety rules and zoning requirements, as well as

various taxes.

However, the regulatory require-

ments applied to industries that are vulnerable

to challenges by online platforms contain mea-

sures designed to protect incumbents, hence

creating barriers to entry. The extent to which

these requirements are compatible with the free

movement principles contained in the Treaty on

the TFEU is thus an important question.

This paper is divided in six parts. Part II de-

scribes Uber’s activities in Europe, as well as the

way in which its smartphone software applica-

tion functions. Part III discusses the Spanish law-

suit, which led to the request for a preliminary

ruling by the Commercial Court No 3 of Barce-

lona. Part IV analyses the various legal issues

raised by the referring court and, in particular,

whether uberPOP is a “transport service” or an

“information society service”. Part V outlines the

implications that the answers given by the CJEU

to the questions raised by the referring court will

have beyond Uber’s activities. Finally, Part VI

concludes.

7 Ron Lieber, A Warning for Hosts of Airbnb Travelers,

New York Times, 30 November 2012, available at www.nytimes.

com/2012/12/01/your-money/a-warning-for-airbnb-hosts-who-may-

be-breaking-the-law.html?_r=0

121

In my view, uberPOP is not a “transport ser-

vice”, but an “information society service”.

Therefore, it falls within the scope of Directives

2006/123 and 2000/31, as well as under Article

49 and 56 TFEU. This does not mean that Uber’s

activities cannot be regulated, but that regula-

tion should be limited to protecting objectives

of public interest and they should be proportion-

ate to the accomplishment of these objectives.

In other words, licensing requirements and other

rules that may apply to Uber’s activities and oth-

er similar companies should be limited to what

is necessary to correct market failures. Measures

that have the object or effect to restrict entry

should be struck down unless they can be ob-

jectively justied.

II. Uber’s activities in Europe

Since its creation in 2009, Uber Technologies

(“Uber”) has developed a large worldwide pres-

ence, including in the European Union. As of

January 2016, Uber was active in over 20 Mem-

ber States and 50 European cities.

Uber has created a smartphone application

that allows it to provide a range of services. In

the EU, the two main services provided by Uber

are: (i) uberPOP , which Uber refers to as a

“peer-to-peer rideshare service” which enables

a rider to “share” the use of a vehicle with the

driver and owner of that vehicle (which Uber re-

fers to as “partner-drivers”) against the payment

of a fee; and (ii) uberX, which is a professional

transportation service provided by licensed

“private hire vehicle” (“PHV”) drivers operating

licensed private hire vehicles.

Thus, a key differ-

ence between these services is that while uberX

drivers are professional drivers who need to sat-

isfy the conditions applicable to licensed PHV

drivers, uberPOP drivers are non-professional

8 See https://www.uber.com/cities (last visited on 14 January 2016)

9 Uber has developed additional services, such as, for in-

stance, UberEATS, a service which allows a customer to order food

from a restaurant and have it home delivered.

drivers, who have to satisfy a number of condi-

tions set by Uber regarding their credentials and

car, such as the ownership of a driving licence,

a clean criminal record, proof of insurance for

the car, certicate of third party liability insur-

ance, etc.

The services which form the basis of the pre-

liminary ruling discussed in this paper, are pro-

vided by means of a smartphone application,

licensed by Uber BV to end user licensees locat-

ed in Spain, i.e. it is downloaded and installed

both by drivers and riders. This application li-

censed by Uber BV, a company located in the

Netherlands, performs a variety of intermedia-

tion functions, such as matching potential riders

with drivers, calculating the price to be paid

by the rider to the driver, facilitating the pay-

ment transaction of the service between the

rider and the driver, allowing drivers and riders

to rate each other through a one-to-ve stars

system, etc. Uber does not employ drivers (they

are self-employed) and does not own any cars

(they are owned by the drivers). It is a software

rm.

Uber’s software application is a source of

considerable efciencies, which has contribut-

ed to its success among users and investors.

However, Uber has a large number of enemies

in the taxi industry. The arrival of Uber in Europe

has triggered massive protest from taxi drivers

and companies on the ground that Uber does

not comply with the licensing and other re-

quirements that apply to them and therefore

engages in “unfair competition”.

This led taxi

companies to le lawsuits in the national courts

and uberPop has been banned in Member

10 See http://www.uberkit.net/blog/uberx/

11 Sarah Mishkin, “Uber raises $1.6bn from Goldman clients”,

Financial Times, 21 January 2015 (indicating that Uber had a $40bn

valuation”)

12 See Matthew Field, “’Uber protest’ by black cab drivers

bringstrafcchaostoWestminster”,The Guardian, 26 May 2015.

122

States, such as Belgium,

Germany,

Italy

and

Spain.

Although some public authorities are

considering changes to the applicable regula-

tory framework in order to accommodate Uber

and similar companies,

the situation remains

largely chaotic.

III. The lawsuit leading to the

preliminary ruling

In 2014, Uber BV started providing its uberPOP

service in Spain in the cities of Barcelona, Ma-

drid and Valencia. Based on the EU free move-

ment legislation on information society services,

Uber BV (or any of its related entities) did not re-

quest any prior authorization to the competent

Spanish authorities before initiating its activities

in Spain.

On 27 October 2014, Asociación Profesion-

al Elite Taxi, a trade association, led a lawsuit

against Uber Systems Spain on the ground that

its uberPOP service allegedly breached Spanish

Act 3/1991 on Unfair Competition (“Unfair Com-

petition Act”).

Specically, the plaintiff request-

13 JamesFontanella-Khan,“€10,000nesthreatforUbertaxis

in Brussels”, Financial Times, 15 April 2015.

14 Eric Auchard and Christoph Steitz, “German court bans

Uber’s unlicensed taxi services”, Reuters, 18 March 2015, available at

www.reuters.com/article/2015/03/18/us-uber-germany-ban-idUSKBN-

0ME1L820150318

15 “Italian court bans unlicensed taxi services like Uber”, Re-

uters, 26 May 2015, available at www.reuters.com/article/2015/05/26/

us-italy-uber-idUSKBN0OB1FQ20150526

16 Harriett Alexander, “Judge in Spain bans Uber taxis”, The

Telegraph, 9 December 2014.

17 Frances Robinson, “Brussels to Propose New Laws Gov-

erning Uber”, 24 November 2014, available at blogs.wsj.com/dig-

its/2014/11/24/brussels-to-propose-new-laws-governing-uber/

18 Inparallel,asimilarlawsuithadbeenledbytheAsocia-

ción Madrileña del Taxi before the Commercial Court nº 2 of Madrid.

On 9 December 2014, this Commercial Court adopted inaudita parte

a preliminary injunction with an order to Uber Technologies Inc.

to cease operations in the whole territory of Spain. The injunction also

ordered local telecommunications and payments operators to block

trafcandtransactionswithwww.uber.comandtoUbersoftware

application. Uber BV complied with the order and suspended its elec-

tronic intermediation of uberPOP in Spain. See “Uber taxi app suspend-

ed the Commercial Court nº 3 of Barcelona to

declare that Uber Systems Spain had infringed

Article 5 (covering misleading practices that

can lead to a signicant distortion of the eco-

nomic behaviour of the consumer) and Article

15(2) (declaring unfair non-compliance with a

provision regulating competitive activity) of the

Unfair Competition Act.

The issue at the core of this case relates to

whether Uber’s services can benet from the

free movement provisions included in the TFEU,

and in particular Article 49 (free establishment)

and 56 (free movement of services), but also

from the guarantees included in:

• Directive 2006/123/EC of the European Par-

liament and of the Council of 12 December

2006 on services in the internal market (the

“Services Directive”),

Article 9(1) of which

provides that Member States should not

make access to a service activity or its ex-

ercise “subject to an authorization scheme

unless the following conditions are satised:

(a) The authorization scheme does not dis-

criminate against the provider in question;

(b) The need for an authorization scheme

is justied by an overriding reason relat-

ing to the public interest ; (c) The objective

pursued cannot be attained by means of

a less restrictive measure, in particular be-

cause an a posteriori inspection would take

place too late to be genuinely effective.”

• Directive 2000/31/EC of the European Par-

liament and of the Council of 8 June 2000

ed in Spain”, BBC News, 9 December 2014, available at www.bbc.com/

news/business-30395093

19 It is important to note that Uber Systems Spain only pro-

vides marketing and support services to Uber BV and its related entities.

It does not operate or license the software application service. There-

fore, Uber Systems Spain, S.L. has argued in the Spanish national

litigation that the subject-matter of the main proceedings – namely

thequalicationoftheserviceprovidedbytheUbercompany–pertains

to Uber BV’s activity not that of Uber Systems Spain.

20 O.J. 2006, L 376/36.

123

on certain legal aspects of information so-

ciety services, in particular electronic com-

merce in the Internal Market (“E-commerce

Directive”),

Article 3.2 of which provides

that Member States may not restrict the

freedom to provide information society ser-

vices from another Member State. Pursu-

ant to Article 3.4, Member States may only

adopt measures derogating from the free-

dom to provide information society services

if such measures are: (i) necessary for rea-

sons of public policy, the protection of pub-

lic health, public security, the protection of

consumers; (ii) taken against a given infor-

mation society service which prejudices the

objectives referred to in point (i) or which

presents a serious and grave risk of preju-

dice to those objectives; (iii) proportionate

to those objectives.

Whether or not uberPOP can benet from

these guarantees depends on how this service is

dened and, in particular, whether it can qualify

as an “information society service” which is “any

service normally provided for remuneration, at

a distance, by electronic means and at the indi-

vidual request of a recipient of services” as per

Article 1.2 of Directive 98/34/EC laying down a

procedure for the provision of information in the

eld of technical standards and regulations as

amended by Directive 98/48/EC.

Against that background the referring Court

suspended its proceedings and decided to

address four questions for preliminary ruling to

the CJEU (which I streamline a bit for reason of

space):

1. Inasmuch as Article 2(2)(d) of Directive

2006/123/EC … excludes transport activities

from the scope of that directive, must the

21 O.J. 2000, L 178/1.

22 Directive 98/48/EC of the European Parliament and of the

Council of 20 July 1998 amending Directive 98/34/EC laying down a

procedurefortheprovisionofinformationintheeldoftechnicalstan-

dards and regulations, O.J.1998, L 217/18.

activity carried out for prot by the defen-

dant, consisting of acting as an intermediary

between the owner of a vehicle and a per-

son who needs to make a journey within a

city, by managing the IT resources … which

enable them to connect with one another,

be considered to be merely a transport ser-

vice or must it be considered to be an elec-

tronic intermediary service or an information

society service, as dened by Article 1(2) of

Directive 98/34/EC … laying down a proce-

dure for the provision of information in the

eld of technical standards and regulations

and of rules on Information Society services?

2. Within the identication of the legal nature

of that activity, can it be considered to be

… in part an information society service,

and, if so, ought the electronic intermediary

service to benet from the principle of free-

dom to provide services as guaranteed in

the Community legislation — Article 56 TFEU

and Directives 2006/123/EC and … 2000/31/

EC?

3. If the service provided by UBER SYSTEMS

SPAIN, S.L. were not to be considered to

be a transport service and were therefore

considered to fall within the cases covered

by Directive 2006/123, the question arising is

whether Article 15 of the Law on Unfair com-

petition … is contrary to Directive 2006/123,

specically Article 9 on freedom of establish-

ment and authorisation schemes, when the

reference to national laws or legal provisions

is made without taking into account the fact

that the scheme for obtaining licences, au-

thorisations and permits may not be in any

way restrictive or disproportionate, that is, it

may not unreasonably impede the principle

of freedom of establishment.

4. If it is conrmed that Directive 2000/31/EC

is applicable to the service provided by

UBER SYSTEMS SPAIN, S.L., the question aris-

124

ing is whether restrictions in one Member

State [regarding] the freedom to provide

the electronic intermediary service from an-

other Member State, in the form of making

the service subject to an authorisation or a

licence, or in the form of an injunction pro-

hibiting provision of the electronic intermedi-

ary service based on the application of the

national legislation on unfair competition,

are valid measures that constitute deroga-

tions from paragraph 2 in accordance with

Article 3(4) of Directive 2000/31/EC.

These questions are not free of ambiguities.

First, it is not clear what the referring court un-

derstands by a situation where uberPOP could

be “in part” an information society service. Sec-

ond, I assume that in question 3, the issue is not

whether Article 15 of the Spanish Unfair Com-

petition Act breaches Directive 2006/123, but

whether its application in the case at hand vi-

olates Directive 2006/123. Moreover, as Article

15 considers as “unfair” a rm’s non-compliance

with a provision regulating a competitive activ-

ity, question 3 seems to presume that uberPOP

breaches Spanish legislation, which is not nec-

essarily a given if it is not a “transport service”, as

I will argue below. Third, while questions 3 and 4

refer to the services provided by Uber Systems

Spain, uberPOP is provided by Uber BV, Uber

Systems Spain’s activities being limited to mar-

keting and support. The fact that the Uber com-

pany the court has effectively in mind is Uber

BV (based in the Netherlands) rather than Uber

Systems Spain is conrmed by the following pas-

sage of question 4 whereby the question aris-

ing is “whether restrictions in one Member State

[regarding] the freedom to provide electron-

ic intermediary service from another Member

State.” (emphasis added)

IV. Analysis of the issues raised by

the referring court

23 See supra, note 1.

The issues at the core of this case are (i) wheth-

er the uberPOP service should be characterized

as a “transport service” or an “information soci-

ety service” or both and (ii) the regulatory impli-

cations of this characterization.

In order to address these issues, this part is di-

vided in four sections. Section A seeks to deter-

mine whether uberPOP should be characterized

as a “transport service” or an “information soci-

ety service” or both. Section B, C and D analyze

the regulatory implications of the answers which

can be given to the nature of Uber’s activities

and, in particular, whether – depending on the

answer given – these activities can be protect-

ed under Directives 2000/31 and/or 2006/123,

and/or the free movement of services and the

freedom of establishment rules contained in the

TFEU.

A. Is uberPOP a “transport service” or an

“information society service” or both?

Directive 2000/31 denes “information soci-

ety services” as services within the meaning of

Article 1(2) of Directive 98/34 as amended by

Directive 98/48. In turn, Directive 98/34 denes

such services as “any service normally provided

for remuneration, at a distance, by electronic

means and at the individual request of a recipi-

ent of services” and species that for the purpos-

es of this denition: “‘at a distance’ means that

the service is provided without the parties being

simultaneously present; ‘by electronic means’

means that the service is sent initially and re-

ceived at its destination by means of electronic

equipment for the processing (including digital

compression) and storage of data, and entire-

ly transmitted, conveyed and received by wire,

by radio, by optical means or by other electro-

magnetic means, and ‘at the individual request

of a recipient of services’ means that the service

is provided through the transmission of data on

individual request.”

125

There is little doubt that uberPOP meets the

different elements required by Directive 98/34

to be dened as an information society ser-

vice. First, uberPOP is provided “at a distance”

since the service is provided without the driver

and the rider being simultaneously present. Sec-

ond, uberPOP is clearly provided “by electron-

ic means”, i.e. a mobile software application.

Third, uberPOP is provided “at the individual

request of a recipient of services” in that Uber

provides its services to both users and drivers,

which request the service by connecting to the

Uber platform. Finally, uberPOP is a “service nor-

mally provided for remuneration”. While Uber’s

services are frequently referred to as “ride shar-

ing”,

there is a clear pecuniary element to the

transaction in that the driver expects a payment

for transporting the rider.

In fact, Uber can be described as an online

“market making” platform in that it connects

producers (in this case drivers) with “consumers”

(riders) and facilitates their interactions and ex-

changes. In other words, Uber does not create

value by performing transport services, but by

enabling direct interactions between two dis-

tinct categories of users. Like other platforms,

such as eBay or Airbnb, Uber’s platform is also

two-sided in that the two sides that the platforms

connects (partner-drivers and riders) are linked

by “indirect network effects” in that a large

number of drivers benets riders, and vice-ver-

sa.

Traditional taxi companies hold none of

these features, although companies that limit

their activities to dispatching services can also

probably be considered as market-making plat-

forms.

Another question is whether uberPOP should

24 See, e.g., Mark Harris, “Uber: why the world’s biggest

ride-sharing company has no drivers”, The Guardian, 16 Novem-

ber 2015, available at http://www.theguardian.com/technology/2015/

nov/16/uber-worlds-biggest-ride-sharing-company-no-drivers

25 See, “Everybody wants to Rule the World”, The Econ-

omist, 29 November 2014, available at www.economist.com/news/

brieng/21635077-online-businesses-can-grow-very-large-very-fastit-

what-makes-them-exciting-does-it-also-make

be considered as a “transport service” and

therefore be excluded from the scope of Di-

rective 2006/123. Article 2(2)(d) of Directive

2006/123 provides that “[t]his Directive shall not

apply to services in the eld of transport, includ-

ing port services, falling within the scope of Title

V of the Treaty [now Title VI TFEU],” but it does not

elaborate further on what a service “in the eld

of transport” within the scope of Title VI means.

However, Recital 21 of Directive 2006/123 pro-

vides that “[t]ransport services, including urban

transport, taxis and ambulances as well as port

services, should be excluded from the scope of

this Directive.”

Although such a conclusion may be tempting,

recital 21’s reference to “taxis” does not neces-

sarily mean that uberPOP should be considered

as a transport service. As pointed out by Advo-

cate General Wahl in its June 2015 Opinion in

Grupo Itevelesa, when a service falls under Title

VI TFEU neither the freedom to provide services

(Article 58(1) TFEU) nor Directive 2006/123 (Arti-

cle 2(2)(d thereof) apply, hence the exercise of

dening what constitutes a “service in the eld

of transport” must be carried out “with care”,

“especially in respect of services which are only

incidental, ancillary or even tangentially con-

nected to transport.”

In this respect, Advocate

General Wahl observes that

“a ‘service in the eld of transport’ must con-

sist of or be inherently linked to the physical act

of moving persons or goods from one place to

another by means of a vehicle, aircraft or wa-

terborne vessel. If the service in question does

not mainly involve actual transport, then the

mere fact that it can be linked in one way or

another to transport does not, in itself mean that

it ought to be characterised as such.” (empha-

26 Opinion of Advocate General Wahl delivered on 3 June

2015, Case C-168/14, Grupo Itevelesa et al. v. OCA Inspección Técnica

de Vehículos SA, Generalidad de Cataluña, [2015] E.C.R I-0000, at §

22.

27 Id.

126

sis added)

In Trijber, Advocate General Spuznar fol-

lowed a similar line of analysis.

