Designing Competitive Markets for Industrial Data - Between Propertisation and Access

AuthorJosef Drexl
Pages257-292
Designing Competitive Markets for Industrial Data
2017
257
4
Designing Competitive Markets
for Industrial Data
Between Propertisation and Access
by Josef Drexl*
© 2017 Josef Drexl
Everybody may disseminate this ar ticle by electronic m eans and make it available for downloa d under the terms and
conditions of the Digital P eer Publishing Licence (DPPL). A copy of the license text may be obtain ed at http://nbn-resolving.
de/urn:nbn:de:0009-dppl-v3-en8.
Recommended citation: Jos ef Drexl, Designing Competiti ve Markets for Industrial Data – Bet ween Propertisation and A ccess,
8 (2017) JIPITEC 257 para 1.
Keywords: Data ownership; access to data; data sharing; data economy; data-driven economy; Internet of
Things; data analytics; database rights; trade secrets protection; EU competition law; refusal to
license; essential facilities; data portability
make the data economy work? Do we need new own-
ership rights in data? Or should regulation focus on
access in order to make data as widely available as
possible? The European Commission is currently try-
ing to formulate answers to these questions. This ar-
ticle aims to assist the Commission by working on a
pro-competitive framework for issues of both own-
ership and access. In so doing, this article undertakes
two things: first, it analyses to what extent intellec-
tual property laws already provide control over data
and then discusses the need and justification for in-
troducing new rules on data ownership. Second, it
analyses whether EU competition law already pro-
vides remedies to promote access to data, and fur-
thermore explores whether and under which condi-
tions the introduction of new access regimes would
be advisable. This article is to be considered as on-
going research. It does not yet take into account
more recent developments in 2017.
Abstract: As part of the project to establish a
Digital Single Market, the European Commission has
launched a ‘Free Flow of Data’ initiative. This initia-
tive is meant to enhance the growth potential of the
emerging data economy, which is characterised by
the digitisation of production (smart factories) and
the advent of digitised products such as smart—
driverless—cars, or smart wearables that will be able
to communicate with each other and the environ-
ment through the Internet of Things. Furthermore,
the enormous amount of data generated and con-
trolled by the industry could serve as a most valu-
able input for other new data-driven services and for
applications in the public interest, such as the oper-
ation of smart cities, smart and resource-efficient
farming, or measures to prevent the spread of in-
fectious diseases. Obviously, this new data econ-
omy has to rely on the commercialisation of data.
But what kind of regulation is needed in order to
2017
Josef Drexl
258
4
A. Introduction
1
The advent of the data economy and the Internet
of Things (IoT) is currently challenging regulators
across the globe. Buzzwords such as ‘big data’ or
‘data as the oil of the modern economy’ are used
everywhere, and questions like ‘Who owns the data?’
are not only asked by the media, but are also heard
and taken up by decision-makers in the political
arena.
2
In the EU, potential new regulation for the data
economy, concerning both data ownership and
access to data, is part of the Commission’s current
priority project to implement a Digital Single
Market.1 In May 2015, the Commission identied 16
key actions for the implementation of this Digital
Single Market,2 including the ‘building of a data
* Dr iur (Munich), LLM (UC Berkeley), Director of the Max
Planck Institute for Innovation and Competition, Munich,
Honorary Professor at the University of Munich.
This article complements the Position Statement of
the Max Planck Institute: Josef Drexl, Reto M Hilty, Luc
Desaunettes, Franziska Greiner, Daria Kim, Heiko Richter
and Gintarė Surblytė, ‘Data Ownership and Access to Data’
(16 August 2016), available at:
en/link/positionpaper-data-2016-08-16.html> (accessed 10
September 2016). The views expressed in this article are
however only those of its author.
This article was rst made available online as Research
Paper No. 16-13 of the Max Planck Institute for Innovation
and Competition Research Paper series on 8 November
2016. The text remains substantially unchanged. It does not
take into account the debate following the EU Commission’s
Communication of 10 January 2017 on Building a European
Data Economy, COM(2017) 9 nal. On this Communication
see the Position Statement of the Max Planck Institute: Josef
Drexl, Reto M. Hilty, Jure Globocnik, Franziska Greiner, Daria
Kim, Heiko Richter, Peter R. Slowinski, Gintare Surblyte,
Axel Walz and Klaus Wiedemann, ‘On the European
Commission’s Public Consultation on “Building a European
Data Economy”’ (26 April 2017) available at:
ip.mpg.de/leadmin/ipmpg/content/stellungnahmen/
MPI_Statement_Public_consultation_on_Building_the_EU_
Data_Eco_28042017.pdf> (accessed 19 October 2017). On
this, see also Josef Drexl, ‘On the Future Legal Framework
for the Digital Economy: A Competition-based Response to
the “Ownership and Access” Debate’, in Reiner Schulze and
Dirk Staudenmayer (eds), Trading Data in the Digital Economy:
Legal Concepts and Tools (Baden-Baden: Nomos, forthcoming)
222-43.
1 Implementation of the Digital Single Market is one of four
‘priority projects’ of the current European Commission
under the aegis of President Jean-Claude Juncker. See
Jean-Claude Juncker, ‘My priorities’, available at:
juncker.epp.eu/sites/default/les/attachments/nodes/
en_01_main.pdf> (accessed 10 September 2016).