In that case,

which concerned the transportation of passen-

gers by open sloop on the internal waterways

of Amsterdam for entertainment purposes, he

reected on the applicability rationae materiae

of Directive 2006/23. In this respect, he observed

that “the meaning and scope of a term must

be determined by considering its usual mean-

ing in everyday language, while also taking into

account the context in which it occurs and the

purposes of the rules in which it is part.”

that basis, he concluded that “where the main

purpose of the activity is not the physical con-

veying of goods or people but others matters

[…], one cannot speak of services in the eld of

transport.”

The CJEU followed the approach proposed

by Advocate General Spuznar in that it ob-

served that while transport by inland waterway

fell within Title VI TFEU, it does not mean that

“any service consisting in the provision of trans-

port by waterway must automatically be classi-

ed as ‘transport’ or ‘urban transport’ within the

meaning of [Directive 2006/123]” as

“A service of that type could include, besides

transport, one or more other elements that fall

within a commercial sphere that the EU leg-

28 Id. at § 28. Note, however, that in its judgment, the CJEU

concluded that the service in question (the provision of roadworthiness

tests) was a transport service. The CJEU observed in that the purpose

of Directive 2009/40 on roadworthiness tests (as amended by Direc-

tive 2014/45) sought to guarantee road safety and “Directives 2009/40

and 2014/45 were adopted on the basis of Article 71 EC and Article

91 TFEU respectively, both of those provisions being included respec-

tively in the EC Treaty and the TFEU under the Title ‘Transport’ and

constituting the legal basis expressly authorizing the EU legislature to

lay down ‘measures to improve transport safety.’”

29 Opinion of Advocate General Spuznar of 16 July 2015,

Joined Cases C-340/14 and C-341/14, R.L. Trijber and J. Harmsen,

2015 [E.C.R.] I-0000.

30 Id. at § 30.

31 Id. at § 37.

islature has included in the scope of Directive

2006/123. In those circumstances, it is necessary

to consider what the main purpose of the ser-

vice at issue is.”

Seen under that light, I submit that the “main

purpose” of uberPOP is not to provide a “trans-

port service”, but an intermediation service

connecting partner drivers with riders. First, un-

like regular taxi services, Uber is not in the busi-

ness of physically conveying people from point

A to point B. Thus, while many people may con-

sider that Uber provides transport services not

unlike those offered by taxi companies in that

“Uber cars” or “Uber taxis” (as they are some-

times referred to), there is no such thing as an

Uber car let alone an Uber taxi. Thus, the fact

that “taxis” are mentioned at recital 21 of Direc-

tive 2006/123 does not exclude Uber’s services

from the scope of that Directive. The purpose of

Uber’s service is not to “transport” passengers.

Uber does not own cars and its “partner drivers”

are independent contractors using their own

vehicles. They connect to the Uber platform

whenever they want for as long (or as short) a

period of time as they want.

In fact, Uber’s services share many features

with well-known online intermediation platforms,

such as Airbnb or Booking.com. For instance,

Uber facilitates the payment transaction of the

service between the rider and the partner-driv-

er like online booking websites typically provide

an electronic means of payment to the people

booking a room through their website. Another

feature that Uber shares with online platforms

is the ability for users to rate the quality of the

service provided. Online ratings are indeed an

essential element to ensure quality of service

and ensure trust amongst users.

Finally, like oth-

32 Judgment of the Court of 1 October 2015, Joined Cases

C-340/14 and C-341/14, supra note 29, at § 51.

33 James Cook, “Uber’s internal charts show how its driv-

er-rating system actually works”, Business Insider, 11 February 2015,

available at uk.businessinsider.com/leaked-charts-show-how-ubers-

driver-rating-system-works-2015-2

127

er online platforms Uber generates revenues by

charging a commission to the service provider,

which it connects to users.

In fact, the distinct advantage of online inter-

mediation platforms is their ability to intermedi-

ate an extremely large number of transactions

between users and providers with limited staff

and physical assets through the use of software

solutions. This is also the strength of Uber, which

is able to intermediate millions of daily transac-

tions with limited physical and human assets,

precisely for the reason that it does not offer a

transport service by operating cars and employ-

ing drivers. Should that be the case, Uber would

be one of the world’s largest employers and

would own for billions of dollars of assets.

In light of the above considerations, I have

no doubt that uberPOP is an information society

and is not a transport service. Based on the in-

formation in my possession, it is the answer that

I would give to question 1 raised by the refer-

ring court. Now, if the CJEU was to disagree and

consider that uberPOP is a transport service, this

would not leave this service without protection

against regulatory interferences, but this protec-

tion would have to be based on the freedom of

services (Article 56 TFEU). We explore this aspect

below.

B. Regulatory implications of dening

uberPOP as “information society service”

and not a “transport service”

If, for the reasons discussed above, uberPOP

is not a “transport service” within the scope of

Title VI TFEU, but an “information society ser-

vice” within the meaning of Directive 98/34, this

means that uberPOP falls within the scope of

both Directives 2000/31 and 2006/123, as well

as Article 56 TFEU. In fact, both directives are

based on the principles developed by the CJEU

in its interpretation of Article 56 TFEU, and they

largely complement each other.

The referring court unbundles this scenario into

two separate questions, with question 3 taking

as a starting point the situation where uberPOP

is not a “transport service” and therefore falls

under the scope of Directive 2006/123, where-

as question 4 takes as a starting point the situa-

tion where uberPOP is an “information society”

service and therefore falls under the scope of

Directive 2000/31. I suppose that the reason for

this approach is that it is not because a service

is not a “transport service” falling under Title VI

TFEU that it is necessarily an “information society

service” within the meaning of Directive 98/34.

For instance, a service that would intermedi-

ate drivers with passengers without charging a

commission would neither be a “transport ser-

vice” nor an “information society service”.

Question 2 also raises a scenario were uber-

POP would be “in part” an “information society

service” and asks whether in that case uberPOP

can benet from the protection against regula-

tory interferences contained Article 56 TFEU, as

well as Directives 2006/123 and 2000/31.

This Section is divided in four subsections. Sub-

section 1 discusses the scenario where uberPOP

would be “in part” an “information society ser-

vice”. Then, subsections 2 and 3 respectively dis-

cuss the regulatory implications of the (non-mu-

tually exclusive) scenarios where uberPOP is not

a “transport” service and is an “information so-

ciety” service. Finally, subsection 4 emphasises

the point that even if Uber’s services fall under

the scope of Directive 2006/123 and/or Direc-

tive 2000/31, and/or Articles 49 and 56 TFEU it

does not mean that they cannot be regulated

by the Member States, but that regulation is sub-

ject to constraints.

1. uberPOP is “in part” an “information society

34 See Directive 2006/123, Recital 1; Directive 2000/31, Re-

cital 1.

128

service”

As mentioned above, the difculty here is

that the referring court does not clearly de-

scribe the scenario it has in mind. One interpre-

tation is that the court considers that uberPOP is

a “transport service” within the scope of Title VI

TFEU, but also an “information society service”

within the meaning of Directive 98/34. Other-

wise, it would not be easy to understand why

uberPOP would only be “in part” an information

society service. Another interpretation would be

that the court considers that uberPOP is com-

posed of two distinct services: (i) a transport

service performed by a “partner driver” (who is

driving his own car) and (ii) an intermediation

service based on Uber’s software application

and platform, which is “detachable” from the

transport service provided by the driver. In my

view, the latter interpretation is preferable be-

cause otherwise it would be hard to understand

why the court’s question would refer inter alia

to the applicability Article 56 TFEU and Directive

2006/123. If uberPOP was a transport service, this

would automatically rule out the application of

Article 56 TFEU and Directive 2006/123.

The legal consequences of the latter inter-

pretation would be that the intermediation

service provided by Uber would benet from

the principle of freedom to provide services as

guaranteed in Article 56 TFEU, as well as in Di-

rective 2006/123 and 2000/31. What this means

in practice will be discussed in subsections 2

and 3 below. This latter interpretation might still

create challenges for Uber, however, as its part-

ner-drivers, who physically convey people from

A to B and therefore provide a “transport ser-

vice” within the meaning of Title VI TFEU, might

be subject to the licensing obligations required

under Spanish law. While that may not impede

Uber’s uberX service, which relies on profession-

ally licensed drivers, it might create obstacles

to uberPOP, which relies on drivers who are not

professionally licensed.

2. uberPOP is not a “transport” service

In this sub-Section, I analyse the scenario

where uberPop is neither a transport service nor

an information society service. In that scenar-

io, uberPOP falls within the scope of Directive

2006/123 and can benet from the free move-

ment provisions it contains.

As we have seen, the claim made by Asocia-

ción Profesional Elite Taxi in the Barcelona court

is that Uber breaches Article 15 of the Spanish

Unfair Competition Act by operating its service

without the necessary authorization, this raises

the question – as asked by the referring court

– of whether the abovementioned provision is

compatible with Article 9 of Directive 2006/123.

As we have seen, Article 9 provides that Mem-

ber States shall not make access to a service

activity or the exercise thereof subject to an

authorization scheme unless : (i) the authoriza-

tion scheme does not discriminate against the

provider in question; (ii) the need for an au-

thorization scheme is justied by an overriding

reason relating to the public interest ; and (iii)

the objective pursued cannot be attained

by means of a less restrictive measure.

Article 15 of the Unfair Competition Act,

which Uber has allegedly breached, consid-

ers as “unfair”, and thus objectionable, the

non-compliance with a provision regulating

competitive activity.

Article 15 does not con-

tain any “substantive” regulatory requirements.

Thus, it seems to me that the question that is ef-

fectively at stake is whether the fact of imposing

the licensing requirements required for taxis or

so-called private hire vehicles (“PHV”) in Spain

on uberPOP is in breach with the requirements

contained in Article 9 of Directive 2006/123. In

my view, this question raises two distinct issues.

35 Ley 3/1991 de Competitia Desleal (Article 15): “Tendrá

también la consideración de desleal la simple infracción de normas jurí-

dicas que tengan por objeto la regulación de la actividad concurrencia”.

129

The rst issue is whether the Spanish authorities

can subject Uber’s activities to licensing require-

ments without breaching Article 9 of Directive

2006/123. This issue triggers two observations.

First, as in many other sectors, licensing require-

ments may be needed to ensure compliance

with important issues of public interest, such as

public safety and protection of the consumer.

When justied by genuine public interest objec-

tives, licensing requirements cannot in them-

selves breach Article 9 of Directive 2006/123

provided that these requirements are not dis-

criminatory and the objective sought cannot be

achieved in a less restrictive manner.

Second,

it is subject to question whether the licensing re-

quirements should be imposed on the platform

(Uber) or on the drivers operating through the

platform. The answer may depend on whether

it is by licensing the platform or the driver that

the objectives sought by the regulator are best

protected.

The second issue is whether the licensing re-

quirements required for taxis or PHVs meet the

above test. This is a factual issue that I am not

well placed to address, but I will nevertheless

make the following observations.

First, while the licensing requirements imposed

on taxis or PHVs often seek to address market

failures, they may also contain measures restrict-

ing competition. This is particularly so when, as I

understand is the case in Spain, the number of

licenses is strictly limited. Historically, these limita-

tions were justied by the need to prevent “ru-

inous” competition between taxis in situations

where there were too many cars chasing too

few passengers.

While one can seriously ques-

tion whether it is the role of the State to regulate

36 In this respect, it is interesting to note that Spanish law itself

seems to provide for a particularly light-touch approach to intermedia-

tion services. The prior authorization regime for intermediation of pas-

senger transport services has been abolished in Spain by Act 9/2003.

37 See Paul Stephen Dempsey, “Taxi Industry Regulation, De-

regulation & Reregulation: The Paradox of Market Failure”, (1996) 24

Transportation Law Journal 73, at 77.

supply and demand, the quotas sometimes im-

posed decades ago have not kept up with de-

mand and many cities are currently undersup-

plied, particularly at certain times of the week/

day.

In most cases, it is hard to see how they

can be objectively justied.

Second, while licensing requirements repre-

sent a burden for the taxi industry, it now seems

that the main objective of taxi companies is to

maintain these requirements and use them as

barriers to entry. As I have written elsewhere,

while licensing requirements may be needed to

correct market failures, regulation should not be

used to prevent Uber and other intermediation

platforms to offer their innovative services to the

benet of consumers. As will be seen below, the

CJEU has made it clear that the protection of

incumbent operators cannot be seen as a legit-

imate public interest objective.

In sum, licensing requirements restricting the

number of authorizations or imposing require-

ments that cannot be justied by market failures

should be inherently suspect under Article 9 of

Directive 2006/123 or, more generally, Article 56

TFEU.

For the sake of exhaustiveness, I note that if

the CJEU was to decide that uberPOP is a trans-

port service within the scope of Title VI TFEU, it

would not mean that national authorities would

be entirely free to regulate the service as they

wish. As the CJEU made clear in the Yellow Cab

case,

national legislation requiring that an au-

thorization be obtained to operate a transport

38 “Pricing the surge”, The Economist, 29 March 2014,

available at www.economist.com/news/nance-and-econom-

ics/21599766-microeconomics-ubers-attempt-revolutionise-taxi-mar-

kets-pricing-surge

39 BenjaminG.EdelmanandDamienGeradin,“Efciencies

and Regulatory Shortcuts: How Should We Regulate Companies like

Airbnb and Uber?”, forthcoming Stanford Technology Law Review

(2016), current draft available at http://papers.ssrn.com/sol3/papers.

cfm?abstract_id=2658603

40 Case C-338/09, Yellow Cab Verkehrsbetriebs v. Lande-

shauptmann von Wien, [2010] E.C.R. I- 13927.

130

service (in that case a transport bus service):

“constitutes, in principle, a restriction of free-

dom of establishment within the meaning of Ar-

ticle 49 TFEU, in that it seeks to restrict the num-

ber of services providers, notwithstanding the

alleged absence of discrimination on grounds

of nationality of the persons concerned. … Con-

sequently, it is necessary to examine whether

the legislation at issue … may be justied objec-

tively.”

Pursuant to the case-law of the CJEU, a leg-

islation “applicable without discrimination on

grounds of nationality, may be justied by over-

riding reasons of general interest, provided that

it is appropriate for ensuring attainment of the

objective pursued and does not go beyond

what is necessary for attaining that objective”.

Thus, licensing requirements needed to ensure

the protection of objectives such car and pas-

senger safety may be justied. The CJEU, how-

ever, observed in its Yellow Cab judgment that:

“[b]y contrast, the objective of ensuring the

protability of a competing bus service, as a

reason of a purely economic nature, cannot, in

accordance with settled case-law, constitute

an overriding reason in the public interest jus-

tifying a restriction of a fundamental freedom

guaranteed by the Treaty.”

Thus, licensing requirements that cannot be

justied by the need to correct a market failure,

but on the contrary seek to protect the eco-

nomic interests of incumbent operators are not

objectively justied.

3. uberPOP is an “information society service”

For the reasons discussed in Section A above,

41 Id. at §§ 45-46.

42 Case C-169/07, Hartlauer, [2009] ECR I-1721, § 44.

43 Id. at § 51.

there is no doubt in my view that uberPOP

should be considered as an “information so-

ciety service” within the meaning of Directive

98/34. Hence, Directive 2000/31 applies. In that

context, question 4 asked by the referring court

is whether restrictions in one Member State re-

garding the freedom to provide electronic inter-

mediation services from another Member States

either in the form of making the service subject

to a licence or in the form of an injunction based

on the legislation on unfair competition, are val-

id derogations from Article 3(2) in accordance

with Article 3(4) of Directive 2000/31. Thus, my

observations in subsection 2 on the type of mea-

sures than can be objectively justied, largely

apply to this subsection as well.

Although, as already observed above, the

drafting of that question is confusing as Uber

Systems Spain does not provide the uberPOP in-

termediation service, there is no question that

the Spanish legislation imposes restrictions on

the provision of Uber BV, a company located

in the Netherlands. As to the question’s refer-

ence to an injunction based on the legislation

on unfair competition, it should be understood

as an allusion to the preliminary injunction ren-

dered by the Commercial Court No 2 of Madrid

ordering on (i) Uber Technologies to cease its

operations in Spain and on (ii) telecommunica-

tions services providers and payment operators

to block the trafc and transactions with

www.

uber.com

and with Uber’s software applica-

tion.

The claim was once again that Uber had

breached the Spanish Unfair Competition Act,

and vice-versa.

Article 3(2) of Directive 2000/31 contains the

general principle that: “Member States may not,

for reasons falling within the coordinated eld,

restrict the freedom to provide information soci-

ety services from another Member State.” How-

ever, Article 3(4) provides that Member States

may nevertheless adopt measures derogating

44 See supra note 16.

131

from the freedom to provide information society

services if such measures are: (i) necessary for

reasons of public policy, the protection of pub-

lic health, public security, the protection of con-

sumers; (ii) taken against a given information

society service which prejudices the objectives

referred to in point (i) or which presents a serious

and grave risk of prejudice to those objectives;

(iii) proportionate to those objectives.

Given the conceptual proximity between

Article 9 of Directive 2006/123 and Article 3(4)

of Directive 2000/31, the analysis performed in

section 2 above is relevant as well here. The no-

tion of public interest contained in Article 9 of Di-

rective 2006/123 certainly covers the reasons of

public policy, protection of public health, public

security and the protection of consumers men-

tioned above. The principle of “proportionality”

is also conceptually close to the “least restric-

tive means requirement” contained in Article 9

of Directive 2006/123.

Thus, it is unlikely that a

measure that would fail under the Article 9 test

would succeed under Article 3(4) of Directive

2000/31.

There is little doubt that license requirements

restrict the freedom to provide information so-

ciety services across Member States. However,

as noted above, there may also circumstances

where regulation may be needed to achieve

public interest objectives, such as public secu-

rity and protection of consumers. It could, for

instance, be argued that by connecting pro-

spective passengers with unsafe drivers (or driv-

ers using unsafe cars), Uber could harm its users

(passengers) and non-users (pedestrian). Thus,

for instance, a measure requiring Uber to com-

ply with safety measures going beyond those it

voluntarily performs may not in itself be seen as

a breach of Article 3(4) of Directive 2000/31 pro-

vided, of course, that the measure adopted is

45 See Opinion of Mr Advocate General Van Gerven delivered

on 5 February 1991, Case C-347/89, Freistaat Bayern v Eurim-Pharm

GmbH., [1991] E.C.R. I-01747, at §§ 9 et seq.

proportionate to the objective sought. By con-

trast, imposing licensing requirements that have

the object or effect of creating barriers to entry

cannot be compatible with that provision.

As to the issuance by the Commercial Court

No 2 of Madrid of a preliminary injunction not

only ordering Uber to cease its activities in Spain,

but also telecommunications services providers

and payment operators to no longer assist Uber

in its transactions, it appears as an overkill. That

is particularly the case considering that the im-

plementation of this measure by Spanish mobile

operators not only prevent smartphone users

to use Uber’s services in Spain, but also in other

Member States. It is not clear why the extension

is necessary to protect a legitimate objective.

4. Regulating Uber in a proportionate manner

One concern that some may have with a

nding that Uber’s activities fall under the scope

of Directive 2006/123 and/or Directive 2000/31

is that these activities may end up being un-

regulated, or at least under-regulated, with the

risks that it leaves its users and non-users without

sufcient protection and give it an unfair com-

petitive advantage compared to taxi and other

regulated forms of services.

However, as not-

ed above, that does not need to be the case.

Neither Directive 2006/123 nor Directive 2000/31

prevent regulating Uber’s activities provided

that the regulatory requirements comply with a

series of principles designed to avoid unjustied

interference with Uber’s freedom to provide its

services.

For instance, the State of California has ad-

opted a regulatory regime specically designed

for “companies that provide prearranged trans-

portation services for compensation using an

online-enabled application (app) or platform to

connect passengers with drivers using their per-

46 Id.

132

sonal vehicles”.

This regime, which was adopt-

ed by the California Public Utilities Commission

(CPUC), establishes 28 rules and regulations for

so-called Transportation Network Companies

(“TNC”), such as Uber, whereby they must ob-

tain a license from the CPUC to operate in Cali-

fornia; require each driver to undergo a criminal

background check; establish a driver training

programme; implement a zero-tolerance policy

on drugs and alcohol; hold a commercial liabil-

ity insurance policy that is more stringent than

the CPUC’s current requirement for limousines,

requiring a minimum of $1 million per-incident

coverage for incidents involving TNC vehicles

and drivers in transit to or during a TNC trip, re-

gardless of whether personal insurance allows

for coverage; and conduct a 19-point car in-

spection.

Closer to home, Estonia adopted a new Pub-

lic Transport Act (“PTA”) in February 2015, which

comprises a regulatory regime applying to

occasional service providers, such as the driv-

ers that typically operate under the uberPOP

banner.

The PTA denes “occasional services”

as “the carriage by road, except for regular

services and taxi services, and the main char-

acteristic of which is the carriage of groups of

passengers constituted on the initiative of the

customer or the carrier.”

Providers of occasion-

al services are subject to a mandatory license

comprising requirements, such as obligations

to register as commercial or non-commercial

entity, to show the absence of criminal convic-

tions, the need to appoint a transport manager,

etc.

Moreover, Estonia is currently considering

47 See CPUC Establishes Rules for Transportation Network

Companies, Press Release, 19 September 2013, available at http://docs.

cpuc.ca.gov/PublishedDocs/Published/G000/M077/K132/77132276.

PDF

48 Id.

49 See https://www.riigiteataja.ee/en/eli/526032015005/con-

solide

50 Id. at § 5.

51 Id. at §§ 41 et seq.

further amending the TPA to allow ride-sharing

services.

The draft legislation would require

intermediation platforms such as Uber to meet

certain transparency and safety standards, for

instance by requiring transparency regarding

the fares are calculated, providing riders with

electronic receipts; and displaying a driver’s

photo and license plate number before the

passenger enters a vehicle. The draft legisla-

tion would not, however, subject Uber or other

equivalent service providers to an authorization

or licensing regime.

While it is early to tell how the Estonian re-

gime will apply in practice, the approach of the

CPUC seems a step in the right direction in that it

ensures that Uber’s activities are regulated, with

requirements that are however adapted to its

intermediation business model.

V. Looking beyond Uber

Uber is probably the best known and most

controversial online intermediation platform

in the world. The reasons why Uber has raised

so much attention is that, on the one hand, it

has attracted a large number of acionados

amongst users who enjoy the cheap and con-

venient service it offers, but, on the other hand,

it has drawn the hatred of taxi companies and

drivers as they see Uber’s services as an existen-

tial threat to their business. This situation is not

unlike what we have seen throughout the 1980s

and 1990s where a large number of sectors of

the economy (air transport, telecommunica-

tions, etc.) were liberalized, with incumbents

complaining that new entrants did not have to

comply with some of the heavy requirements

(such as, for instance, universal service obliga-

tions) that applied to them.

But Uber is only a pioneer in a segment of the

52 See Avi Pau, “Estonia to regulate Uber and Taxify ride-shar-

ers”, 15 February 2016, available at news.postimees.ee/3583509/esto-

nia-to-regulate-uber-and-taxify-ride-sharers

133

economy that is likely to grow as various activi-

ties historically carried out by humans (such as,

for instance, taxi dispatchers or booking agents)

can now be performed by software applica-

tions.

Already Airbnb has become a powerful

player in the hospitality industry drawing criti-

cism from hotels it is largely free from the regu-

latory burden they are subject to.

Hence, like

Uber is criticized for engaging in unfair compe-

tition and lawsuits abound.

Similar claims will

be made every time a popular platform creates

losers, although they generate signicant bene-

ts for both providers and users.

The answers that will be given by the CJEU to

the questions of the referring court are thus likely

to have an impact beyond the dispute oppos-

ing Uber and taxis. The broad question is how to

reach the dual objective of ensuring that online

intermediation platforms are allowed to provide

their (usually efcient and attractive) services,

while ensuring that they comply with the regu-

latory requirements needed to correct clearly

identied market failures.

While Uber’s services

have been subject to challenges in many coun-

tries inside and outside the European Union,

I strongly believe that the right approach for

regulatory authorities is to adopt regulatory re-

gimes that achieve the dual objective identied

above. This of course requires independence (as

there are strong vested interests) and creativity,

although the principles contained in Directives

53 Marc Andreessen, “Why Software Is Eating The World”,

The Wall Street Journal, 20 August 2011, available at www.wsj.com/

articles/SB10001424053111903480904576512250915629460

54 Asha Barbaschow, “Airbnb raises $100m and maintains

$25.5b valuation”, ZDNet, 23 November 2015, available at www.

zdnet.com/article/airbnb-raises-100m-and-maintains-25-5b-valua-

tion/?ftag=YHRe31c277

55 See, e.g., Jason Abbruzze and Jessica Plautz, “New

York Goes to War Against Airbnb for Disrupting Hotel Business”,

Mashable, 26 April 2014, available at mashable.com/2014/04/26/

new-york-vs-airbnb/#lSXkJBuoIEq1

56 Damien Geradin, “Uber and the Rule of Law: Should Spon-

taneous Liberalization be Applauded or Criticized”, forthcoming Com-

petition Policy International (2016), draft available at http://papers.ssrn.

com/sol3/papers.cfm?abstract_id=2693683

2006/123 and 2000/31, as well as the case-law

of the CJEU on articles 49 and 56 TFEU may offer

useful guidance.

VI. Conclusion

The answers that will be given by the CJEU to

the questions asked by the Commercial Court

No 3 of Barcelona regarding the nature of uber-

POP, as well as the compatibility of Spanish leg-

islation restricting these activities with Directives

2006/123 and 2000/31, and the free movement

rules contained in the TFEU, will have a funda-

mental impact on the way Uber, but also oth-

er online intermediation platforms will be regu-

lated in the future. The judgment of the CJEU

could result in forcing Member States to rethink

the regulatory frameworks that apply to taxis

and/or VTCs in a manner that truly serves com-

petition, innovation and user choice.

In my view, uberPOP is not a “transport ser-

vice”, but an “information society service”.

Therefore, it falls within the scope of Directives

2006/123 and 2000/31, as well as under Article

56 TFEU. This does not mean that Uber’s activi-

ties cannot be regulated, but regulation should

be limited to protecting objectives of public in-

terest and they should be proportionate to the

accomplishment of these objectives. In other

words, regulatory requirements that may apply

to Uber’s activities and other similar companies

should be limited to what is necessary to correct

market failures. Measures that have the object

or effect to restrict entry should be struck down

unless they can be objectively justied.

134

Competition Policy

in Consumer

Financial Services:

The Disparate

Regulation of

Online Marketplace

Lenders and Banks

By Thomas P. Brown and Molly E. Swartz

I. INTRODUCTION

In October 2014, Washington D.C. City Coun-

cil passed legislation that effectively allowed

Uber to operate in the District. David Plouffe,

formerly an advisor to the President and now an

executive with Uber, greeted the new legisla-

tion with the following observation:

Obviously what we’re doing doesn’t neces-

sarily in all cases t in existing regulation. I think

that’s what Washington really wrestled with and

decided they needed to chart a new pathway

forward. So rather than say how do we t this

new technology and service into existing regu-

lations, let’s look at how do we create new regu-

lations that give citizens of the city the right kind

of condence on things like safety, on things like

insurance.

Thomas P. Brown is a partner and Molly E. Swartz is an

associate at Paul Hastings LLP. The views expressed are entirely their

own.Theirviewsdonotrepresenttheviewofthermoranyclientof

therm.

1 Dana Rubenstein, “Uber Publicly Embraces Regulation, the

‘Modern’ Sort,” Politico New York, October 29, 2014,

Uber is just one of many startups struggling to

t their businesses into existing regulatory frame-

works. As technological innovation leads to new

business models, there is increasing friction be-

tween these new companies and the existing

regulatory regime.

The tension between regulated entities and

new entrants is particularly acute in the con-

text of online marketplace lending.

While bank

lenders enjoy regulatory privileges that enable

them to lend immediately to consumers in all

50 states, non-bank lenders are forced to en-

gage in resource-intensive analyses to satisfy

state-specic compliance requirements. As

non-bank lenders expand access to credit to

those currently under served by banks—provid-

ing new underwriting methodologies, real-time

data transmission and new nancing mecha-

nisms—disparate regulation of banks and non-

bank lenders appears problematic.

In the past, where new entrants have chal-

lenged existing regulatory frameworks, restruc-

turing has occurred to ensure a functioning

market. This continues to happen in a number

of industries, with the Uber-led transformation of

taxi regulation being the most prominent. This

kind of regulatory reorganization is also need-

ed in the lending space. The existing framework

for regulating the delivery of nancial services

works against the interests of consumers, com-

petition, regulators and society as a whole. A

state-by-state legal regime serves as barrier to

entry protecting incumbent banks from compe-

tition and depriving consumers of alternatives.

http://www.capitalnewyork.com/article/city-hall/2014/10/8555647/

uber-publicly-embraces-regulation-modern-sort.

2 For purposes of this paper, we use the phrase “marketplace

lenders” to distinguish between bank lenders and technology driven

lenders. In our experience, however, this taxonomy is a bit narrow in

that technology driven lenders follow two variations—(1) those that in-

volve discrete sources of third-party capital and are generally described

as “marketplace lenders” such as Prosper, Lending Club, and OnDeck;

and (2) those that lend off their own balance sheet such as PayPal Credit

but that use an essentially identical origination model and are generally

described as “platform lenders.”

135

There is simply no reason why banks should en-

joy access to the common market while non-

bank lenders cannot.

This is not to say that banks and non-banks

should be treated similarly on all counts. There

are numerous situations in which it is appropriate

for banks to maintain regulatory privileges inac-

cessible to non- banks. In fact, in the context of

nancial services, banks tend to bear a greater

regulatory burden than non- banks (e.g., appli-

cation of customer identication program re-

quirements, required maintenance of leverage

ratios, etc.). In the lending context, however,

banks’ unique ability to offer products on a na-

tionwide basis remains largely unjustied.

In Part I below, we provide an overview of on-

line marketplace lending. We suggest that mar-

ketplace lenders offer value that is not currently

replicable by banks. Part II examines market-

place lending across state lines, recognizing the

near impossibility of full compliance. Part III pro-

vides examples of cases in which new entrants

have successfully challenged existing regulatory

frameworks. In these cases, regulatory change

reinvigorated competition to the benet of con-

sumers. Finally, in Part IV, we suggest the need

for reorganization of the existing lending regu-

latory framework. The current bifurcated regu-

latory framework increases costs to consumers,

limits consumer choice and insulates banks from

competition.

II. Online Marketplace Lending Ben-

ets Both Underserved Borrowers and

Investors

In the past few years, marketplace lending

has emerged as an alternative to traditional

bank lending. In the wake of the 2008-2009 -

nancial crisis, banks tightened credit guidelines.

This left many consumers and small businesses

without access to bank-issued credit. Total con-

sumer lending fell by 6.1 percent between Jan-

uary 2009 and March 2010.

At the same time

that they tightened credit standards, banks

found themselves a safe haven for deposits

even as yields on those deposits plummeted.

The simultaneous tightening of credit stan-

dards and drop in yields created an opportu-

nity for new credit intermediaries to emerge.

Marketplace lenders lled this gap. In their initial

incarnation, rms such as Prosper and Lending

Club enabled lenders to fund loans to borrow-

ers. They and other alternative lenders simulta-

neously expanded the pool of available credit

and enabled yield-starved investors to obtain a

positive rate of return on funds that would have

generated no return had they been left on de-

posit at banks and other depository institutions.

Marketplace lenders differ from traditional

nancial institutions in a number of ways. First,

marketplace lenders often serve demograph-

ics that are underserved by bank lenders. Mar-

ketplace lenders have enabled “thin le” bor-

rowers and small business borrowers to access

credit that traditional nancial institutions were

unwilling to extend. Borrowers rendered ineli-

gible by traditional bank underwriting models

may nd investors on online marketplaces will-

ing to nance their credit needs. Alternative un-

derwriting models may enable such lenders to

extend credit to thin le borrowers who would

not qualify for credit based solely on traditional

underwriting criteria such as FICO score.

Second, marketplace lenders rely on tech-

nology to reduce the cost of connecting bor-

rowers with lenders. They use algorithms, rather

than lending ofcers, to screen borrowers, and

they provide granular information about repay-

ment risk to investors. Further, many such plat-

forms have eliminated unnecessary or unwant-

ed services associated with traditional lenders,

3 “Epic Consumer Credit Crunch Unfolding,” Seeking Alpha,

March 2, 2010, http://seekingalpha.com/article/191517-epic- consum-

er-credit-crunch-unfolding.

136

such as branches and other physical locations.

Through better underwriting and more efcient

operations, marketplace lenders and other

lending platforms have lowered the cost of ob-

taining loans and are able to offer borrowers

credit on better terms.

Third, platform lenders offer value to investors.

Marketplace lenders have enabled investors to

diversify their investment portfolios by investing

directly in individual loans. Even to the extent

that investors choose to fund pools of loans rath-

er than individual loans, marketplaces may be

able to pass a larger portion of the interest that

those loans generate to the investors that fund

their loans.

III. Marketplace Lending Across

State Lines Triggers Signicant Com-

pliance Obligations

In lending across state lines, marketplace

lenders, like other non-bank lenders and, indeed,

all non- bank providers of nancial services,

confront a complex, unstable and fragmented

regulatory regime. The regulatory thicket that

surrounds the nancial services industry in the

United States, particularly the lending business,

is Byzantine. A rm that is considering launching

a product that provides liquidity to customers

must grapple with a long list of Federal laws and

regulations, including the Truth-in-Lending Act,

the Fair Credit Reporting Act,

the Electronic

Funds Transfer Act,

the Equal Credit Opportu-

nity Act,

Regulation Z,

and Regulation E

(to

4 Andrew Verstein, “The Misregulation of Person-to-Person

Lending,” U.C. Davis Law Review 45, (2011): 445, 457.

5 Id.

6 15 U.S.C. §§ 1601 et seq.

7 15 U.S.C. §§ 1681 et seq.

8 15 U.S.C. §§ 1693 et seq.

9 15 U.S.C. §§ 1691 et seq.

10 12 C.F.R. §§ 1026 et seq.

name but a few). Individual states have their

own laws. California, for example, regulates

non-bank lenders through the California Consti-

tution,

the Finance Lenders Law

and, in some

instances, the Consumer Legal Remedies Act.

How and whether any one of these laws or

regulations applies turns on a number of factors,

including the following: (1) whether the service

is provided for household use; (2) whether the

service provider is a bank (or other federal in-

sured deposit taking institution); (3) whether the

service creates a debt enforceable against

the customer; (4) whether the service involves

a nance charge on a loan or a “time-price”

charge associated with a sale; (5) whether the

service is associated with a prepaid account

but not a deposit account; and (6) whether the

information on which the decision to provide li-

quidity is collected from the customer directly,

third parties that have a direct relationship with

the customer, or third parties that collect infor-

mation from others about the customer.

This body of law and regulation is also unsta-

ble. Regulators, courts, and, of course, legisla-

tures change the rules from time-to-time, and

these changes can have signicant repercus-

sions for industry participants. The Second Cir-

cuit’s recent decision in Madden v. Midland

Funding, LLC,

provides one timely example.

Madden arose from a dispute between a con-

sumer and purchaser of debt owed by the

consumer to the bank that had issued the con-

sumer a credit card.

The consumer sued the

debt collector in New York state court alleging

that the fees charged by the debt collector ex-

11 Cal. Const. art. XV, § 1.

12 Cal. Fin. Code §§ 22000 et seq.

13 Cal. Civ. Code §§ 1750 et seq.

14 Madden v. Midland Funding, LLC, 786 F.3d 246 (2d Cir.

2015).

15 Id. at 247-49.

137

ceeded the cap set by New York usury law.

The Second Circuit held that federal preemp-

tion was not available to the debt collector in

collecting the debt pursuant to the terms of the

loan agreement because the debt collector

was acting on behalf of itself rather than the

bank.

The court deected criticism that its de-

cision would undermine the sale of charged-off

debt by banks by arguing that it “would not

signicantly interfere with any national bank’s

ability to exercise its powers under the [National

Bank Act].”

Among other things, Madden illustrates that

the regulatory burdens and benets are not

evenly distributed in the lending space. On its

face, the Second Circuit’s decision creates a

special privilege for banks relative to non-banks.

A bank purchaser of another bank’s debt can,

under the Second Circuit’s analysis, invoke its

ability to preempt state law to block a consum-

er’s challenge to the fees collected by the sec-

ond bank based on the loan originated by the

rst. Most non-banks not exercising the powers

of a national bank, according to the Second

Circuit, have no such right.

Both of the pub-

licly traded platform lender, Lending Club and

OnDeck, saw their valuations decline relative to

traditional lenders in the wake of the decision,

and commentators have attributed the relative

severity of the decline to regulatory risk.