2 See Communication of the Commission of 6 May 2015 to the
European Parliament, the Council, the European Economic
and Social Committee and the Committee of the Regions—A
Digital Single Market Strategy for Europe, COM(2015)
192 nal. See also European Commission, ‘A Digital Single
Market for Europe: Commission sets out 16 initiatives to
make it happen’, Press Release of 6 May 2015, available
economy’. This ‘action’ is supposed to contribute to
the third pillar of the Digital Single Market project,
aiming at ‘maximising the growth potential of the
digital economy’.3 More concretely, the Commission
announced a ‘Free Flow of Data’ initiative for 2016,
which would address in particular the restrictions
on the free movement of data beyond the protection
of personal data with the objective of enhancing
the cross-border use of data in a world of big data
and the Internet of Things. Yet the initiative also
includes a mandate to look at the issue of ownership.
The announcement reads as follows:
The Commission will propose in 2016 a European ‘Free ow of
data’ initiative that tackles restrictions on the free movement
of data for reasons other than the protection of personal data
within the EU and unjustied restrictions on the location of
data for storage or processing purposes. It will address the
emerging issues of ownership, interoperability, usability and
access to data in situations such as business-to-business,
business to consumer, machine generated and machine-to-
machine data. It will encourage access to public data to help
drive innovation.4
3
As regards ownership, the mandate does not indicate
the direction in which later regulatory actions may
ultimately go. In the light of the objective to promote
access to data, one could expect the Commission to
consider whether existing ‘ownership’ regimes are
in need of additional exceptions and limitations
to promote access. This would have been in line
with the debate in other fora, such as OECD in
particular, where a study of 2015 highlighted the
need to promote access to big data in order to
generate maximum benets for society.5 Rather
than taking data ownership as the starting point
of the regulation of the data economy, the OECD
study recommends developing and improving ‘data
governance regimes’ that ‘overcome … barriers to
data access, sharing and operability’.6
4
As regards the EU, however, the debate quickly
shifted direction. While the responsibility to work
on the initiative was allocated to the Digital Value
Chain unit of DG CONNECT, which is also responsible
for the open data policy of the EU as regards public
sector information in particular, it was the German
Commissioner Günther Oettinger responsible for DG
CONNECT who publicly contributed to the impression
that the Commission would soon propose legislation
at:
en.htm> (accessed 10 September 2016).
3 See Chapter 4.1 of the Commission Communication (supra n
2) at 14-15.
4 Ibid, at 15. (Emphasis added.)
5 OECD, ‘Data-Driven Innovation: Big Data for Growth and
Well-Being’ (2015) 195-98, available at:
org/sti/data-driven-innovation-9789264229358-en.htm>
(accessed 10 September 2016).
6 Ibid, at 195-99 (in particular at 198).
Designing Competitive Markets for Industrial Data
2017
259
4
on a new ‘data use right’ (Datennutzungsrecht).7
5
The data economy and its regulation attract particular
attention in Germany, where the industry is deeply
involved in the development of new business models
of the Internet of Things. In Germany, in 2011, the
‘Industrie 4.0’ initiative was launched as a joint
initiative of the government, the private business
sector and the public research sector to manage and
promote a fourth industrial revolution characterised
by the integration of manufacturing in modern
information and telecommunications networks,
including the Internet of Things.8 This initiative
not only aims at optimising the manufacturing
process, whereby the product itself, in the various
production phases, communicates with, and steers,
the production process. It also targets the logistics
sector, aiming to foster an ‘Internet of Services’
that builds on smart products as a basis for new
kinds of services provided to consumers. This early
initiative may also explain why, in Germany, legal
regulation of the industrial dimension of the data-
driven economy, namely, beyond the issues of
7 See, for instance, ‘Oettinger: Versicherungen brauchen
mehr digitale Produkte’, Der Standard (25 November 2015),
available at:
Oettinger-Versicherungen-brauchen-mehr-digitale-
Produkte> (accessed 20 May 2016) (reporting on a talk
by the Commissioner at a conference of the German
insurance industry association in November 2015 where
the Commissioner called upon the insurance industry to
take part in the discussion on such a right). See also the
association’s website: ‘Versicherungstag 2015: Es geht mehr
denn je um den Kunden’ (25 November 2015), available
at:
chancen-der-digitalen-welt/> (accessed 10 September
2016). The author of this paper personally attended
another talk given by the Commissioner at a conference
of the Forschungsinstitut für Wirtschaftsverfassung und
Wettbewerb (FIW) in Innsbruck on 25 February 2016, where
the Commissioner made similar statements. See ‘Rede
(Kommissar Oettinger) auf dem 49. FIW-Symposion (2016) in
Innsbruck zur Digitalisierung’ (25 February 2016), available
at:
rede-kommissar-oettinger-auf-dem-49.-w-symposion-
2016-in-innsbruck-zur-digitalisierung> (accessed 10
September 2016) (reporting on the Commissioner asking
who owns the data that are produced by modern cars
in a world of the Internet of Things). In a more recent
speech at the occasion of a Commission conference on the
‘Free Flow of Data’ initiative, however, the Commissioner
did not repeat this claim for a data usage right. See
Günther Oettinger, Speech at the Conference ‘Building
European Data Economy’ (17 October 2016), available at:
announcements/speech-conference-building-european-
data-economy_en> (accessed 30 October 2016).