16 Id.

17 Id. at 245-53

18 Id. at 249.

19 Id. at 251 (stating that “[i]n most cases in which NBA pre-

emption has been applied to a non-national bank entity, the entity has

exercised the powers of a national bank—i.e., has acted on behalf of a

national bank in carrying out the national bank’s business. This is not

the case here.”). See also Marquette Nat. Bank of Minneapolis v. First

of Omaha Service Corp., 439 U.S. 299, 308, 313- 316 (1978) (holding

that a federally chartered bank may offer loans to consumers in any of

the other 49 states at any interest rate allowed by the bank’s state of

residence regardless of whether the consumer’s home state recognizes

a lower usury cap); Smiley v. Citibank (South Dakota), N.A., 517 U.S.

735 (1996) (holding that the NBA preempts state limits on fees as well

asnancecharges).

20 Leena Rao, “Once-hot Online Lending Companies Go Cold

The existing regulatory framework for regu-

lation of non-bank lenders is a patchwork of

complicated and overlapping state laws and

regulations. Each state sets a different maxi-

mum interest rate that parties may contract for,

and this rate may vary depending on whether

the credit will be used for personal, household

or family purposes (i.e., consumer credit) or for

business purposes (i.e., commercial credit). In

many states, consumer and/or commercial

lenders may be authorized to charge interest

above a state’s usury cap if they obtain a state

lender license—a time-consuming and expen-

sive process. For example, a marketplace lend-

er may contract with a Utah-based borrower for

any rate of interest without a license.

In Virgin-

ia, a lender must obtain a lender license to offer

consumer loans to Virginia residents at interest

above 12 percent per annum.

In California, a

license is required to engage in the business of a

nance lender, regardless of what interest rates

are offered.

Lender license applications can also be quite

burdensome and appear designed to deter

applications. The applications often require ap-

plicants to submit background checks and n-

gerprints on all persons owning or controlling 10

percent or more of the lending entity, nancial

statements, and surety bonds. Nevada, for ex-

ample, requires lenders to maintain a physical

ofce in the state—a requirement that is partic-

ularly onerous for online lenders with no physical

location.

IV. To Avoid State Lending Laws,

Marketplace Lenders Are Forced to

Partner with Banks

in Face of Skepticism,” Fortune, June 30, 2015,

http://fortune.com/2015/06/30/lending-club-ondeck-shares.

21 Utah Code Ann. § 15-1-1 (West).

22 Va. Code §§ 6.2-1501, 6.2-303.

23 Cal. Fin. Code § 22100(a).

24 Nev. Rev. Stat. §§ 675.090(2)(a); 675.090(3).

138

To avoid this morass of state lending laws, a

number of marketplace lenders have chosen to

partner with banks. A regulatory regime where

the burden of compliance is so high that com-

panies are forced to partner with competitor

incumbents to provide cost-effective products

seems unequivocally problematic.

Both Prosper and Lending Club were, in their

original incarnations, fairly novel. They enabled

investors to fund loans extended to individuals

without a traditional nancial institution, either

a bank or licensed lender, serving as originator.

Yet although Prosper and Lending Club were

serving as intermediaries between borrowers

and investors, neither used the form that has

dominated the consumer lending business in

the United States since the early part of the

Twentieth Century—i.e., a chartered nancial

institution such as a bank, credit union or thrift.

And it was not at all clear how either compa-

ny thought that it was complying with the raft

of Federal and state laws related to consumer

lending.

But times have changed. In the almost ten

years that have passed since Prosper got its start,

Prosper and Lending Club have almost com-

pletely reinvented their businesses. Today, both

companies rely on banks to originate loans.

Likewise, both companies have jettisoned the

direct investment approach. Under the model

that both companies have now adopted, inves-

tors no longer directly fund loans to borrowers.

Rather, the companies interpose intermediaries

that own the right to the receivables generat-

ed by borrowers, and those intermediaries then

25 Eileen Ambrose, “Peer-to-Peer Lending Alternative Runs

into a Regulatory Wall,” Baltimore Sun, December 7, 2008, http://

articles.baltimoresun.com/2008-12-07/business/0812060058_1_lend-

ing-sites-peer-to-peer-lending-prosper-loans (stating that “Peer-to-peer

lending . . . recently has come into regulators’ sights . . . [sidelining]

the largest peer-to-peer lending site, Prosper.com,” and also quoting

Lending Club CEO’s statement that he “concluded that the industry was

headed toward regulation.”).

pay investors based on the repayment history of

borrowers. Although the platforms offer inves-

tors far more visibility into the performance of

particular loans, the structure of the relationship

between investors in the loans and borrowers is

similar in form to traditional securitization.

Viewed through this lens, the “new” platform

lending businesses look pretty similar to “old”

consumer lending businesses. That is, a non-

bank contracts with a bank to help the bank

acquirer borrowers, underwrite those borrow-

ers, service those borrowers and manage the

resulting portfolio for the benet of third-party

investors. Although some of the details have ob-

viously evolved, the basic components of the

“new” platform- lending model should be famil-

iar to anyone who has followed the credit card

industry since General Motors offered the GM

Rewards card in the 1980s.

In fact, the 1996

Narratives to the Ofce of the Comptroller of the

Currency handbook issued for the supervision of

credit card lending describes the component

parts of a credit card business in terms that mir-

ror the relationship between platform lenders,

their bank origination partners, consumers, and

investors.

Having chosen to partner with a state char-

tered bank for the origination of the loans,

Lending Club and Prosper have subjected

themselves to regulatory supervision in more or

26 Jane Kim, “Peer-to-Peer Lender Relaunched,” Wall

Street Journal, April. 28, 2009, http://www.wsj.com/articles/

SB124088142201761953; Prosper Funding LLC, August 2015 prospec-

tus for up to $500,000,000 in principal amount of Borrower Payment

Dependent Notes at *10, https://www.prosper.com/Downloads/Legal/

Prosper_Prospectus_2015-08-13.pdf ; Lending Club, August 2014 pro-

spectus for Member Payment Dependent Notes at *7, https://www.lend-

ingclub.com/fileDownload.action?file=Clean_As_Filed_20140822.

pdf&type=docs.

27 S. Evans and Richard Schmalensee, Paying with Plastic

(Cambridge: The MIT Press, 2005), 78-79.

28 See OCC Credit Card Handbook 1996 at 12 (“Although

some institutions develop their own scoring models, most are built by

outside vendors.”); id. at 5 (“Issuing banks often employ outside ven-

dors to perform solicitation, servicing, collections, or other functions

….”).

139

less the same way that non-bank technology

providers have been subjecting themselves to

regulatory supervision for decades. The loans

are bank products, and the banks that origi-

nate them are answerable to their regulators

for the nancial performance of those loans as

well as the many regulatory issues that arise in

connection with the issuance of such loans. In

short, Lending Club and Prosper have achieved

regulatory compliance by relying on banks’ pre-

emptive privileges.

V. Varied State and Federal Regu-

lation in the Airline, Telecommunica-

tions and Taxi Industries Demonstrate

the Need for Regulatory Reorganiza-

tion

The fact that Lending Club and Prosper felt

compelled to partner with a bank to reduce

the regulatory burden should be understood

as broad indictment of that regulatory regime.

In other industries where new business models

have challenged existing regulatory frame-

works, the government has been willing to revise

the overarching regulatory framework to ensure

a functioning market. In the airline, telecom-

munications, and taxi industries, for example,

existing regulations unfairly advantaged incum-

bents, thus precluding competition. To ensure

a functioning market, regulatory reorganization

was necessary.

Prior to passage of the Airline Deregulation

Act of 1978

(“ADA”), airlines were heavily reg-

ulated by the Civil Aeronautics Board (“CAB”).

The CAB had jurisdiction to control route entry

and exist of air carriers, regulate fares, award

subsidies, and control mergers and inter-carrier

agreements.

The inexibility of this federal reg-

29 Airline Deregulation Act of 1978, Pub. L. No. 95-504, 92

Stat.1705(codiedasamendedinscatteredsectionsof49U.S.C.).

30 Andrew R. Goetz and Paul S. Dempsey, “Airline Deregula-

tion Ten Years After: Something Foul in the Air,” Journal Air Law and

ulation made it increasingly difcult for carriers

to comply. A number of studies determined that

economic regulation resulted in excessively

high fares and a net economic loss to society

at large.

In an effort to avoid this stringent federal reg-

ulation, some carriers began investing in intra-

state travel—a market that remained outside of

CAB jurisdiction. Carriers operating in the unreg-

ulated intrastate markets were able to offer low-

er fares to consumers and avoid CAB regulation

all together.

As Lewis A. Engman, then chair-

man of the Federal Trade Commission stated,

If you have any doubt that one consequence

of the CAB’s control over rates and routes is

higher prices, you need only look at what hap-

pened some years ago in California when Pa-

cic Southwest Airlines, an intrastate carrier not

subject to CAB regulation or entry restrictions

entered the San Francisco/Los Angeles market

with rates less than half those being charged by

the interstate CAB certied carriers TWA, West-

ern, and United.

Fares were 30 percent less for the unregulat-

ed intrastate airlines in Florida.

Eventually, economists determined that

economic regulation in the airline industry was

distorting the efcient performance of the

marketplace. With leadership from Sena-

tor Edward Kennedy, Congress eventually

passed the ADA. The ADA rescinded CAB’s au-

Commerce 54, (1986): 927, 929.

31 Id. at 930.

32 Id.

33 Christine Chmura, “The Effects of Airline Regulation,” The

Freeman, Foundation for Economic Education, August 1, 1984, http://

fee.org/freeman/the-effects-of-airline-regulation/ quoting Lewis A. En-

gman, “Regulating Industry,” Washington Post, October 15, 1974.

34 Id. quoting Robert Lindsey, “Airline Deregulation is

Weighted,” New York Times, February 7, 1975, p. 39.

140

thority over route entry and exist, airline fares,

and mandated that the CAB be dissolved by

1984. In essence, the government acknowl-

edged that there was a problem: consumers

were poorly served by a system that incentiv-

ized airlines to provide only intrastate travel.

Similarly, the telecommunications faced sig-

nicant organization where state and federal

regulation were set up so as to encourage mo-

nopolistic behavior. Prior to 1969, the telecom-

munications industry was regulated as a lawful

monopoly.

Local telephone service was pro-

vided by an operating company of the AT&T-

owned Bell System or by one of approximately

1,600 independent telephone companies. Long

distance telephone service was provided by

the long Lines Department of AT&T in partner-

ship with the Bell operating companies.

In 1969, however, the Federal Communica-

tions Commission approved an application

submitted by AT&T competitor MCI to construct

and operate a long distance telephone system

between Chicago and St. Louis.

Effective-

ly, however, to provide long distance service,

MCI would need to rely on AT&T-owned inter-

connections and local distribution facilities. Al-

though MCI and AT&T attempted to negotiate

a permanent agreement regarding access to

this infrastructure, negotiations failed. Among

other things, MCI claimed that AT&T was unlaw-

fully denying it interconnections and that it was

being charged excessive and discriminatory

prices for local distribution facilities.

MCI led

suit. Shortly thereafter, the Department of Jus-

tice (“DOJ”) began an investigation.

35 MCI Communications Corp. v. American Tel. &

Tel. Co., 708 F.2d 1081, 1093 (7th Cir. 1983).

36 Id.

37 Id. at 1094.

38 Id. at 1096.

Again, consumers were unable to bene-

t from competition in the market. And again,

the government was forced to step in. After a

protracted lawsuit, AT&T settled with the DOJ.

Among other things, AT&T agreed to divest it-

self of the operating companies that provided

the local exchange service. Challenging AT&T’s

established monopoly, new entrant, MCI effec-

tively transformed the existing regulatory para-

digm, opening telecommunications up to mul-

tiple providers and offering consumers greater

choice.

This trend—new business models threatening

existing regulatory frameworks—continues to-

day. As noted at the beginning of this article,

Uber poses a tremendous threat to the incum-

bent taxi industry. While common carrier regu-

lations are well intentioned, these regulations

were written in a time before geolocation- en-

abled smartphones and ride-sharing appli-

cations. They reect and benet regulatory

concerns associated with taxi service, not peer-

to-peer ride-sharing. Yet as consumers continue

to use Uber’s services and demand regulatory

changes to support Uber’s business, state gov-

ernments have begun to revise state utility laws

to accommodate Uber—despite taxi industry

protests.

In California, for example, Uber was success-

ful in lobbying the California Public Utilities Com-

mission (“CPUC”) to create a new category

of regulated entities (“Transportation Network

Companies”) to cover peer-to-peer ride-sharing

services. Recognizing the value of Uber’s prod-

uct, the CPUC altered its regulatory framework,

thus expanding the market for transportation

services and consumer choice.

VI. Leveling The Playing Field

Between Banks and Non-Banks

In the same way that new entrants have

forced re-examination of the regulatory frame-

141

work for the airline, telecommunications and

now, taxi industries, the effort of Prosper, Lend-

ing Club and countless others to reinvent nan-

cial services should lead regulators to re-evalu-

ate the regulatory framework for that industry.

The fact that Lending Club and Prosper have

effectively joined the club by partnering with

incumbents does not give regulators in this in-

dustry a pass.

Banks have a vested interest in preserving

the regulatory status quo. Banks benet from

the complexity, instability and fragmentation of

regulation in two ways. First, banks are incum-

bent providers of services that others would like

to offer, and as incumbents, the complex and

unstable regulatory regime serves as a barrier to

entry. Second, banks have a unique ability to

export the terms of the loans that they offer from

the states in which are chartered to the states in

which their consumers reside.

There is no policy justication for giving banks

and other chartered nancial institutions a mo-

nopoly on the ability to export contract terms

from one state to another. Although banks are

subject to prudential supervision, there is no dis-

cernable connection between onsite govern-

ment supervision to protect against the systemic

risks that massively leveraged institutions create

for the economy as a whole and banks’ unique

ability to exploit the efciencies associated with

the common market. Exportation of product

terms is not a source or solution to the system-

ic risks created by the enormous leverage that

lurks on bank balance sheets. In short, the risks

that uniquely justify much of the supervision of

banks do not also justify their sole ownership on

exportation. After all, the massive risks of lever-

aged institutions simply are not present for online

lending marketplaces or other alternative lend-

ers. To the extent that exportation of product

terms creates regulatory issues, those regulatory

39 12 U.S.C. § 1831d; Greenwood Trust Co. v. Mass., 971 F. 2d

818, 826 (1st Cir. 1992).

issues fall in the realm of consumer protection,

and in the wake of the passage of Dodd-Frank,

that playing eld has been largely leveled with

the creation of the Consumer Financial Protec-

tion Bureau.

The bank monopoly on national contracting

is also a relatively recent creation. Until the mid-

1960s, the prevailing rule in U.S. courts when

faced with disputes about which law to apply

to a lending agreement— the law of the do-

micile of the lender or the law of the domicile

of the borrower—did not turn on whether the

lender was a bank or an unchartered nancial

institution. Courts generally enforced the law

of the lender, rather than the borrower.

When

the prevailing judicial approach to conict of

laws changed in the 1960s, banks sought new

ways to ensure that their contracts could be en-

forced on a nationwide basis, and courts even-

tually latched on to the pre-emptive force of

federal banking statutes. Although non-banks

cannot currently claim a similar right, they could

regain the ability to export terms if courts simply

reverted to the conict rule that used to apply

to lenders regardless of charter—i.e., that the

law of the state of the lender, not the borrower,

governs the relationship between the two.

V. Conclusion

The broader point goes well beyond giving

non-banks the same ability to contract across

state lines as banks. In the nancial services

industry today, as in the telecommunications

and transportation industries a generation ago,

competition has essentially been lost as a guid-

ing regulatory principle. Regulatory compliance

has become an economic moat that existing

providers are using to fend off disruptive com-

petition. Rather than looking for ways to force

upstarts to join with those incumbents, regula-

tors in this industry should look for inspiration in

40 Ury v. Jewelers Acceptance Corp., 227 Cal. App. 2d 11 (Ct.

App. 1964).

142

the examples of the past and nd ways to level

the regulatory playing eld. Leveling the playing

eld will ensure greater consumer access to bet-

ter nancial products.

143

Legal boundaries

of Competition in

the Area of

Internet:

Challenges and

Judicial Responses

By Zhu Li

Abstract

Some new characteristics of competition in

the Internet industry, e.g., competition for atten-

tion, innovation competition, cross-market com-

petition etc., have brought about new chal-

lenges and difculties for the legal regulation of

competition. In virtue of the theoretical innova-

tion and the innovation of law applicability, Chi-

nese courts gave creative judicial responses in

the scopes of Anti-Unfair Competition Law and

antitrust Law, claried the legal boundaries of

competition and effectively regulated compe-

tition in the online environment. Certain trends

and rules implicit in this kind of judicial responses

are worth noting.

The growth of the Internet industry in our

country has become a new engine for the

development of domestic economy. The on-

line technology has been updating rapidly;

new products, new services and new business

models have been emerging uninterruptedly,

impacting existing models and established or-

der of the traditional commercial community,

changing the structure of commercial interests

in the economy and the society. This process is

accompanied with the increasingly erce com-

petition in the Internet area and the continuous

emergence of new kind of competition that

constantly challenge legal boundaries. At the

early stage of Internet development, the chal-

lenge had mostly appeared to be the difculties

in protecting online copyright. In pace with the

maturity and progress of the Internet technolo-

gy, the scale of e-commerce enlarges rapidly,

new business models relying on the Internet are

emerging uninterruptedly, therefore the men-

tioned challenge has expanded to areas such

as trademark protection, competition regula-

tion, etc. in the Internet. By examining specic

cases, Chinese courts have claried and delim-

ited competition rules on the internet, lled the

legal gaps, and thus, played an important role

in the regulation of competition on the Internet.

First, the paper describes the characteristics of

competition on the Internet. Second, it explores

the special challenges this kind of competition

has brought to the judicature, and third, the pa-

per will analyze the creative judicial responses

to the challenge, and endeavor to reveal the

trend and the regular pattern in such judicial

action

I. Basic features of competition in

the internet area

A. Competition for attention

The expansion of information continues con-

stantly in Internet, whereas everyone's time and

Zhu Li, the Judge of the Intellectual Property Division of

Supreme People’s Court, JD. The article represents only the author’s

own views beyond the position of its organization. Certain content of

this paper was discussed and reviewed in “Chinese IP judges Forum”

ofSouthwestUniversityofPoliticalScience&Law,№4,aswellasin

the second meeting of the Forum Internet Competition Policy and Inno-

vation, organized by the Electronics Intellectual Property Center of In-

dustry and Information Technology Ministry. Thanked the participants

of the forum for their comments, certainly the author is responsible for

any error in this paper.

144

energy are limited and valuable. Hence, Inter-

net users’ attention has become a scarce re-

source. Each Internet service provider tries to

do its best to obtain customers’ attention and

to focus it on its product by virtue of its particu-

lar tools, operations and services in purpose to

gain its business benets through customers’ at-

tention.