8 See the public announcement of the initiative made on
the occasion of the 2011 Hanover trade fair: Henning
Kagermann, Wolf-Dieter Lukas and Wolfgang Wahlster,
‘Industrie 4.0: Mit dem Internet der Dinge auf dem Weg
zur 4. industriellen Revolution’ (1 April 2011), available at:
Industrie-40-Mit-Internet-Dinge-Weg-4-industriellen-
Revolution> (accessed 10 September 2016).
protection of personal data in particular, attracted
attention much earlier than in other parts of the
EU, both from the academic community9 and from
the stakeholders’ side. As regards the latter, the
Bundesverband der Deutschen Industrie (BDI, German
Industry Association) published a study on the legal
ramications of the data-driven economy that, inter
alia, argued against the introduction of a new right of
data ownership.10 A report of the Bavarian Industry
Association (Vereinigung der Bayerischen Wirtschaft)
argued that ownership for single pieces of data and
small datasets could lead to a scarcity of data and
distort innovation through big data analytics.11
6
Indeed, scepticism about introducing a new
intellectual property right expressed by the industry
that is expected to rely on this right for protecting
its own investments is not something that experts
in intellectual property law would necessarily
expect. However, the same scepticism was voiced
by the representatives of the ‘Industry 4.0’ sector
who were invited to a hearing of DG CONNECT on
the ‘legal regime t for an efcient and fair access
to and usage and exchange of data’ in Luxembourg
on 17 March 2016.12 The hearing concentrated
on the legal protection of the investment in data
collection capabilities and the exploitation of the
value represented by that data. The hearing was not
least held for the purpose of learning more about
the legal instruments that are used and needed to
9 More in favour of such a right Herbert Zech, ‘Daten als
Wirtschaftsgut—Überlegungen zu einem “Recht des
Datenerzeugers”’ (2015) Computer und Recht 137; most
recently see Herbert Zech, ‘A legal framework for a data
economy in the European Digital Single Market: rights to
use data’ (2016) 11 J Int Prop L & Prac 460; Herbert Zech, ‘Data
as tradeable commodity’ in Alberto de Franceschi (ed.),
European Contract Law and the Digital Single Market (Insentia:
2016 forthcoming) 51; against such a right Michael Dorner,
‘Big Data und “Dateneigentum”, Grundfragen des modernen
Daten- und Informationshandels’ (2014) Computer und Recht
617. See also Alberto De Franceschi and Michael Lehmann,
‘Data as Tradable Commodity and New Measures for their
Protection’ (2015) 51 Italian LJ 51 (seemingly supporting the
recognition of a ‘data usage right’).
10 Konrad Żdanowiecki, ‘Recht an Daten’ in Peter Bräutigam
and Thomas Klindt (eds), Digitalisierte Wirtschaft/Industrie
4.0, Study of Noerr LLP for BDI (November 2015) 18-
28, available at:
digitalisierung/downloads/20151117_Digitalisierte_
Wirtschaft_Industrie_40_Gutachten_der_Noerr_LLP.pdf>
(accessed 10 September 2016).
11 Zukunftsrat der Bayerischen Wirtschaft, ‘Zukunft digital—
Big Data: Analyse und Handlungsempfehlungen (July 2016)
at 99, available at:
big_data/vbw_zukunftsrat_handlungsempfehlungen_
langfassung_v15_rz_web.pdf> (accessed 10 September
2016).
12 The author of this paper took part in this ‘Round Table’ as a
representative from the academic community. The results
of the event are documented in a synthesis report not
publicly available of the Unit for Data Value Chain (available
from the author).
2017
Josef Drexl
260
4
implement new business models based on big data.
Unanimously, the industry participants stressed that
they were able to implement their business models
involving data-sharing by relying on contract law.
‘Ownership’ was even considered a concept that does
not t the needs of the data economy; introduction
of a new right was seen as a form of government
intervention that needs to be avoided. At the same
time the need to promote access, with a potential role
of competition law, was discussed. Ultimately, the
Digital Value Chain Unit’s representative indicated
that the Commission would come forward with
policy conclusions in the form of a Communication,
which was published in January 2017.13
7 There seem to be two obvious, yet related, reasons
why the industry rejects the introduction of new
property rights for data: rst, many rms are
producers of data and have to rely on access to
data of other players at the same time. Hence, it
is not clear to them whether the introduction of
new rights would provide them with more benets
than drawbacks. Second, the criteria on who would
qualify as the owner of the new right are not at all
clear. Many stakeholders, in one way or another,
contribute to the same data-based business model
and may have very diverse kinds of interests.
Therefore, allocation of data ownership is indeed
a major issue.14 This is also an issue of considerable
complexity because of the particularities of the
specic sectors. The interests of stakeholders
regarding the data collected by the sensors of a car,
in which public authorities also have an interest, so
as to protect the environment or to increase driving
safety, are likely to be different than those in the
case of health-related data derived from blood tests
of patients for which a patented diagnostic tool is
used, which, taken together with similar data from
other labs, may help authorities around the globe to
ght the spread of infectious diseases. The difcult
question to whom the new data ownership should
be allocated led the BDI study to conclude that the
legislature should refrain from creating such a right
from the outset.15 In such a situation it should not
come as a surprise that rms, which cannot foresee,
and do not have any legitimate expectation, that
they will be recognised as owners of data rights, will
13 European Commission, Communication from the Commission
to the European Parliament, the Council, the European Economic
and Social Committee and the Committee of The Regions Building
A European Data Economy, COM (2017) 9 nal, Bruxelles,
10.1.2017. As mentioned, this article does not yet discuss
this Communication. For further references see at n * above.
14 The OCED (supra n 5) at 196, lists ten different kinds of
stakeholders. It thereby relied on literature—David Loshin,
‘Knowledge integrity: Data ownership’ (2002) (no longer
available on the Internet)—that predates the big data
debate and, in particular, does not yet take account of big
data analyses and big data brokerage.
15 Żdanowiecki (supra n 10) at 28.
be hesitant to support any additional legislation. If
it was accepted that there should be ownership of
everybody to whom specic data can be allocated,
the result would be multiple ownership of the same
data with considerable negative effects on access to
that data.16
8
This article aims to produce additional insights
on how the data economy should be regulated as
regards data collected by the industry. Ideally, it
will also assist the European Commission in its task
of designing its regulatory approach to promoting
the data economy in the interest of society. For that
purpose, the article looks at the issues of both data
ownership and access to data.
9
As a starting point, this article argues that the
question ‘Who owns the data?’ is fundamentally
misguided. This is so for two reasons: rst, it skips
the prior question of whether there is a need to
recognise any ownership. There is no natural law
that says that data as an asset, although it may have
economic value, has to be owned by anybody. Rather,
recognition of any new right should, as is the case
in intellectual property in general, be considered
a form of government regulation of the market,
which is in need of a particular justication. In
terms of data ownership, which enables its owner
to commercialise data, this justication needs to be
an economic one.17
10
The second reason is that identifying the owner
does not resolve all issues of ownership. In the eld
of intellectual property law, the legislature has to
decide upon a series of issues: rst and foremost, the
subject-matter of protection has to be determined.
Hence, the law would have to clarify what is meant
by ‘data’ in the context of ‘data ownership’. And
then there is the issue of ‘how’ ownership should
be protected. In other words, the legislature has to
decide on the scope of protection—namely, what
kind of interests and uses are protected— whether
there are certain exceptions and limitations that
16 This could be considered a situation of a ‘tragedy of the
anti-commons’ in which too many property rights in the
same asset lead to inefcient underuse of that asset. See
Michael A. Heller, ‘Tragedy of the Anti-Commons: Property
in the Transition from Marx to Markets’ (1998) 111 Harv L
Rev 621.
17 This distinguishes ‘data ownership’ from the protection of
personal data. It is to be noted that data protection rules
in the EU only protect natural persons. See Regulation (EU)
2016/679 of the European Parliament and of the Council of 27
April 2016 on the protection of natural persons with regard
to the processing of personal data on the free movement of
such data, and repealing Directive 95/46/EC (General Data
Protection Regulation), [2016] OJ L 119/1. Corporate entities
may also have an interest in keeping back information that
has the potential of harming their ‘corporate reputation’.
However, this can be seen as part of their commercial
interests. In this context, trade secrets rules may provide
some protection. On this, see at C.II. below.
Designing Competitive Markets for Industrial Data
2017
261
4
take into account conicting interests and, nally,
which remedies will be made available to the right-
holder. In making such decisions on the framing of
the new right’s regime, the economic arguments that
justify the recognition of a new right as such have
to play a key role.
11 In addition, any new legislation on data ownership
should take into account the public interest in
maintaining competition in the market. Additional
rights regarding data as an asset may enhance market
power deriving from the control of data. As in other
elds of intellectual property law, the guidepost
should be that both property rights and competition
pursue the goal of enhancing innovation.
18
If the data
ownership right is supposed to create incentives
to invest in new data-based business models by
controlling the use of data, and if competition is
designed to maintain competitive pressure on the
right-holder to maintain its incentives to invest,
the best approach will be to take the competition
dimension into account as a core consideration for
the design of the property rules. This approach has
the advantage of reducing the need for later reliance
on competition law as a countervailing legal regime.
Accordingly, the interest in maintaining access to
data in the interest of society would have to be one
of the criteria that guide any future legislation on
data ownership.
12 In the following, the article will rst take a look at
the phenomenon of the emerging data economy and
how value is generated in that economy (section B.
below). Then, it will explore to what extent there
is already control over data, in the form of either
factual control or legal control based on specic
protection regimes (section C. below). Against this
backdrop, it will be possible to discuss whether and
to what extent there is an economic justication
for additional protection (section D. below).
Furthermore, the article will explore the different
issues concerning the design of an additional
protection regime (section E. below). Yet the analysis
is not limited to the question of whether additional
ownership rights are needed. Rather, in part F.
this article will analyse and discuss legal regimes,
including competition law and more targeted forms
of legislation, to enhance access to data in order to
promote a pro-competitive data economy.
18 See Communication from the Commission—Guidelines
on the application of Article 101 of the Treaty on the
Functioning of the European Union to technology transfer
agreements, [2014] OJ C 89/3, para 7.
B. The phenomenon of
the data economy
13
For the purpose of this article, a number of
particular features of the data economy need to be
understood properly in order to answer the policy
issues surrounding data ownership. This includes a
description of the use of data as an asset in the data
economy and the economic and societal benets of
that economy (part B.I. below), the phenomena of
‘data’ and ‘big data’ in this context (part B.II. below),
specic features of how value is generated in this
economy (part B.III. below), and nally the interests
of specic stakeholders that need to be taken into
account in designing any future legislative action
(part B.IV. below).