There are two ways to turn customers’

attention into business benets: 1) to develop

value-added services for interested customers

and thus gain prot and 2) to sell the gathered

customers’ attention to a third party that seeks

this kind of resource. Baidu, for instance, has

gathered great amount of Internet users via its

search engine platform and then began to gain

business benet by selling advertising or by of-

fering value-added services.

The competition to obtain customers’ atten-

tion in online environment has derived three

subsidiary features. First, the zero price (or nega-

tive price) competition. Internet service provid-

ers have found in the competition of obtaining

customers’ attention that the most effective

way to attract customers’ attention is zero price,

or even negative price. Zero prices for basic

services have become a mainstream business

model of Internet service providers. Recently, for

instance, in areas of instant messaging, search,

social networks, security, e-commerce etc. all

providers offer their basic services for zero prices.

In some areas of intensive competition providers

prefer even negative price competition by subsi-

dizing customers in order to continuously attract

and maintain users’ attention. Second feature

is the platform competition. In the online envi-

ronment the operations and services offered by

the providers are increasingly platform-related.

Services, offered by Internet service providers,

turn to be the platform, connecting two or more

groups of different entrepreneurs who take part

1 “Many network providers are competing in order to obtain

the limited attention of consumers, their featured products and services

are in fact just the tools to gain the customers’ attention in competi-

tion.” David S. Evans, “Attention Rivalry Among Online Platforms,”

University of Chicago Institute for Law & Economics, no. 627 (2013):

3. Available at SSRN: http://ssrn.com/abstract=2195340.

of business via the platform; in this form of busi-

ness the benet of one group of participants by

joining the platform depends of the scale and

the development of another group of entrepre-

neurs joining this platform. Internet service pro-

viders who offer the platform service will “create

value by virtue of reducing the conict between

different participants of the platform or lowering

the transaction cost.”

Internet service provid-

er will continuously increase the user's depen-

dence on their platforms and improve customer

stickiness by virtue of offering added particular

function that other platforms cannot propose

or supplying the function of better quality than

that of other platforms. Third of these features

is the motto “the customer is the king.” Internet

service providers pay high attention to custom-

ers’ needs, constantly update their products or

offer new type of services, and this way attempt

to improve the users’ experience, to enhance

the attractiveness of their products and ser-

vices, and nally to keep customers.

B. Innovation competition

Given the new ways to disseminate informa-

tion and the speed to do so, Internet, to a large

extent, provides endless resources and spaces

to expand innovation. The magic weapons in

the competition are rapid innovation and seize

and keep customers as soon as possible. In the

area of Internet the focus of innovation is shift-

ing from traditional innovation of technology

and production to innovation of business mod-

els, and the leading innovation factor is to shift

from a closed mode of innovation to an open

one. That is, innovation is no longer limited to the

companies’ development centers, but opened

to the community, providers and end-users, and

innovation is carried out in accordance with

customers’ demands and ideas. No doubt that

in online environments the customer is the de-

2 CitedfromDavidS.Evan,“MarketDenitionandMarket

Power for Internet Industries” (discourse in Seminar of Implementation

of Antitrust Law in ICT Industry of the Forum Internet Competition

Policy and Innovation, organized by the Electronics Intellectual Proper-

ty Center, 24 march, 2015).

145

cisive factor for the survival of the enterprise,

creativity is the leading factor for development

of the enterprise. Competition in online environ-

ments appears to be a dynamic competition,

where operators attempt to build their compet-

itive advantage by virtue of constant innova-

tion. “The period for leading enterprises of many

industries to keep their dominant position is get-

ting shorter. Enterprises resting on their domi-

nant position and revealing their former laurels

will soon be replaced by more innovative com-

petitors.”

Meanwhile products and services

differentiation offered by different providers is

becoming more relevant. Providers compete

for customers by offering different functions and

user experience.

C. Cross-market competition

Another obvious feature of competition in

the area of Internet is cross-market competi-

tion. Products and services, offered by a typical

network provider, are mostly based on software

and provided via Internet, where the market

entry threshold is not very high and the cost of

changing the scope of products and services

is relatively low. Thus, the network provider can

easily bring its existing customers into a new area

by adding services without customer churn, if

he has already obtained massive customer re-

source. Therefore, in the area of Internet, service

providers who offer different products and ser-

vices are used to look at each other as compet-

itors and do not stop to compete for customers’

attention.

D. Huge impact

he scope, magnitude, breadth, and depth

of impact caused by competition behavior

may expand unlimitedly and spread to the

entire Internet environment in a very short pe-

riod of time; it is difcult to eliminate the result

3 Jian-cai Zhang, “Dynamic competition and enterprise core

competitiveness,” East China Economic Management, no. 4 (2002):

81.

of such impact effectively. An Internet service

provider that faces intensive competition could

suffer customer churn for a couple of days, lose

its competitive advantage and even be forced

to withdraw from the market.

II. Challenges brought to court by

competition in the area of internet

The abovementioned characteristics of

the competition in the area of Internet have

brought about new challenges to court, con-

cerning at least three aspects: changes of judi-

cial functions; industry’s stronger expectation of

reasonably determined legal rules; right holders’

higher requirements for the prompt and effec-

tive judicial relief.

A. Changes of judicial functions

It is often difcult for the law to timely respond

to changes, occurring when technology and

business models are developing dramatically.

At the same time it is also difcult for the admin-

istrative law enforcement organs to promptly in-

vestigate the competitive behavior in the area

of Internet due to lack of a clear basis for en-

forcement. The court is thus inevitably pushed to

the forefront of solving disputes of network com-

petition. According to an incomplete statistics

report, Chinese courts have heard 126 cases

involving disputes of unfair competition in the

internet industry as of October 2014.

These dis-

putes demonstrate the following characteristics:

disputes occur frequently and easily shift along

with hot new technology and revolution of busi-

ness models; numerous disputes occur due to

new type of competitive behaviors that are not

yet clearly regulated by law and are to be ad-

justed by applying the guidelines of anti-unfair

competition law; many disputes are tentative,

i.e. the purpose of the litigation parties is not

4 For particularities of dispute of competition in the area of

Internet see Qin-kun Zhang, “A empirical analysis of unfair competition

cases of Internet development in China,” Electronics Intellectual Prop-

erty, no. 10, (2014): 26-37.

146

just ghting the interests of the pros and cons,

but rather requiring the judiciary clear industry

rules and code of conduct; disputes between

Internet service providers, spreading to disputes

between traditional enterprises and Internet

service providers, appear to be intense com-

petition for each one’s own interest. Thus, it be-

comes a major issue for the courts to provide

proper administered judicial rules for new type

of competitive behaviors, to guide and regulate

healthy competition while resolving disputes.

That means that the justice must assume not

only the functions of legal performer of compe-

tition law, but also the functions of competition

policy/rule maker. Along with implementing

national strategy of establishing an innovative

country and accelerating the pace of innova-

tion-driven development, the role of competi-

tion policy is increasingly prominent. “The closer

the forefront of knowledge, the higher the com-

plexity and uncertainty.”

The lower the accu-

racy and efciency of industrial policy and the

greater the role and value of competition pol-

icy, the more urgent of demands to maintain

a healthy competitive environment. The actu-

al situation requires justices to dene the legal

boundaries of the legality of acts by virtue of

creative application of the law in every specic

case, “bridging the gap between law and con-

stantly changing reality.”

To perform this function properly and to cre-

atively apply the law, the justice needs to be

not only prociency in the spirit of the law and

rules, but also to have a deep understanding of

the competition reality in the area of Internet,

knowledge of information technology and a

good understanding of the business develop-

ment model and innovation requirement. In the

process of dealing with some knotty disputes of

5 Rajneesh Narula, Globalization and technology: Interde-

pendence, innovation systems and industrial policy, (Blackwell Pub-

lishing Ltd., 2003).

6 Aharon Baraka, The Judge in a Democracy, trad. Bi Hong-

hai (Law Press, 2011), 17.

competition in the area of Internet, the court,

applying guidelines of Article II of the Anti-Unfair

Competition Law, has carried out certain explo-

ration of the standard of legitimacy of compet-

itive behavior. In Baidu, Inc. vs. Qihoo case of

violation of robots protocol, the trial court, for

instance, has put forward the rule of procedures

of “consultation-notice.”

In Baidu, Inc. vs. Qi-

hoo case of inserted standard, the trial court has

put forward the principle of “no interference in

case of non-public necessity.”

These explora-

tions have, to a certain extent, deepened the

understanding of the principle of Article II of the

Anti-Unfair Competition Law. These explorations

have, of course, sparked considerable contro-

versy that demonstrates a fact that a relatively

broad social consensus is not yet established in

this area.

The justice still has to take a heavy

burden and to embark on a long road for prop-

erly shaping the rule of law, responding to de-

mands of market and uniting social consensus.

B. Difculty of legal evaluation

The competition for obtaining customers’ at-

tention, innovative competition and cross-mar-

ket competition with the retrot of competition

means in the area of Internet leave a gray area

where competition bears the characteristics

between legality and illegality that cause seri-

ous difculty of legal evaluation. The difculty is

reected in the scopes of anti-unfair competi-

tion and antitrust laws.

This difculty is mainly reected in two aspects

of unfair competition. One of them is the prob-

lem of judging the competitive relationship. In

civil cases of unfair competition, the traditional

7 Ltd. vs. Beijing Qihoo Technology Co., Beijing Baidu

FMCT no. 2668 (Network Technology Co., Ltd. Case of unfair compe-

tition dispute (violation of robots protocol), (2013).

8 Baidu Online Network Technology (Beijing) Co., Ltd v.

Beijing Qihoo Technology Co., Ltd. (Beijing High Court Civil Judg-

ment (2013) Gao Min Zhong Zi no. 2352)

9 Xue Jun: Question “the principle of no interference in case of

non-public necessity,” Electronics Intellectual Property, (2015): 66-70.

147

theory and practice of the anti-unfair compe-

tition law sets the existence of competition be-

tween the plaintiff and defendant as the pre-

conditions for relief of the plaintiff. In the online

environment the real and strong competition

exist also between the operators of different

products and services due to the competition of

obtaining customers’ attention, platform com-

petition and cross-market competition. Many

victims of unfair competitive behaviors cannot

be protected by law, if the existence of compe-

tition (even direct competition) remains to be

the precondition for legal remedy. It is thus ap-

pears to be the theoretic and practical issue the

justice should address to re-identify the compet-

itive relationship and to re-dene the competi-

tive relationship in anti-unfair competition law.

Another aspect of the difculty mentioned

above is to judge the legitimacy of competi-

tive behavior. Unfair competitive behaviors in

the Internet environment are carried out mostly

by means of the implementation of information

technology and often in the name of tech-

nological innovation with powerful technical

features. Operators tend to defend their com-

petitive behaviors on the excuse of necessary

measures to meet customers’ demands. The

competitive behavior of one operator usually

does not immediately have an impact on its

competitor, but it does through the customer as

the intermediary, i.e. the customer’s interest is

kidnaped by the Internet service providers who

exploit their customers as a shield to ght their

competitors for their own business benet. In

comparison with traditional competition, com-

petition in the Internet environment is more neu-

tral, at least on the surface; its boundaries of le-

gitimacy are blurred. Technological and neutral

features of competitive behaviors make more

difcult to dene the subjective fault of the op-

erator and the economic effects on competi-

tion. Under certain conditions, the competitive

behavior, though injuring the competitor, can

enhance consumer’s welfare or it might impact

consumer’s welfare for a short time, but can

enhance consumer welfare in the long term.

The traditional anti-unfair competition law sim-

ply denes the boundary of legality based on

the virtue of typical characteristics of conducts.

Such a method could not be used in an online

environment.

The “technologization” of com-

petition increases the relevance and interde-

pendence of competing operators: products

and services offered by one provider, inevitably

relate to products and services offered by other

providers, and thus will impact the business of

the latter. The standard of competition and busi-

ness achievement is characterized by compet-

ing for its own business results and trying not to

disturb others. Such standard increasingly shows

its drawbacks in judging the justication of com-

petition since it has wider scope of attack and

hampers free competition.

Meanwhile, the in-

novation in area of Internet is frequent, compe-

tition is complex and evolving. It is thus difcult

to form commonly recognized business ethics

timely. It is increasingly difcult to seek an ethical

consensus in a fragmented, divided society. The

moral evaluation criteria, applied in traditional

anti-unfair competition law, is fallen in straitened

circumstances, it is thus not an easy task for us to

extract stable and clear legal criteria.

Competition in the area of Internet also se-

riously challenges the applicability of antitrust

law. First of all, the difculty in dening the rel-

evant market. The zero price competition and

the platform competition and other features

10 The so-called typed features of behavior refer to the typical

characteristicsofabehaviorreectedonthesubject,object,subjectivity

and objectivity which are the four constituent elements of a behavior.

Constituent elements of unfair competition are commonly the follow-

ing: the subject is the participant of various market trade operations;

objectivity is the behavior which violates the principle of good faith or

generally accepted business ethics; the perpetrator is subjectively fault;

the object of the infringement is the interest of the operator, the interest

of the customer and the public interest of the society. See Wang Xian-

lin, Competition Law, (China Renmin University Press, 2006), 96-98.

11 The so-called performance competition refers to the perfor-

mance of their goods or services at competitive prices or to start com-

petition of their own business activities, also known as the effectiveness

competition. See Fan Chang-jun, A study of German Anti-Unfair Com-

petition Law, (Law Press China, 2010), 114-115.

148

of competition in the area of Internet have

brought about greater uncertainty in dening

the relevant market. Theorists and practitioners

are debating whether the free market under

conditions of zero price competition is the rele-

vant market in sense of antitrust law and how to

carry out Hypothetical Monopolist Test (“HMT”)

in this circumstance. Inuenced by the platform

effect due to the existence of bilateral or mul-

tilateral market, the market boundaries in the

area of Internet are characterized by the high

ambiguity and thus far less clear than that of tra-

ditional market. The widely applicable method

for market denition in traditional areas cannot

therefore be directly applied to dene the rele-

vant market in the area of Internet. Another issue

is the increasing difculty to dene the abuse of

dominant market position. Under the condition

of blurred market boundaries the market share

has signicantly reduced its role of indicators,

which used to measure the enterprise market

forces. The probability of miscarriage of justice

will increase in case of dening the dominant

market position based solely on market shares. It

is hence an urgent issue, which the justice must

resolve, to improve the rationality of denition of

the dominant market position and the abuse of

dominant market position.

C. New demands of judicial relief

The competition in area of Internet is a

far-reaching behavior, it increases rapidly, is

hardly eliminated so that the survival of some

enterprises will be threatened. The character-

istics of Internet competition have proposed

new requirements of the timeliness and effec-

tiveness judicial protection. If the innovation

and fair competition of an enterprise cannot

be promptly protected, the enterprise might fall

into the dilemma of “won the lawsuit, but lose

the market,” even if it could ultimately succeed

in the litigation. From the point of view of judi-

cial precedents the most serious problems the

Parties complain about are insufcient com-

pensation, delayed temporary remedial mea-

sures etc. Among more than 120 cases of dis-

putes of unfair competition in area of Internet,

the amount of highest compensation is RMB 5

million, only 4 percent of the claim of the Party

of this case.

The case of genuine “Kaixin001”

had lasted more than one year, the corpora-

tion, though had won the litigation and got RMB

400,000 for compensation, had lost its market

share, eroded by its competitors, and had not

been able to reemerge. The function of judicial

relief, suffering great shortage in the timeliness

and effectiveness, cannot yet fully meet the de-

mand of competition in area of Internet.

III. Creative responses and adjust-

ments from the judiciary

A. Innovation and breakthrough of com-

petition theory

The new characteristics of competition in the

online industry (e.g. competition for attention,

cross-market competition etc.) have redened

the competitive relationships and the role of ju-

dicial relief in unfair competition cases. Judicial

precedents have to break through the limita-

tions of existing theories by resorting to theoret-

ical innovation. In Baidu Inc. vs. Unicom Qingd-

ao, Osun Network and others, for instance, the

appeal court held:

There exists competition between Unicom

Qingdao and Baidu Inc, insofar as Unicom Qin-

gdao has carried out the commercial activity

of poping-up ads prior to the result of Baidu

search. Thus, Unicom Qingdao competes with

the paid search activity of Baidu, although

Unicom Qingdao (network provider) and Baidu

Inc. (provider of search service) offer entirely dif-

ferent services.

12 Qihoo vs. Tencent of “QQ guards”, civil judgment of CF no.

5 (the Supreme People’s Court. 2013).

13 Shandong Province Higher People's Court, the civil judg-

mentofLCSF№5-2.

149

This denition has broken through the

theoretic limitation of direct competition and in-

cluded indirect competition. Summing up prec-

edents, the Supreme Court further points out in

its justice policy: “the competitive relationship

should be correctly dened, i.e. among all com-

petitors that take part in the market and are

impacted by the unfair competition. Between

them, there exists competitive relationships, no

matter if it is direct competition or not.”

Ac-

cording to this justice policy, the competitive

relationship between operator A and other op-

erators, whose business is impacted by the com-

petitive behavior of operator A, can be dened

as such. In Heyi Information Technology (Beijing)

Co., Ltd vs. Security Software Company Kingsoft

Corp. about “Cheetah browser shielded video

advertising” the appeal court had proposed

two conditions to dene a relationship of com-

petition: whether the behavior of the operator

can harm the interest of other operators; wheth-

er the operator can gain actual or potential

benet by virtue of this behavior. Meanwhile

the trial court held that the criterion to dene

a competitive relationship does not rely on

whether competitors are of the same industry.

It can clearly be seen that the position of the

trial court is actually the specic application of

the Supreme Court’ justice policy. To a larger

extend, competitive relationship means that the

importance of a competitive relationship is de-

clining to nd unfair competition. Moreover, trial

courts do not even carry out particular investi-

gations of the competitive relationship between

the parties in many cases, which means that the

competitive relationship can no longer be an

obstacle to dene unfair competition. Thanks

to the theoretical innovation the justice has bro-

ken away with the legal predicament based on

the narrow understanding of competitive rela-

tionship, and thus has met the characteristics of

competition in the online industry.

14 Xi Xiao-ming, Vice President of Supreme People's Court, in

the forum of national IPR trial courts (28 November, 2011).

15 The Beijing First Intermediate People's Court, the civil

judgmentofFICF№3283.(2014).

B. Adjustments of judging the fairness of

competitive behavior

Given the new features of competition on In-

ternet such as technicality, neutralization, and

complexity of the effect of interests, courts have

proposed new alternatives to judge the fairness

of competitive behavior and new methodolog-

ical adjustments.

First, the criteria of business ethics tend to be

objective. In the traditional commercial sectors,

business operators have formed commonly ac-

cepted business ethics. These ethics become

dominant in determining the fairness of com-

petition behaviors. “All competitive behaviors,

violating conventional honest practices in com-

mercial activities, constitute unfair competi-

tion.”

The justice policy of the Supreme Court

points out:

Any action, even if it is not forbidden

by a particular provision of the anti-un-

fair competition law, can be regulated in

accordance of the provisions of the ap-

plicable principles, if such action can be

dened as unfair by harming the legitimate

rights and interests of other operators and

violating the principle of good faith and

commonly accepted business ethics, and

the fair competition order cannot be main-

tained without ceasing such action

Meanwhile, to prevent generalization and

subjectivity of moral judgment, the judicial

practice denes the business ethics as “stan-

dards of behavior, generally recognized and

accepted in specic business areas, its objec-

tivity is embodied in its common acceptance

and commonality.”

The criterion of business

16 Paris Convention for the Protection of Industrial Property,

September 28, 1979, Article 10.bis.

17 The Supreme People’s Court: “Some advices in consider-

ation of full trial function of intellectual property, promotion of devel-

opment and prosperity of socialist culture, promotion of the coordinated

and autonomy development of economic”, Article 24.

18 TheSupremePeople’sCourt,theciviljudgmentofCTF№

1065. (2013).

150

ethics continues to play an important role in the

denition of fairness in the new type of compe-

tition in Internet. Nevertheless, the commonly

accepted business ethics of Internet business is

still at the stage of formation and development

due to the innovation and rapidly developing

competition in this area. The court has to seek

a more objective form of business ethics when

applying the business ethics as the standard of

the competition fairness evaluation. In the cas-

es of “QQ Guards” and “Robots Protocol,” the

courts had considered the industry standards

and self-regulation in Internet as an important

origin of the criteria for discovery and denition

of standards of conventional industry behaviors

and commonly accepted business ethics.

The

standard of behavior is the means of adminis-

tration, whereas self-regulation is the means of

self-management of operators in the industry,

both means do not exactly accord to commonly

accepted business ethics. Therefore, the courts

emphasizes that it is necessary “to rely on the

judgment of their legality, impartiality and ob-

jectivity, (the mentioned) relevant means can

be taken as references for dening standards of

conventional industry behaviors and commonly

accepted business ethics in Internet.”

Second, evaluating the effect of competitive

behavior has become more important in the

judgment of its fairness. The objectivity of busi-

ness ethics in Internet is getting more difcult

because of the technicality and neutralization

of competitive behavior. Therefore, courts be-

gan to pay attention to the evaluation of the

effect of the competitive behavior and seek to

justify the fairness of the behavior by assessing

the impact of the behavior on the legal interest

protected by competition law. In the case of

“Cheetah browser shielded video advertising,”

19 TheSupremePeople’sCourt,theciviljudgmentofCTF№

5 (2013) and The Beijing First Intermediate People's Court, the civil

judgmentofFICF№2668.(2013).

20 TheSupremePeople’sCourt,theciviljudgmentofCTF№5.(2013).

the civil court had investigated and analyzed

the harm that Heyi Information Technology (Bei-

jing) Co., Ltd has suffered due to the behavior of

“Cheetah browser shielded video advertising”

based on the behavior’s perspective effects,

and assessed the long-term impact on the inter-

est of customers and public interest. Then came

to the conclusion that the behavior of “Chee-

tah browser shielded video advertising” consti-

tuted unfair competitive behavior.

More cases demonstrate a new mode of

evaluating the fairness of behavior: the judg-

ment is carried out on the basis of comprehen-

sive evaluation of the impact of competitive

behavior on the interest of competitors, the in-

terest of customers and the public interest of so-

ciety. The analysis of the harm that the operator

suffers as the result of competitive behavior is set

to be the logical starting point. For the opera-

tor’s interest is the object, directly protected by

the anti-unfair competition law. All concerned

cases have considered the impact on opera-

tors through competitive behavior. The custom-

er is the object of competitive behavior that

bears the result of competition and accepts the

market products; it is therefore the nal aim of

justice to enhance consumers’ welfare. It should

be an integral part of the judgment of fairness

of competitive behavior to consider the impact

of competitive behavior on the customers’ in-

terest with respect of enhancement of consum-

ers’ welfare and their fundamental benet. The

customers’ interest is itself multilevel and relative

due to differentiation of consumer groups and

their divergent interests. Trial courts have been

paying attention to this issue and assessing the

different interests of consumers with different

weights. Trial courts have been following more

closely the inuence of competitive behav-

ior on the right of customers to know and to

choose; to harm this type of customers’ inter-

est will more likely be conrmed to constitute

21 The Beijing First Intermediate People's Court, the civil

judgmentofFICF№3283.(2013).

151

unfair competition. Present judicial precedents

demonstrate that the positive customer experi-

ence and other kind of customers’ interest that

can be resolved via the market do not have a

signicant impact on evaluating the fairness of

competitive behavior. On this basis, many judi-

cial precedents have investigated the impact

of competitive behavior on the public interest

of society and analyzed whether this kind of

behavior could harm the healthy mechanism

of market competition. The method of compre-

hensive evaluation of negative and positive re-

sults is very close to the principle of rationality for

antitrust analysis; it indicates the integration and

interoperability of the anti-unfair competition

law and the antitrust law.

Third, the trend of multi-angle evaluation. Be-

cause there is no conict between criteria of

moral evaluation and criteria of evaluation of

competition result, the nationality and persua-

siveness of the evaluation result will improve if

the fairness of competitive behaviors is inspect-

ed under various angles, e.g. moral evaluation,

efciency competition, principle of proportion-

ality, assessment of competitive effects etc. The

typical cases of application of this method are

“QQ guards” and “Cheetah browser shielded

video advertising.”

C. Innovation in application of antitrust law

The competition in the Internet industry obvi-

ously differs from the competition in traditional

areas. The logic of analysis and method, widely

applicable for monopoly behaviors in tradition-

al areas, cannot be applied directly in the Inter-

net industry. The justice has carried out targeted

adjustments and innovation in accordance with

the competitive features on the Internet.

First, innovation in the analysis of abuse of

dominant market position cases in the Internet

industry. In the traditional antitrust law there are

three patterns of analysis of abuse the dominant

market position: patter 1 “relevant market-mar-

ket power- competitive effects (R-M-C)”, for

which the denition of relevant market is the

insurmountable starting point of analysis of the

monopoly. Pattern 2 “market power - compet-

itive effects (M-C)”, for which the starting pint

of analysis is the denition of market power that

can be tested and veried by virtue of direct or

indirect evidence. Under this pattern of analy-

sis the denition of relevant market can be cir-

cumvented. Pattern 3 “behavior - competitive

effects (C-C)”.

For traditional antitrust judicial cases, pat-

tern 1 is currently the leading pattern of anal-

ysis in European and American courts, where-

as pattern 2 is rarely applied in practice, and

pattern 3 has not yet been applied. In Inter-

net, the boundaries of relevant market are ob-

scured even more due to the competition for

customers’ attention, the platform competition

and cross-market competition. Thus, the justice

should keep the necessary caution when ap-

plying R-M-C patter and apply the two latter

patterns as the preferred tool of analysis. Chi-

nese courts have carried out valuable attempt

in this relation. In Qihoo 360 Technology Co Ltd

vs. Tencent computer system Co. Ltd. of abuse

the dominant market position, the trial court, in

an in-depth investigation of the relevant market

concerned and analyzing the market power of

Tencent in this market, had come to the con-

clusion that Tencent company does not possess

the dominant market position. Yet the trial court

did not cease the investigation and analysis at

this point, but had further evaluated the actual

or potential effect of the respondent monopo-

listic behavior on the market competition and

had carried out the nal judgment on this ba-

sis.

For this method of analysis of the relevant

market, the market power and the effect of

competitive behavior are considered related

and referenced factors, but not separate stag-

es of analysis; the rationality of the denition is

22 TheSupremePeople’sCourt,civiljudgmentofCTF№4.

(2013),

152

thus improved due to a cross-verication of all

mentioned factors. For this method of analysis,

these three patterns can be exibly chosen in

accordance with a particular case. The C-C

pattern could be chosen when it seems to be

difcult to dene the relevant market and the

market dominant position, thus the relevant

market could lie beyond a clear denition. The

caution should be kept in nding the indicator

effect of market share on the basis of features

of online competition even when the relevant

market and market share are dened by using

the former two patterns. In this case factors such

as market entry, market behavior and econom-

ic results become the focus of attention

The competition in circumstance of Internet is

highly dynamic, the boundaries of relevant mar-

ket are thus far less clear than that in traditional

areas, the indicator effect of market share in this

case should not be overvalued, more attention

should be paid to those factors such as market

entry, market behaviors of operators and their

inuence on the competition etc. which can

help to determine the facts and evidence of

dominant market position.

Meanwhile trial courts have applied more

exible methods to analyze various factors with

respect to the particularities of the competi-

tion in Internet: when dening relevant market

of goods and services of relative platform fea-

tures, possessing certain but not very close sub-

stitution, such products and services could be

involved in consideration of the inuence on the

behavior of hypothetical monopolist, no matter

whether they can be included in the scope of

relevant market or not. When determining the

dominant market position, the analysis of the re-

sult of competition should not be abandoned

even after certain preliminary conclusions, the

consideration of the result of competition and

the verication of the accurateness of judg-

23 Id.

ment of the dominant market position should be

carried out further.

Second, adjustment of Hypothetical Monop-

oly test (“HMT”). In Qihoo 360 Technology Co Ltd

vs. Tencent computer system Co. Ltd., the trial

court had explored the applicability of HMT in

the online industry and the method of its spe-

cic application. The trial court held that, as a

method to analyze relevant market, HMT has

universal applicability, yet the particular analysis

by means of HMT should be carried out in de-

pendence of the area of market competition,

concerned in the specic case, and the rele-

vant data that can be obtained. Competitors

in the area of Internet pay more attention to

quality, services, innovation etc., but not price.

Customers have very high price sensitivity, so

to them, it would seem a great change of the

features of products and the business model, if

free products or services would have turned to

paid ones. The HMT in terms of price increment

is thus not fully applicable in area of Internet, yet

the alternative forms of this method, e.g. HMT

on the basis of quality degradation, could still

be applied.

Third, innovation in the method of analyzing

relevant market and market power, related to

multi-sided platform cost-free for users. Econo-

mists usually apply the method of conversion

analysis for dening relevant market of platform

products or services for charge-free users by re-

garding the charge-free basic services as the

investment for platform products with the pur-

pose of converting the market of platform prod-

uct to common and paid single market.

This

method, though simple and convenient, might

not only exaggerate the inuence of platform

competition, but also to certain extent neglect

the connection and interaction of both ends

of the platform; it is hence neither scientic nor

24 Herbert Hovenkamp, The antitrust enterprise: principle

and execution, trad. Wu XU-liang (University of Finance & Econom-

ics Press, 2011), 48.

153

accurate. In Qihoo 360 Technology Co Ltd vs.

Tencent computer system Co. Ltd. the trial court

did not apply this method of analysis, but, had

investigated whether the features of platform

competition can affect the denition of relevant

market based on “whether the competition be-

tween online platforms, competing for custom-

ers’ attention and advertisers, totally steps over

the boundary determined by the features of

products or services and thus imposes sufcient-

ly strong competitive constraint on operators,”.

The trial court had chosen the free instant mes-

saging service as the criterion to dene the rel-

evant market because of the absence of exact

empirical data that could prove that platform

competition has imposed sufciently strong

competitive constraint on operators, as well as

in case that the competitive behaviors involved

in litigation occur mostly at the free users’ end.

At the same time, the trial court did hot neglect

the inuence of platform competition by taking

it in proper consideration when dening market

position and market power. We see that the tri-

al court has exibly applied the new method of

analysis for specic case when economics is not

yet able to supply any more persuasive pattern

of analysis for relevant judicial cases. The revela-

tion is: for platform-related products or services

there is no xed pattern of analysis, the relevant

market and dominant market position should be

determined depending on each specic case.

What the court has to do is to take into consider-

ation features of platform and the interaction of

both ends of the platform, and then accurate-

ly identify the actual or potential competitive

constraint the operators can face.

Fourth, advocate for an objective and ef-

fect-oriented method of analysis. The antitrust

action follows closely whether the respondent

monopolistic behavior will distort and destruct

the healthy, orderly and energetic competition

mechanism. Such action “has involved in itself

neither moral content nor ethic law, appropri-

ately designed for business.”

In the antitrust

case the important thing is industry reality and

economic rationality, the moral thinking must

thus be avoided. One should search the origi-

nal sin of monopoly merely in competition result

and economic reality. In Qihoo 360 Technolo-

gy Co Ltd vs. Tencent computer system Co.

Ltd. the trial court did not carry out any moral

evaluation of “either-or” and other behaviors

of Tencent, but had focused on the effect of

the respondent monopoly behaviors. After a

comprehensive evaluation of the actual and

potential passive and positive results that these

behaviors caused, the court came to the con-

clusion that such behaviors were legal. In this

process, the trial court had followed the meth-

od of investigation that especially focuses on a

specic industry and in a specic behavior. In

other words, to investigate a specic behavior

of a specic industry with respect to the charac-

teristics of the respondent monopoly behaviors

and its impact on the competition, to consid-

er in detail platform effect and network effect

on each specic case, and thus to more accu-

rately determine the impact of behavior on the

competition.

Fifth, the creative combination of legal judg-

ment and economic analysis. The analysis and

judgment of legality of the monopolistic behav-

ior usually relies on the economic analysis, yet

the ultimate decision will be carried out by the

judge. The economic analysis only makes avail-

able different tools for the proper legal judg-

ment; the judge cannot thus transfer the right to

rule to economists. Therefore the judge should

creatively combine the legal judgment and

the economic analysis in the antitrust case by

properly applying the conclusion of econom-

ic analysis for improving the accuracy of legal

judgment of monopoly behavior. In Qihoo 360

25 For price-related method of analysis see Peter Davis and

Eliana Garcés, Quantitative Techniques for Competition and Antitrust

Analysis, trad. Zhou De-fa. (China Renmin University Press, 2013),

139.

154

Technology Co Ltd vs. Tencent computer sys-

tem Co. Ltd. the trial court, analyzed whether

the acquisition costs of mobile terminal equip-

ment constituted an obstacle for including the

mobile instant message service in the relevant

market. In doing so, it considered that for the

customer who possesses both mobile terminal

device and a PC, the acquisition costs of mobile

terminal equipment had already become sunk

costs. This fact cannot be changed due to any

current or future decision of the provider, the

acquisition costs cannot any way inuence on

the customer’s choice between mobile termi-

nal device and PC as the preferred equipment

for instant message. Thus, the court came to

the conclusion that acquisition costs of mobile

terminal equipments will not constitute any ob-

stacle for including the mobile instant message

service in the relevant market in this case. This is

a typical example of applying economic com-

monsense in a specic case.

According to economists, the trial court has

carried out even more professional price-relat-

ed analysis determining whether social network

Weibo and instant message can be included

in the same relevant market. This is an example

of combination of economic analysis and legal

judgment. For a correct and reasonable appli-

cation of economic analysis, the methodolog-

ical errors should be prevented and the limita-

tion of data and constraint of conditions should

be properly considered. A judge should break

away from arrogance and prejudice when ap-

plying the method of economic analysis and

adopting a conclusion. It is important for justice

that the judge, besides the necessary econom-

ic knowledge, maintains the principle of “effect

rst.” In other word, the judge should follow close-

ly the actual or potential effect of respondent

monopoly behaviors on competition, and in do-

ing so, adequately apply the economic analysis

and stay away from possible methodological

errors, limitation of data and constraint of condi-

tions. The more realistic, more rational and more

accurate conclusion of the legal identication

of monopolistic behavior can nally be drawn

as long as the essence of the monopolistic be-

havior that causes actual or potential negative

effect on the competition will be grasped. The

result of the economic analysis based on direct

evidence will be veried pursuing to the “effect

rst” principle. This could explain the judicial ac-

tion in Qihoo 360 Technology Co Ltd vs. Tencent

computer system Co. Ltd., where the trial court,

applying an economic analysis, investigated

whether the monopoly’s actions will exclude

or restrict competition on the basis of almost all

testimonies related to the effect of its behaviors.

IV. Conclusion

From the point of view of interpretation and

application of substantive law, the analysis

abovementioned demonstrates that the jus-

tice has creatively responded to the competi-

tive behaviors in an online environment. It also

claried legitimacy boundaries of competitive

behaviors, and thus effectively regulated com-

petition on Internet. Some responding measures

and innovation of applicable judicial methods

has had a profound inuence on international

justice. Yet, needless to say that there are a lot

of shortcomings of the responding measures

and methods mentioned, e.g., there are many

misunderstandings, even mistakes in dealing

with the correlation between anti-unfair com-

petition law and antitrust law with respect of

applying the substantive law. There exist more

or less specious and vague criteria for judging

the legitimacy of competition; the economic

analysis of monopolistic behavior and the as-

sessment of the competitive effects are not yet

well skilled; the lack of timely and effective ju-

dicial relief seriously impacts and limits the fully

implementation of judicial efciency. A further

improvement is to be planned in the future judi-

cial practice.

155

Can Big Data

Protect a Firm from

Competition?

Anja Lambrecht and Catherine E. Tucker

December 18, 2015

Executive Summary

There is plenty of hype around big data, but

does it simply offer operational advantages, or

can it provide rms with sustainable competitive

advantage? To answer this question, we look

atbigdatausingaclassicframeworkcalledthe‘re-

source-basedviewoftherm,’whichstates that,

for big data to provide competitive advan-

tage, it has to be inimitable, rare, valuable, and

non-substitutable.

Our analysis suggests that big data is not in-

imitable or rare, that substitutes exist, and that

by itself big data is unlikely to be valuable. There

are many alternative sources of data available

to rms, reecting the extent to which custom-

ers leave multiple digital footprints on the in-

ternet. In order to extract value from big data,

rms need to have the right managerial toolkit.

Thehistoryofthedigitaleconomyoffersmanyex-

amples,likeAirbnb,UberandTinder, whereasim-

pleinsightintocustomerneedsallowedentryin-

tomarketswhereincumbentsalready had access

to bigdata.

Anja Lambrecht is an Assistant Professor at London Busi-

ness School, London NW1 4SA, UK. Catherine Tucker is the Distin-

guished Professor of Management Science at MIT Sloan School of

Management, MIT, Cambridge, MA 02139, USA. The authors thank

the Computer and Communications Industry Association for generous

funding of this research. All mistakes are our own.