14 All of these issues are closely linked to new business
models that are currently evolving in very diverse
sectors of the data economy. This means that the
following analysis has to do with very dynamic
phenomena of high complexity and variety. Anybody
who engages in this topic has to understand what
is actually going on in the market concerning the
underlying business models; also, generalisations
need to be considered with caution. This is already
an important lesson for the legislature. Any rule that
is adopted against the backdrop of one case scenario
also has to t other scenarios to which it may apply.
In addition, property legislation in particular should
not only respond to the needs of today’s economy,
but also the needs of tomorrow. This argues against
precipitate legislative action, despite the enormous
speed of the development of the data economy, at
least as regards the recognition of new property
rights without a clear understanding of the business
models that will be affected now and in the future.
Such new rights have the potential of increasing
market power, creating barriers to access to
important data and, ultimately, curbing rather than
fostering the data economy.19
19 An example of such premature legislation was the
introduction of the sui generis database right by the EU
legislature in 1996. See Directive 96/9/EC of the European
Parliament and of the Council of 11 March 1996 on the
legal protection of databases, [1996] OJ L 77/20. Ten years
after its adoption, the Commission had to admit that there
was no evidence that the Directive had indeed produced
the expected positive economic effects as regards the
information market in the EU. The Commission even
considered a withdrawal of that protection, without,
however, recommending it. See DG Internal Market and
Services Working Paper—First evaluation of Directive
96/9/EC on the legal protection of databases (12 December
2005), available at:
copyright/docs/databases/evaluation_report_en.pdf>
(accessed 10 September 2016).
2017
Josef Drexl
262
4
I. Data as a most important asset
of the data economy and the
societal benefits deriving from it
15
In the data economy, data have become the key asset
for conducting business. This explains why data are
often called the ‘oil’ of the new economy.
16
Beyond the use of this buzzword, it is more
important to understand why and how data are used.
Different forms of use relate to different stages of the
development of the Internet. At its rst stage, the
Internet was used as a tool for providing information.
This was the time when politics started to realise
that an ‘information society’ with new services was
emerging that was in need of new legislation.20 At
this rst stage of development the Internet emerged
as an information and selling platform (web 1.0).
17
At the second stage, new business models developed
that provided consumers with other kinds of
services, yet still related to information, without
charging them a price. These services, such as search
engines or social platforms that connect people with
people (web 2.0), were often exclusively nanced by
advertising. Whereas, at the rst stage, information
was largely limited to information as an object of
the service; at the second stage, personal data
became a most important input for new kinds of
business models that were information-related. The
advertising value of a service or platform increases
with its attractiveness for private users who, in turn,
provide its operator with personal data as the key
input for such business models.
18
In the Internet of Things, physical objects get
connected with each other and with the environment.
This brings about another major boost of the data
that are collected and an extension of the data that
enter into big data collections and business models.
At this stage of Internet development, any data that
is collected by somebody for a particular purpose can
become a most important asset for other economic
players or public entities for very different purposes.
For instance, smart cars nowadays collect data for
steering driverless cars and for providing better and
timely—even predictive—maintenance services. But
cars may also register the driving habits of the driver,
in which the insurance companies are interested, the
geographical location of the car at a given moment
20 In the EU, see in particular Directive 2000/31/EC of the
European Parliament and of the Council of 8 June 2000 on
certain legal aspects of information society services, in
particular electronic commerce, in the Internal Market
(‘Directive on electronic commerce’), [2000] OJ L 178/1;
Directive 2001/29/EC of the European Parliament and of
the Council of 22 May 2001 on the harmonisation of certain
aspects of copyright and related rights in the information
society, [2001] OJ L 167/10.
can inform providers of geographic data, such as
Google Maps, about a change of the direction of a
one-way road, and inform the public authorities
about the volume of use and trafc conditions of
roads at a given time. The social benets of data will
even increase with the inclusion of the data in larger
datasets that bring together data from different
sources, such as from different car manufacturers
to get a more comprehensive picture of the concrete
trafc conditions in a particular geographic area.
The innovative character of this kind of use of data
consists in linking large datasets in order to answer
many different questions based on mere correlations
between different kinds of data (often called ‘data
mining’) in the interest of individual businesses or
the public.
19 In this big data world, it also seems that the role of
the state is beginning to change. At an earlier stage
of the development of the Internet, states started to
realise that it is becoming increasingly important
to grant private businesses access to publicly held
data (so-called ‘public sector information’, PSI)
for commercial re-use in order to promote new
commercial information services.21 Conversely,
the modern private data economy is increasingly
producing data from which big data analytics in
particular can extract new knowledge that can
optimise public decision-making—whether it is about
increasing trafc security based on data collected by
cars, protecting the environment, for instance, by
relying on information that is collected by machines
used in the agricultural sector, or revolutionising
health care around the world by collecting and
analysing the clinical, genetic, environmental, and
behavioural data from myriad sources.22 In other
words, the public sector is a major contributor, as
well as a beneciary of the data economy and big
data analyses.23
20
In sum, in the development of the ‘data economy’
a shift of focus can be observed. Whereas the
business models of major Internet platform
operators are built on the use of personal data and,
accordingly, may give rise to particular concerns
about effectively protecting the use of personal
data, the data economy will no longer be limited
21 See Directive 2003/98/EC of the European Parliament and
of the Council of 17 November 2003 on the re-use of public
sector information, [2003] OJ L 345/90, as revised by Directive
2013/37/EU of the European Parliament and of the Council
of 26 June 2013, [2013] OJ L 175/1; consolidated version
available at: http://eur-lex.europa.eu/legal-content/EN/
TXT/PDF/?uri=CELEX:02003L0098-20130717&from=EN>
(accessed 10 September 2016).
22 On the benets for health care, see in particular the study
by OECD (supra n 5) at 331-78.
23 The OECD argues that the governments should ‘lead by
example’ in promoting data-driven innovation by granting
access to public-sector information. See OECD (supra n 5) at
404-48.
Designing Competitive Markets for Industrial Data
2017
263
4
to the use of personal data for advertising and
marketing purposes. There are two more important
innovation-driven features of the data economy that
can be witnessed. On the one hand, in the era of the
Internet of Things, data collection by sensors will
allow consumers to be provided with innovative
smart products and services that will increasingly
replace traditional products. On the other hand, the
data collected in this industry will be of particular
utility to private actors in very different business
sectors and to public entities. Hence, data collected
by smart products will become an important input,
both for other businesses and for the government.
II. What do we mean by
data and big data?
21
Asking the question of who owns the data suffers
from the terminological weakness of what is meant
by the term ‘data’. There are two aspects to the
problem. First, more precision is needed in dening
the individual data. The second aspect relates to the
aggregation of larger datasets and their protection.
22
The rst issue relates to the question of the potential
object of protection of data. Take the following
example: a smart car of manufacturer A, through
the sensors attached to its dampers, locates a
pothole. This information is not yet noticed by any
natural person; however, it is stored in the form of
digital data on a server of manufacturer A. If the law
recognised ownership of A in this data, the question
arises whether ownership relates to the pure digital
dataset in the form of bits and bytes, or to the
‘information’ the digital dataset contains. This makes
a major difference from a competition-oriented
perspective. The pothole can of course be registered
by the smart cars of different manufacturers (A and
B) that follow each other. Hence, the ‘information’
in which the public road authority is interested
could be extracted from two different (competing)
datasets.
23
This example shows that the concept of data is in
need of additional precision. When we use the term
‘digital data’, we typically refer to ‘machine-readable
encoded information’.24 However, the interest
in ‘protecting data’ relates to the information
encoded in these bits and bytes. As regards this
information, in turn, a distinction can be made in
terms of semiotics between the different levels of
information.25 For data protection, the distinction
24 Denition used by Herbert Zech, ‘Data as tradeable
commodity’ in Alberto de Franceschi (ed.), European Contract
Law and the Digital Single Market (Insentia: 2016 forthcoming)
51, at 53.
25 On this distinction see also Maximilian Becker, ‘Rechte
an Industrial Data und die DSM-Strategie’ (2016/1) GRUR
between the syntactic and the semantic level is key.
The syntactic level regards the representation of
information in particular signs, for instance as a text,
a photograph or a video. In contrast, the semantic
level relates to the meaning. Take the example of a
camera at a public square that produces a video. The
syntactic information is the video as such, which
can be stored on different carriers. In contrast, the
meaning that can be extracted from that video, for
instance, how many people or vehicles cross the
square on a single day, is placed on the semantic
level. These distinctions can be further illustrated by
the example of a novel printed as a book. The book is
the physical carrier of the information. The syntactic
information consists in the text printed in a sequence
of letters and words. The semantic information is the
story told by the novel. If somebody does not speak
the language in which the novel is written, to this
person the information will only be accessible on
the syntactic level.
24
Hence, whenever the law protects ‘data’, it has
to make clear what it really protects. There is no
general argument against protecting semantic
information. Indeed, trade secrets protection and
private data protection relates to the semantic level
of information.
26
The know-how of a rm consists
in technical knowledge; it does not matter whether
this knowledge arises from a drawing, a text or a
combination of both, or whether this knowledge
is stored in a digital format or not. Similarly,
individuals are protected against unauthorised
processing of information relating to them, whether
this information is contained in a text, photographs,
or audiovisual recordings. In contrast, in the
abovementioned example on the potholes in the
street, it would be better to avoid protecting the
semantic information the sensors of a car collect.
Hence, the question of whether the law should
protect the semantic or the syntactic information,
or even only the integrity of the digital le, will
depend on the circumstances. This analysis would
seem to argue for context-specic regulation. Even
a general regime on the protection of industrial
data would thus appear problematic since, in some
instances, protecting semantic information such as
in the case of trade secrets seems the right approach,
while protection of data collected through sensors in
the public sphere should probably not be extended
to the meaning these data are able to convey. To
Newsletter 7, available at:
leadmin/daten_bilder/newsletter/2016-01_GRUR_
Newsletter.pdf> (accessed 10 September 2016); Andreas
Wiebe, ‘Protection of industrial data—a new property right
for the digital economy’ (2016) Gewerblicher Rechtsschutz und
Urheberrecht Internationaler Teil (GRUR Int) 877, at 881; Zech
(supra n 24) at 53-54.
26 Art 4(1) General Data Protection Regulation (supra n 17)
denes ‘personal data’ as ‘any data relating to an identied
or an identiable person’.
2017
Josef Drexl
264
4
draw the line between the two approaches is not
an easy task. Constitutional rights can argue in
favour of protecting semantic information, such as
in the case of personal data. Yet in other instances
constitutional rights and competition policy will
argue against ownership in semantic information,
if such protection has the potential of undermining
the free ow of information.27
25 The second problem arises from the fact that rms
do not only hold individual pieces of data. Data
are collected and then included in larger datasets.