Therefore, to build sustainable competitive

advantage in the new data-rich environment,

rather than simply amassing big data, rms

need to focus on developing both the tools and

organizational competence to allow them to

use big data to provide value to consumers in

previously impossible ways.

I. Introduction

The digitization of the ofine and online econ-

omy alike means that rms are naturally collect-

ing ’big data’, distinguished by its volume

, va-

riety of formats spanning text, image and video,

and velocity, meaning that data is recorded in

real time.

There is plenty of hype around big data.

Firms are constantly exhorted to set strategies in

place to collect and analyze big data (Bughin

et al., 2010; Biesdorf et al., 2013), and warned

about the potential negative consequences of

not doing so. For example, the Wall Street Jour-

nal recently suggested that companies sit on a

treasure trove of customer data but for the most

part do not know how to use it.3 Recent articles

such as McGuire et al. (2012) and McAfee et al.

(2012) have made cases for why big data of-

fers a short-term operational advantage, both

in terms of cost and performance, for rms who

nd ways of using it successfully.

1 Companies such as Amazon and Walmart already work with

petabytes of data in a single data set (McAfee et al., 2012).

2 Thisfunctionaldenitionofbiddatadoesnotspecifythe

depth of consumer insight it can provide. Big data

spansanonymizeduserdata,personallyidentiableinformation,search

query data, web browsing data or data on consumer sentiments or pur-

chaseintentions.Dependingonthespecictypeofdataunderconsider-

ation,howvaluableitistothermmaydiffer.

3 http://www.wsj.com/articles/the-untapped-value-of-cus-

tomer-data-1444734633?mod=djem_jiewr_MK_domainid4https://hbr.

org/2015/01/why-nordstroms-digital-strategy-works-and-yours-proba-

bly-doesnt. This article highlights that because digital technologies are

visible and accessible to competitors, it is hard to generate a competitive

advantage.

156

However, big data’s long-term strategic, rath-

er than operational, implications for rms are

less clear. Academic opinion differs on whether

it will lead to a new type of competitive advan-

tage (McGuire et al., 2012) or not.

The question

of whether big data can indeed confer sustain-

able competitive advantage is critical for rms

but has, to our knowledge, received surprisingly

little systematic attention.

To evaluate the strategic role of big data as

a source of sustainable competitive advantage

or as a barrier to entry, we use a classic frame-

work in strategic management sometimes re-

ferred to as the ‘resource-based view of the rm’

(Wernerfelt, 1984; Barney, 1991; Peteraf, 1993;

Barney, 2001). This literature is useful because

it sharply distinguishes factors that enhance an

entire industry from a ‘sustainable competitive

advantage’ that bts a single rm. For there to

be a sustainable competitive advantage, the

rm’s rivals must be unable realistically to dupli-

cate the benets of this strategy or input.

Specically, Barney (1991) suggests that for a

rm resource to be a source of competitive

advantage, the resource has to be inimitable,

rare, valuable, and non-substitutable. In a sim-

ilar spirit to Markman et al. (2004)’s analysis of

patents, we examine along each of these di-

mensions whether big data is a source of sus-

tainable competitive advantage to rms.

II. Is Big Data Inimitable?

For big data to be inimitable, no other rm

should easily be able to replicate the advan-

tage. There are two underlying economic rea-

sons for why big data in many instances is unlike-

ly to be inimitable. First, big data is non-rivalrous,

4 https://hbr.org/2015/01/why-nordstroms-digital-strategy-

works-and-yours-probably-doesnt. This article highlights that because

digital technologies are visible and accessible to competitors, it is hard

to generate a competitive advantage.

meaning consumption of the good does not

decrease its availability to others. Second, big

data has near-zero marginal cost of production

and distribution even over long distances (Sha-

piro and Varian, 1999). These two basic charac-

teristics, combined with the fact that customers

constantly leave footprints on the internet, have

lead to a thriving industry where consumer big

data is resold.

This type of commercially available big data

allows new entrants to gain insights similar to

those available to rms that own big data on

a large number of customers. There are many

examples of large commercially available data

sets. Acxiom has ‘multi-sourced insight into ap-

proximately 700 million consumers worldwide’

with over 1,600 pieces of separate data on

each consumer; Datalogix asserts that its data

‘includes almost every U.S. household.’

Com-

cast is planning to license TV viewing data col-

lected through set-top boxes and apps.

Other

companies, such as the Oracle- owned Bluekai,

sell cookie-based user information online to al-

low for targeting advertising based on a user’s

past activities or demographics. Bluekai states

that it has data on ‘750 million unique users per

month with an average of 10-15 attributes per

user.’

To protect both their customers and

themselves, such companies undertake to en-

sure that their data collection complies fully with

data protection rules.

Given the different possible types of big data,

an obvious question is whether this analysis ex-

tends to cases where the big data has what

appears to be unique or individual insights. For

5 See Acxiom Corp., 2013 10K Annual Report for the Period

Ending March 31, 2013 and Staff of S. Comm. on Commerce, Sci., and

Transp.,OfceofOversight&Investigation,AReviewoftheDataBro-

ker Industry: Collection, Use and Sale of Consumer Data for Marketing

Purposes.

6 http://www.wsj.com/articles/comcast-seeks-to-harness-

trove-of-tv-data-1445333401

7 https://docs.oracle.com/cloud/latest/daasmarketing_gs/D

157

example, recently the retail store Target hit the

headlines because of its alleged ability to use

its retail shopping data to predict a pregnan-

cy even before close relatives knew about it.

However, even such highly specic and timely

data-driven insights are easy to imitate for rms

that do not own a national database of retail

sales. For example, a marketing unit of the cred-

it-scoring agency Experian sells frequently up-

dated data on expecting parents, along with

income and rst-birth information.

In addition, data that is available due to indi-

vidual consumer-level tracking is complement-

ed by the explosion of user-generated content

where consumers themselves create a footprint

of their behavior, likes, opinions and interests

across the internet. Recent research in comput-

er science has emphasized that by combining a

myriad of external online proles, external rms

can gain huge insights into any one customer

(Narayanan and Shmatikov, 2008; Calandrino

et al., 2011). Firms can also use such content as

a direct substitute for customer data. For exam-

ple, Edelman (2015) discusses that Zillow.com

was able to build a successful home-buying

digital platform by relying on existing town as-

sessment data.

In short, where a market for data exists, it is

unlikely that big data is inimitable.

III. Is Big Data Rare?

For Big Data to be a ‘rare’ resource would

8 http://www.nytimes.com/2012/02/19/magazine/shop-

ping-habits.html?_r=0 There are some doubts over the origin of this

story and whether Target actually did this - see for example http://www.

kdnuggets.com/2014/05/ target-predict-teen-pregnancy-inside-story.

html

9 http://www.experian.com/small-business/prenatal-lists.jsp

10 He highlights the interaction between publicly available

information and user generated content, saying, ‘Zil- low’s initial in-

formation was good enough to attract consumer interest, at which point

property owners happily contributed corrections, photos, and other in-

formation. Indeed, real estate agents were soon willing to pay to show

their advertisements in and around Zillow’s property listings.’

mean that few other rms possess it. However,

there are two reasons why this is unlikely to hold.

First, large shifts in supply infrastructure have

rendered the tools for gathering ‘big data’

commonplace (Greenstein et al., 2013). Cloud-

based resources such as Amazon, Microsoft,

and Rackspace make these tools not depen-

dent on scale

and storage costs for data con-

tinue to fall, so that some speculate they may

eventually approach zero

. This allows ever

smaller rms to have access to powerful and

inexpensive computing resources. Free open

source technologies such as Hadoop that allow

users to analyze large datasets are widely avail-

able and accessible.

Second, as consumers’ lives increasingly shift

to the web, consumers leave traces of their

needs and preferences everywhere. Firms who

embrace these low-cost digital technologies

have many opportunities to gather customer

data. Telecom companies can collect data on

calling behavior and browsing on their phones;

Amazon, Macy’s and Walmart collect detailed

consumer-level purchase data, while platforms

such as Bluekai collect a large range of detailed

consumer browsing and purchasing information

across multiple website.

Indeed, such ‘multi-homing’, that is the use

of multiple different digital services by consum-

ers, means that similar pieces of information are

often available to many different companies.

11 http://betanews.com/2014/06/27/comparing-the-top-three-

cloud-storage-providers/

12 http://www.enterprisestorageforum.com/storage-manage-

ment/can-cloud-storage-costs-fall-to-zero-1.html

13 The European Commission spoke similarly in 2014 when

concluding its investigation into Facebook’s acquisition of WhatsApp.

Itconcludedthat‘therearecurrentlyasignicantnumberofmarketpar-

ticipants that collect user data alongside Facebook, including Google,

Apple, Amazon, eBay, Microsoft, AOL, Yahoo, Twitter, IAC, LinkedIn,

Adobe and Yelp and that, in addition, ‘there will continue to be a large

amount of Internet user data that are valuable for advertising purposes

and that are not within Facebook’s exclusive control’. See (Tucker and

Wellford, 2014) as well as ”Case No COMP/M.7217 - FACEBOOK/

WHATSAPP”, http://ec.europa.eu/competition/mergers/cases/ deci-

sions/m7217_20141003_20310_3962132_EN.pdf

158

Take, as an example, consumers who use mul-

tiple online social media such as Facebook,

Twitter, LinkedIn or Instagram and share broad-

ly similar information through each of them. Or,

consider access to information in the app eco-

system: Many apps, and not only those related

to location or weather, regularly ping location

data - as many as hundreds of times a week -

meaning that a user’s location is always avail-

able to a wide range of rms (Almuhimedi et al.,

2015). Of course, as we later discuss, these rms

still have to invest in ensuring that they have the

technical skills to transform this data into valu-

able insights.

Seeing that big data is not inimitable or rare,

we turn to the question of whether and when

big data is valuable for rms.

IV. Is Big Data Valuable?

Much of the current managerial literature is

focused on whether or not ‘big data’ is indeed

valuable for rms in that it enhances a rm’s abil-

ity to have protable relationships with custom-

ers (Chen et al., 2012). Cuzzocrea et al. (2011)

point to three open problems currently chal-

lenging analysts and researchers faced with

ensuring that big data is valuable to organiza-

tions. We discuss these challenges in turn and

conclude that by itself big data is not sufcient

to create prot-enhancing opportunities. The

rst challenge limiting the value of big data to

rms is compatibility and integration. One of the

key characteristics of big data is that it comes

from a ‘variety’ of sources. However, if this data

is not naturally congruent or easy to integrate,

the variety of sources can make it difcult for

rms to indeed save cost or create value for cus-

tomers. Such hindrances may prove particular-

ly burdensome in industries such as healthcare,

where prior research has shown that rms have

strategic incentives to ensure that data is siloed

and hard to integrate (Miller and Tucker, 2014).

The second challenge to making big data

valuable is its unstructured nature. As discussed

by Feldman and Sanger (2007), specialized ad-

vances are being made in mining text-based

data, where context and technique can lead

to insights similar to that of structured data, but

other forms of data such as video data are

still not easily analyzed. One example is that,

despite state-of-the-art facial recognition soft-

ware, authorities were unable to identify the

two bombing suspects for the Boston Marathon

from a multitude of video data, as the software

struggled to cope with the full-frontal nature

of the photo of their faces.

Given the chal-

lenges of unstructured data, rms tend to nd

big data most valuable when it augments the

speed and accuracy of existing data analysis

practices. In oil and gas exploration, big data is

used to enhance existing operations and data

analysis surrounding seismic drilling. However,

as emphasized by Feblowitz et al. (2013), ‘Ge-

ologists, geophysicists, and reservoir engineers

have been using massively parallel processing

capabilities of high-performance computing

(HPC) to perform analysis on petabytes (PB) of

data to inform exploration since the late 1990s.’

In other words, though big data may be a new

label for such practices, and the volume of data

may have increased, big data is valuable in oil

and gas as an extension of existing practices

and infrastructure. Most rms’ ability to analyze

the ‘variety’ of types of big data does not yet

match their ability to record its volume and

velocity.

The third challenge, and in our opinion the

most important factor that limits how valuable

big data is to rms, is the difculty of establishing

causal relationships within large pools of over-

lapping observational data. Very large data

sets usually contain a number of very similar or

virtually identical observations that can lead

to spurious correlations and as a result mis-

lead managers in their decision making. The

Economist recently pointed out that ‘in a world

14 http://www.wired.com/2013/05/boston-marathon-investi-

gation/

159

of big data the correlations surface almost by

themselves’ (Economist, 2010) and a Sloan

Management Review blog post emphasized

that while many rms have access to big data,

such data is not ’objective’,

since the difculty

lies in distilling ‘true’ actionable insights from the

data. Similarly, typical machine learning algo-

rithms used to analyze big data identify correla-

tions that may not necessarily offer causal and

therefore actionable managerial insights. Do-

mingos (2012) suggests that machine learning

algorithms should be used as a ‘guide to further

investigation’ in order that we might be able

to ‘predict the effect of our actions.’ In other

words, the skill in making big data valuable is

being able to move from mere observational

correlations to correctly identifying, potentially

outside of big data, what correlations should

form the basis for strategic action.

One well-known example of big data is Goo-

gle Trends, which uses Google’s records of ag-

gregate search queries. However, it is also an

example of a case where the fact that the data

is merely correlational limits its usefulness. But-

ler (2008) argued that this data could be

used to project the spread of u. However,

later researchers found that because the

data was backward-looking rather than for-

ward-looking, using search data only marginal-

ly improved performance relative to a ‘simple

autoregressive model’ (Goel et al., 2010).

To take a more specic example, imagine a

shoe retailer that advertises to consumers who

have previously visited their website on oth-

er websites. Raw data analysis would suggest

that customers exposed to these ads are more

likely to purchase shoes. However, these con-

sumers, who have previously visited the website

have already demonstrated their interest in

15 http://sloanreview.mit.edu/article/for-better-decision-mak-

ing-look-at-facts-not-data/

16 There were also other critiques of the usefulness of the ini-

tial predictive model: (Cook et al., 2011; Lazer et al., 2014)

the specic retailer even prior to viewing the

ad, and so are more likely than the average

consumer to purchase (Lambrecht and Tuck-

er, 2013). Was the ad effective? It is hard to

say.

Indeed, big data here does not allow any

causal inference about marketing commu-

nication effectiveness. To understand whether

such ads are effective, the retailer needs to run

a randomized experiment, where one subset

of consumers is randomly not exposed to the

ad.

By comparing the purchase probabil-

ities across consumers who were exposed to

the ad and those who were not, the company

can then determine whether exposing con-

sumers to an ad made them more likely to buy.

Value is delivered in such instances not primar-

ily by the access to data, but by the ability to

design and implement meaningful experiments.

Therefore, experiments are the main way rms

can understand whether a data relationship

is merely correlational or might be predictive

(because it is causal). Implementing eld exper-

iments, drawing the right conclusion and tak-

ing appropriate action is not necessarily easy

(Lambrecht and Tucker, 2015)

. However, suc-

cessful companies have developed the ability

to design, implement, evaluate and then act

upon meaningful eld experiments. It is this ‘test

and learn’ environment, coupled with the skill to

take action on the insights, that can make big

data valuable.

Thanks to diminishing returns to increasingly

17 This is emphasized by work such as Lewis et al. (2011) who

show that this kind of activity bias, that is the mere fact of being present

on a website signalling something about the consumer, makes the use

of non-experimental data in assessing advertising effectiveness almost

impossible.

18  Across many industries, eld experiments have widely

been used to evaluate advertising effectiveness (Lambrecht and Tucker,

2013; Draganska et al., 2014; Lambrecht et al., 2015; Lewis and Rao,

2015; Blake et al., 2015; ?)

19  https://hbr.org/2015/11/run-eld-experiments-to-make-

sense-of-your-big-data?utm_campaign=HBR&utm_source=face-

book&utm_medium=social

20 Note that even when using insights from experiments, man-

agersneedtocarefullyconsiderthescopeofanyndingsandhowrep-

licable they will be in different contexts (Ioannidis, 2005).

160

large data samples, such experimentation does

not necessarily require big data. For example,

Google reports that it typically uses random

samples of 0.1% of available data to perform

analyses (Varian, 2014). Indeed, a recent article

suggested that the size of big data can actually

be detrimental as ‘the bigger the database, the

easier it is to get support for any hypothesis you

put forward’.

In other words, because big data

often offers overlapping insights, a rm can get

similar insight from one-thousandth of the full

dataset as from the entire dataset.

Experimentation is not the only method com-

panies can use to infer valuable insights from

big data. Another potential skill rms can devel-

op is the ability to build better algorithms to deal

with big data. One example for such algorithms

is recommender systems. Recommender sys-

tems rely on algorithms trained on correlational

data to recommend the most relevant prod-

ucts to a customer. Yet, again, it is not the size

of the underlying data, but the ability to identify

the critical pieces of information that best pre-

dict a customer’s preferences. For example, it

has been shown that to predict preferences for

movies, ten movie ratings alone are more help-

ful than extensive metadata (Pilászy and Tikk,

2009). Indeed, often not the size of the data

but the machine learning algorithm used deter-

mines the quality of the results.

While predictive

power may increase with the size of the data

available, in many instances the improvements

in predictions show diminishing returns to scale

(Junqué de Fortuny et al., 2013).

Our analysis demonstrates that, by itself, big

data is unlikely to be valuable. It is only when

combined with managerial, engineering, and

21 https://www.london.edu/faculty-and-research/lbsr/diie-nov-

drowning-in-numbers#.Vk-OZvmrRNO

22 http://www.slideshare.net/xamat/10-lessons-learned-from-

building-machine-learning-systems,http://stackoverflow.com/ques-

tions/25665017/does-the-dataset-size-inuence-a-machine-learning-

algorithm

analytic skill in determining the experiment or

algo- rithm to apply to such data that it proves

valuable to rms.

This suggests that for rms, the

primary challenges lie in determining a big data

strategy,

implementing the systems and tools

to analyze the data

and adapting organiza-

tional capabilities (McAfee et al., 2012; Bughin

et al., 2010).

Given that our previous analyses suggest that

big data is neither rare nor inimitable, we con-

clude that the search for competitive advan-

tage in the new digital economy should focus

on attracting the kind of skilled workers who are

able to transform big data into valuable tools.

V. Is Big Data Non-Substitutable?

For a resource such as big data to provide a

sustainable competitive advantage, there has

to be no other means of achieving success in

the specic industry. Yet, in the digital world,

perhaps more so than ofine, there are many

examples of rms that came from nowhere and,

without any embedded data advantage, were

still able to disrupt an industry and attract more

customers because of a superior value prop-

osition. In this section, we discuss ve settings

where alternative rm capabilities have proved

to be compelling substitutes to big data and

consequently where big data has not been a

sufciently sustainable source of competitive

advantage.