This raises the issue of whether there should be
protection of each and every data information or
whether there should be protection of the whole
dataset in its particular composition.
26
This second issue directs the attention to the
features of big data, the technical features of big
data analytics and, ultimately, big data business
models. At the outset, it should be stressed that big
data analyses are only one application where data
held by one person is used by another person in the
data economy. The purpose of big data analyses is to
optimise decision-making. The decision-maker can
be any person or entity, usually a rm or a public
entity. The following three features are key to the
technical understanding of big data: volume, velocity
and variety (the so-called ‘3 Vs’).28 ‘Volume’ relates
27 In this context, the Magill competition law case of the
European Court of Justice (now Court of Justice of the EU,
CJEU) should be recalled. Since British and Irish copyright
law recognised copyright protection for the mere listings
of TV programs, TV stations were able to monopolise the
downstream market for printed TV programs and prevent
the emergence of comprehensive TV guides combining the
programs of different TV stations. The case gave rise to
the EU case-law on refusal to license. For more detail see
at F.II.1. and F.II.2. below. Copyright protection blocked
access to the ‘information’ contained in the TV listings
and, thereby, gave rise to dominance of TV stations in the
upstream information market and allowed the TV stations
to eliminate competition in the downstream market. See
Judgment in RTE and ITV v Commission (‘Magill’), C-241/91 P
and C-242/91 P, ECLI:EU:C:1995:98, [1995] ECR I-743. For this
case, it can be argued that copyright went too far in the rst
place by blocking access to information. On this case see
also at F.II. below.
28 See, for instance, Amir Gandomi and Murtaza Haider,
‘Beyond the hype: Big data concepts, methods and analytics’
(2015) 35 Int’l J. Inf. Manag. 137, 138; Stephen Kaisler, Frank
Armour, J. Alberto Espinosa and William Money, ‘Big Data:
Issues and Challenges Moving Forward’, (2013) 46th Hawaii
International Conference on System Sciences 995, available at:
hicss/2013/4892/00/4892a995.pdf> (accessed 10 September
2016); Daniel O’Leary, ‘Articial Intelligence and Big
Data’ (2013) IEEE Intelligence Systems 96, available at:
fall2014/301/chapters/chapter1/mex2013020096.pdf>
(accessed 10 September 2016); Paul Zikopoulos and Chris
Eaton, Understanding Big Data: Analytics for Enterprise
Class Hadoop and Streaming (2011). The rst author to have
hinted at these three features seems to be Doug Laney, ‘3-D
to the exploding volume of data that is produced by
different sources, including the Internet of Things
and social media. Big data is dened by the fact that
the volume of data to be analysed transcends the
current capacity of storage and processing systems.
‘Velocity’ relates to the dynamic nature of big data.
Indeed, big data constantly changes as new data is
produced. To keep up with the speed of this process
is key in big data analytics because the users of
the results of such analyses will usually have to
rely on real-time analyses for decision-making in
a constantly changing world. ‘Variety’ relates to a
wide range of different kinds and formats of data.
Data may originate from very different sources, such
as machine sensors, websites or social platforms;
it may be structured or unstructured; and it may
consist in texts, pictures, audio or video. While it
would be important to combine different kinds of
data in big data analyses, the large variety of data
constitutes a major technological challenge to big
data analytics.29
27 These technical features also need to be taken into
account when it comes to the policy decision of
whether additional data ownership rights should be
created. The general claim to be made is that data
ownership should not create obstacles to big data
analyses, because it is through these analyses that
new insights and social benets will be generated.
The issue of volume indicates the difculty of storing
all data that needs to enter into an analysis on one
server. This means that big data analyses may have
to take place in a decentralised manner. Either the
‘code has to be brought to the data’ or individual
datasets need to be screened rst for the critical
data, which is then transferred for the analysis.30 In
both cases, it is clear that the big data analyst is in
need of access to different data sources and that the
different data sources cannot ex ante be considered as
substitutes for each other. Creating new data rights
at the upstream level of holding such datasets could
therefore considerably obstruct big data analyses.
28
Velocity may be an even more important feature
to be taken into account for the regulation of
ownership. Velocity indicates that ‘data’ should
not generally be considered as a ‘commodity’ that
Data Management: Controlling data volume velocity and
variety’ (2001), available at:
doug-laney/les/2012/01/ad949-3D-Data-Management-
Controlling-Data-Volume-Velocity-and-Variety.pdf>
(accessed 10 September 2016). From a competition law
perspective see Daniel L Rubinfeld and Michal S Gal, ‘Access
Barriers to Big Data’ (16 August 2016) at 8-9, available
at: (accessed 10
September 2016).
29 On the technique and process of data analytics see Gandomi
and Haider (supra n 28) at 140-143.
30 These two solutions are identied by Kaisler et al. (supra n
28) at 997.
Designing Competitive Markets for Industrial Data
2017
265
4
can be traded like other commodities. Rather, the
modern data economy typically has to rely on real-
time information. Hence, a concept of ownership in
data, similar to copyright in a work, which would
invariably be protected for a xed period of time,
would not serve the needs of such data services
and big data analytics. Big data analyses that are
confronted with dynamic processes and have to
serve a purpose in a dynamic environment, such as
steering the trafc management system of a smart
city, will have to rely on permanent access to real-
time data sources. Ownership in individual data will
hardly be able to constitute the backbone of such a
service.