First, it is natural to focus on an industry where

23 One potential way of evaluating whether this insights holds

inaspeciccontextistoexamine thepricingofdatarelativetorm

processing skills. Data being very cheap relative to processing skills

suggests that processing skills are more important than data in creating

valueforarm.

24 http://www.cio.com/article/2395010/data-management/the-

big-data-challenge--how-to-develop-a-

winning-strategy.html

25 http://sloanreview.mit.edu/article/overcoming-legacy-pro-

cesses-to-achieve-big-data-success/

161

data has, even before the internet, offered op-

erational advantages. The communications in-

dustry has long used large data sets to both im-

prove operations (e.g. data network ow) and

offer better value to customers (e.g. through

pricing plans that meet customer needs, see for

example Lambrecht and Skiera (2006); Ascar-

za et al. (2012)). Many traditional commu-

nications rms such as AT&T and Verizon as

well as newer online rms such as Skype and

Facebook have large data sets covering mes-

saging services. Despite this, the messaging

app WhatsApp became a serious competitor

to established messaging and social network

services by offering a product that satised so-

cial media users’ latent needs - an easy-to-use

interface and an extremely low-cost messaging

solution. Even when acquired by Facebook for

$22 billion, WhatsApp had only 55 employees,

suggesting its success was not due to large-

scale data analytics capacity.

A similar exam-

ple is Snapchat, which succeeded in com-

peting in this space without access to big

data because of its insight that people wanted

to share personal information more privately.

Another industry where big data could pro-

vide insights into consumer preferences and

therefore give advantages to large digital rms

when launching new products, is online gaming.

Yet, King Digital Entertainment was not among

the dominant digital gaming companies, nor

supported by rms with access to big data such

as Google and Facebook, when it launched the

smartphone hit Candy Crush Saga. By 2014, 93

million people played Candy Crush Saga more

than 1 billion times a day.

The fact that Candy

Crush is playable in short sessions and does not

26 http://www.forbes.com/sites/parmyolson/2014/10/06/face-

book-closes-19-billion-whatsapp-deal/,

http://www.businessinsider.com/why-facebook-buying-whatsapp-

2014-2?IR=T, http://www.bloomberg.com/ news/articles/2014-10-28/

facebook-s-22-billion-whatsapp-deal-buys-10-million-in-sales

27 http://www.theguardian.com/technology/2014/mar/26/can-

dy-crush-saga-king-why-popular, https:

//thinkgaming.com/app-sales-data/2/candy-crush-saga/ While Candy

Crush Saga is free to download and play, it makes its money from in-

app purchases of extra moves, lives and power-ups, with estimated daily

revenues of over $700,000, as of November 23, 2015

require extensive time investment explains its

appeal to the non-gaming population of time-

strapped parents, or commuters, ’from ofce ju-

niors through to CEOs’.

This example illustrates

that a superior value proposition to a new group

of consumers can be more important than ac-

cess to data, even in a sector where companies

routinely have access to big data.

Second, it is natural to ask whether there is

a substitute for insights from big data in sectors

where there has historically been little use of

data. It is possible that in such contexts, rms in

adjacent sectors who do have big data have

an executional advantage in terms of mod-

ernizing these sectors. However, the rise of the

new ‘sharing economy’ provides evidence that

to build up entirely new digital industries in tra-

ditional sectors does not require access to big

data. Uber and Lyft had no superior access

to data compared to established taxi services,

but they were better at putting together a prod-

uct that met consumer needs for a convenient

and reliable taxi service. AirBnB entered a highly

competitive industry where large travel compa-

nies have access to large swathes of data and

regularly run experiments to interpret their data

in a meaningful way to constantly improve busi-

ness practices. Yet, despite the lack of data,

AirBnB quickly became a dominant player be-

cause of its superior value proposition. Google’s

purchase of ITA along with its ight data and

data-processing capabilities did not give Goo-

gle a signicant presence in the ight search

market. This contrasts with the growth of Kayak

- a travel search engine - which grew from 2004

from a small start up with no user data to be-

ing acquired in 2012 by Priceline for $1.8 billion.

Indeed, recent spectators have argued that

for the sharing economy the secret sauce is not

data by itself, but instead the systems that such

platforms build around ensuring there is ‘trust

28 http://www.theguardian.com/technology/2014/mar/26/can-

dy-crush-saga-king-why-popular

29 http://thenextweb.com/insider/2012/11/08/priceline-com-ac-

quiring-travel-company-kayak-for-1-8b-in-cash-and-stocks/

162

and reputation’ among users of the platform.

Third, industries where data is important for

delivering a personalized experience, and

where this personalized system of recommen-

dations is particularly important for customer ex-

perience, may be another natural setting where

big data might have few substitutes. One obvi-

ous example of such an industry is online dating,

where the difculty of predicting human rela-

tionships likely puts a premium on the availability

of large data sets. However, Tinder entered the

online dating market in September 2012 with no

access to existing data and quickly became a

dominant player with 1.6 billion Tinder proles,

making more than 26 million matches per day

(as of April 2015). More than 8 billion matches

have been made since Tinder launched.

Tin-

der succeeded not because of big data but

because it offers a better solution for its market.

Critically, this included a simple user interface

that does not require long surveys and allowed

users to express interest using a simple game-like

‘swipe right’ and a ‘double opt-in’ for matches,

where both users must agree before they can

message each other. To build up its user base,

Tinder did not advertise or use mass emails

based on big data bases, but instead hosted

‘exclusive’ parties on college campuses with

admittance based on having downloaded the

app.

Fourth, another natural place to look for

non-substitutability is industries with switching

costs and network effects. Switching costs are

the costs (both perceived and real) incurred by

customers when they switch brands or suppli-

ers. Network effects occur when the usefulness

of a product, service or platform increases as

more people use it. Historically, switching costs

and network effects have been highlighted by

30 http://sloanreview.mit.edu/article/data-at-the-heart-of-the-

sharing-economy/?utm_source=facebook&utm_medium=social&utm_

campaign=sm-direct

31 https://en.wikipedia.org/wiki/Tinder_(app)

32 https://www.quora.com/How-did-Tinder-grow-so-quickly

economists as potential sources of incumbent

competitive advantage, especially in digital

environments (Farrell and Klemperer, 2007).

Therefore it is natural to ask whether big data, in

combination with switching costs and network

effects, might lead to a setting where poten-

tial rivals struggle to compete or nd sufcient

substitutes to compete with. Social network sites

exhibit both potential network effects, because

consumers value being able to communicate

with their friends, and switching costs, as cus-

tomers invest time and money in curating their

online proles.

However, the history of social networking sites

suggests that big data has not protected larg-

er rms in this industry (Tucker and Marthews,

2011). Rather, this industry has experienced a

succession of large rms, even though at each

point in time the incumbent had access to big

data whereas the new entrant was, in terms of

data availability, at a disadvantage. For exam-

ple, MySpace replaced Friendster and was then

replaced by Facebook as the leading social

network site. What ultimately made Facebook

successful was the ability to build a product that

was more focused on customer needs. This in-

cluded giving customers more control over their

social media interactions. For example, Face-

book allowed users more control over what

content observers could see about a user, rela-

tive to the public nature of MySpace. MySpace

was seen by many as too cluttered, and Face-

book offered a much cleaner design.

Fifth, one potential way that big data could

be non-substitutable is if it is necessary for at-

tracting capital investment. However, it is nota-

ble that venture capital does not view big data

as ‘non-substitutable’, in that it continues to fund

startups to compete in spaces where other rms

33 Decisions on the size, quality and placement of ads on MyS-

pacewerelessinuencedbyneedsoftheusersandmorebytheimper-

ative to monetize the site, leading to an even more ad-cluttered site. For

a comprehensive account of what happened to MySpace, see http://

www.bloomberg.com/bw/magazine/content/11_27/b4235053917570.

htm#p3

163

are demonstrably in possession of ‘big data’. For

example, despite ‘Amazon Fresh’ and ‘Google

Express’ having access through their parent

companies to big data about potential custom-

ers, there is vibrant funding of new startups that

are trying to compete in the local delivery

space who do not have this data advantage.

For example, Instacart has received $275M in

funding

, Jet has received $220M in fund-

ing,

and Postmates has received $138M in

venture capital funding.

Overall, big data is not a non-substitutable

requirement for offering online services, though

ownership of big data is often the natural con-

sequence of being successful in offering such

online services. Instead, in a similar manner to

the ofine world, what determines success on-

line is a superior ability to understand and meet

customer needs. The unstable history of digital

business offers little evidence that the mere pos-

session of big data is sufcient protection for an

incumbent against a superior product offering.

VI. Implications

Can big data confer a sustainable competi-

tive advantage for rms, which can help them

persistently deect current and future competi-

tion? To analyze whether big data can act as a

barrier of entry in this manner, we use the classic

resource-based view of strategic management,

which emphasizes that to qualify as a sustain-

able competitive advantage a resource needs

to meet four criteria. It has to be inimitable,

rare, valuable and non-substitutable. For a wide

range of examples from the digital economy

we demonstrate that when rms have access

to big data, at least one, and often more, of the

34 https://www.crunchbase.com/organization/instacart\#/entity

35 https://www.crunchbase.com/organization/jet\#/entity

36 https://www.crunchbase.com/organization/postmates\#/entity

four criteria which are required for a resource

to constitute a sustainable competitive advan-

tage are not met.

Our aim is not to suggest that rms cannot

derive benets from owning and evaluating big

data. Instead, we highlight that the simple act

of amassing big data by itself does not confer

a long-term competitive advantage. We con-

clude that to build up a competitive advan-

tage related to big data rms need to develop

two new competencies.

First, rms need to attract employees who

have the ability to develop and train algorithms

or to design and/or to set up and run meaning-

ful experiments, since it is insights from such ef-

forts that may be able to turn big data into a

meaningful competitive advantage. This builds

on earlier work such as Porter and Millar (1985)

who argued that information technology can

confer a competitive advantage but that the

simple presence of data is not sufcient for com-

petitive success. Instead rms needs to develop

complementary organizational skills.

Second, rms need to use big data to look

forward and understand evolving customer

needs rather than simply use past historic big

data to make incremental improvements to

their current product offering or service. The un-

stable history of digital business offers little evi-

dence that the mere possession of big data is a

sufcient protection for an incumbent against a

superior product offering. To build a sustainable

competitive advantage, the focus of a digital

strategy should therefore be on how to use dig-

ital technologies to provide value to customers

in ways that were previously impossible.

In addition to our managerial implications

this paper also contributes to a policy literature.

This literature is concerned with the question

whether big data can constitute a barrier of en-

164

try which is in a sense the ipside of the ques-

tion we focus on whether big data constitutes

a competitive advantage. In contrast to this

largely legal literature, which grapples with how

to frame big data in the context of traditional

antitrust analysis (Stucke et al., 2015; Grunes and

Stucke, 2015; Tucker and Wellford, 2014), we use

a long-established strategic framework to eval-

uate whether big data indeed merits consid-

eration as a source of sustainable competitive

advantage

References

Almuhimedi, H., F. Schaub, N. Sadeh, I. Ad-

jerid, A. Acquisti, J. Gluck, L. F. Cranor, and Y.

Agar- wal (2015). Your location has been shared

5,398 times!: A eld study on mobile app priva-

cy nudging. In Proceedings of the 33rd Annual

ACM Conference on Human Factors in Com-

puting Systems, pp. 787–796. ACM.

Ascarza, E., A. Lambrecht, and N. Vilcassim

(2012). When talk is free: The effect of tariff struc-

ture on usage under two-and three-part tariffs.

Journal of Marketing Research 49 (6), 882–899.

Barney, J. (1991). Firm resources and sus-

tained competitive advantage. Journal of

manage- ment 17 (1), 99–120.

Barney, J. B. (2001). Resource-based the-

ories of competitive advantage: A ten-year

retrospective on the resource-based view. Jour-

nal of management 27 (6), 643–650.

Biesdorf, S., D. Court, and P. Willmott (2013).

Big data: What’s your plan. McKinsey Quarterly

2, 40–51.

Blake, T., C. Nosko, and S. Tadelis (2015). Cus-

tomer heterogeneity and paid search effective-

ness: A large scale eld experiment. Economet-

rica forthcoming .

Bughin, J., M. Chui, and J. Manyika (2010).

Clouds, big data, and smart assets: Ten tech-en-

abled business trends to watch. McKinsey Quar-

terly 56 (1), 75–86.

Butler, D. (2008). Web data predict u. Na-

ture News 456 (7220), 287–288.

Calandrino, J., A. Kilzer, A. Narayanan, E.

Felten, and V. Shmatikov (2011, May). You might

also like: Privacy risks of collaborative ltering. In

Security and Privacy (SP), 2011 IEEE Symposium

on, pp. 231–246.

Chen, H., R. H. Chiang, and V. C. Storey

(2012). Business intelligence and analytics:

From big data to big impact. MIS quarterly 36

(4), 1165–1188.

Cook, S., C. Conrad, A. L. Fowlkes, and M. H.

Mohebbi (2011). Assessing google u trends per-

formance in the united states during the 2009

inuenza virus a H1N1 pandemic. PloS one 6 (8),

e23610.

Cuzzocrea, A., I.-Y. Song, and K. C. Davis

(2011). Analytics over large-scale multidimen-

sional data: The big data revolution! In Pro-

ceedings of the ACM 14th International Work-

shop on Data Warehousing and OLAP, DOLAP

’11, New York, NY, USA, pp. 101–104. ACM.

Domingos, P. (2012, October). A few useful

things to know about machine learning. Com-

mun. ACM 55 (10), 78–87.

Draganska, M., W. R. Hartmann, and G. Stan-

glein (2014). Internet versus television advertis-

ing: A brand-building comparison. Journal of

Marketing Research 51 (5), 578–590.

Economist, T. (2010). Data, data everywhere.

The Economist Newspaper Limited . Edelman,

B. (2015, April). How to launch your digital plat-

form. Harvard Business Review .

165

Farrell, J. and P. Klemperer (2007). Coordi-

nation and lock-in: Competition with switching

costs and network effects. Handbook of indus-

trial organization 3, 1967–2072.

Feblowitz, J. et al. (2013). Analytics in oil and gas:

The big deal about big data. In SPE Digital Energy

Conference. Society of Petroleum Engineers.

Feldman, R. and J. Sanger (2007). The text

mining handbook: advanced approaches in

analyzing unstructured data. Cambridge Uni-

versity Press.

Goel, S., J. M. Hofman, S. Lahaie, D. M. Pen-

nock, and D. J. Watts (2010). Predicting con-

sumer behavior with web search. Proceedings

of the National academy of sciences 107 (41),

17486–17490.

Greenstein, S. M., A. Goldfarb, and C. Tucker

(2013). The Economics of Digitization. Edward

Elgar.

Grunes, A. P. and M. E. Stucke (2015). No mis-

take about it: The important role of antitrust in

the era of big data. Antitrust Source (Apr. 2015).

Ioannidis, J. P. (2005). Why most published re-

search ndings are false. Chance 18 (4), 40–47.

Junqué de Fortuny, E., D. Martens, and F. Pro-

vost (2013). Predictive modeling with big data:

Is bigger really better? Big Data 1 (4), 215–226.

Lambrecht, A. and B. Skiera (2006). Paying

too much and being happy about it: Existence,

causes, and consequences of tariff-choice bi-

ases. Journal of Marketing Research 43 (2), 212–

223.

Lambrecht, A. and C. Tucker (2013). When

does retargeting work? Information specicity

in online advertising. Journal of Marketing Re-

search 50 (5), 561–576.

Lambrecht, A. and C. Tucker (2015). Field

experiments in marketing. Working paper (Sep-

tember 10, 2015).

Lambrecht, A., C. Tucker, and C. Wiertz

(2015). Advertising to early trend propagators:

Evidence from twitter. Working paper (March

18, 2015).

Lazer, D., R. Kennedy, G. King, and A. Vespig-

nani (2014). The parable of Google Flu: Traps in

big data analysis. Science 343 (14 March).

Lewis, R. A., J. Rao, and D. Reiley (2011). Here,

there, everywhere: Correlated online behaviors

can lead to overestimates of the effects of ad-

vertising. In International World Wide Web Con-

ference.

Lewis, R. A. and J. M. Rao (2015).

The unfavorable economics of measuring the

returns to advertising*. The Quarterly Journal of

Economics .

Markman, G. D., M. I. Espina, and P. H. Phan

(2004). Patents as surrogates for inimitable and

non-substitutable resources. Journal of Man-

agement 30 (4), 529–544.

McAfee, A., E. Brynjolfsson, T. H. Davenport, D.

Patil, and D. Barton (2012). Big data. The man-

agement revolution. Harvard Bus Rev 90 (10),

61–67.

McGuire, T., J. Manyika, and M. Chui (2012).

Why big data is the new competitive advan-

tage. Ivey Business Journal 76 (4), 1–4.

Miller, A. R. and C. Tucker (2014). Health

information exchange, system size and in-

formation silos. Journal of Health Economics 33,

28 – 42.

Narayanan, A. and V. Shmatikov (2008). Ro-

bust de-anonymization of large sparse datasets.

166

In Security and Privacy, 2008. SP 2008. IEEE Sym-

posium on, pp. 111–125. IEEE.

Peteraf, M. A. (1993). The cornerstones of

competitive advantage: A resource-based

view. Strategic management journal 14 (3),

179–191.

Pilászy, I. and D. Tikk (2009). Recommending

new movies: even a few ratings are more valu-

able than metadata. In Proceedings of the third

ACM conference on Recommender systems,

pp. 93– 100. ACM.

Porter, M. E. and V. E. Millar (1985). How in-

formation gives you competitive advantage.

Shapiro, C. and H. R. Varian (1999). Information

rules: a strategic guide to the network econo-

my. Harvard Business Press.

Stucke, M., A. Grunes, et al. (2015). Debunk-

ing the myths over big data and antitrust. Anti-

trust Chronicle 5.

Tucker, C. and A. Marthews (2011). Social net-

works, advertising, and antitrust. Geo. Mason L.

Rev. 19, 1211.

Tucker, D. S. and H. B. Wellford (2014). Big

mistakes regarding big data. Antitrust Source,

American Bar Association.

Varian, H. R. (2014). Big data: New tricks for

econometrics. The Journal of Economic Per-

spectives , 3–27.

Wernerfelt, B. (1984). A resource-based view

of the rm. Strategic management journal 5 (2),

171–180.

167