29
Velocity is closely linked to another ‘V’ that is
increasingly mentioned as an additional feature of
big data and which is key from a legal perspective,
namely, ‘veracity’.
31
Data needs to be reliable to serve
the purposes of a data economy. Where real-time
data are needed, but not delivered, the service also
misses the requirement of veracity. From a legal
perspective, veracity indicates that the supply of
data should also come with particular responsibility.
30
In this regard, it is worth noting that the EU is
currently moving in the direction of xing uniform
standards of ‘quality’ of ‘digital content’ that
need to be respected if digital content is supplied
under a contract with a consumer.32 The Proposal
for a Directive on the supply of digital content
denes ‘digital content’ as ‘data which is produced
and supplied in digital form, for example video,
audio, applications, digital games and any other
software’.33 The Directive would have the effect
of creating a harmonised regime of contractual
liability for both physical goods, which are also
often sold over the Internet, and data. This, however,
does not automatically lead to the recognition of
ownership in the underlying data.34 Whether there is
contractual liability if digital content does not meet
the quality that is to be expected under the contract
and whether the supplier transfers ownership in
the framework of such a contract are two separate
legal issues. Most importantly, ownership implies a
31 An example is big data analytics in the healthcare sector;
see Wullianallur Raghupathi and Viju Raghupathi, ‘Big
data analytics in healthcare: Promise and potential’
(2014) 2(3) Health Information Science & Systems 1, at 2,
available at:
track/pdf/10.1186/2047-2501-2-3?site=hissjournal.
biomedcentral.com> (accessed 10 September 2016).
32 Article 6 of the Proposal of Commission of 9 December
2015 for a Directive of the European Parliament and of the
Council on certain aspects concerning the supply of digital
content, COM(2015) 634 nal.
33 Ibid, Art 2(1)(a).
34 See, however, De Franceschi and Lehmann (supra n 9) at 59-
60 and 71 (relying on the corresponding rule contained in
the previous draft for a Common European Sales Law and
attributing a property dimension to this proposal).
third-party effect while the proposed Directive only
creates rights and obligations between the parties to
the sales contract.35
31
In addition, also as regards big data analyses, the
difference between the syntactic and semantic level
of data is to be taken into account. Big data analytics
consists in reading large datasets to discover ‘new’
meaning—in the sense of (semantic) information—
that has so far not been observed. Big data analytics
acts like a person who is able to read the data in a
different way by identifying correlations between
different data—again in the sense of information—to
draw conclusions from those correlations. Hence, the
information that big data analyses produce is already
hidden in the pre-existing datasets. However, it is big
data analytics that allows us to discover this semantic
information. This explains how problematic it
would be to recognise protection of all semantic
information contained in the pre-existing datasets
for those who control access to these sets. It is indeed
the contribution of the data analyst that leads to
the discovery of that information and, hence, any
right in this information should be vested in the data
analyst36 rather than the holder of the datasets that
are analysed.
III. From value chains to
value networks
32
For considering whether new property rights in data
are to be recognised from a functional perspective,
it is crucial to understand who generates economic
value and, as a follow-on question, whether this
contribution depends on the recognition of a
property right. In this regard, it is important
to understand that in the data economy, value
is generated differently than in the traditional
economy.
33
In the traditional economy, the still dominant
paradigm relates to vertical value chains.
Manufacturers purchase input for the production
of goods in upstream markets and then sell them
through distribution chains—often including
wholesalers and distributors—to consumers. At each
level of the production and distribution chain, some
economic value is added.
35 As regards the recognition of ownership in the download
of a computer program by the CJEU in the Judgment in
UsedSoft, C-128/11, ECLI:EU:C:2012:407, paras 45-52, see at
C.V. below. See also De Franceschi and Lehmann (supra n
9) at 60-63 (relying on this decision in their yet cautious
support of data ownership).
36 Such information can constitute trade secrets. On trade
secrets protection see at C.II. below.
2017
Josef Drexl
266
4
34
In contrast, in a world of smart goods and the
Internet of Things, economic value is increased in
very complex and dynamic value networks, which
can be disruptive for traditional value chains,37
through collaboration of the different participants in
the network. This paradigm shift from value chains
to dynamic value networks is identied as a core
feature of the current digital transformation of the
industry.
35
Four sub-factors are relevant for this shift:38 (1)
Improving decisions based on data: sensor-generated
industrial data and analysis of big data help rms
optimise their decisions. For instance, predictive
maintenance becomes possible. (2) Full automation:
Automation through digital technology, including
robotics, revolutionises production and the use of
products (e.g. driverless cars). Automation increases
the speed of production and decreases the likelihood
of defects. (3) Connectivity: Objects and machines
within the factory and beyond get connected over
the Internet and allow supply and production to
be steered from the perspective of the need of the
customer, which results in quicker production and
distribution while saving resources. (4) Increasing
role of Internet intermediaries: The intermediaries
from the Internet sector who have the best access
to and knowledge of the needs of consumers
and of controlling the data interfaces between
different markets gain a competitive advantage
in the industrial sector where smart products are
produced. This explains why Google and other
rms are today trying to expand their activities
into the industrial sector. Google, or Alphabet as
Google’s parent company, may now already have
considerable competitive power for entering
the market for smart, driverless cars based on its
control of geographic data, and may provide most
efcient transport services to passengers who, in
the future, will no longer buy their own cars but
become passengers of Google transport services. At
the same time, by expanding their activities to the
production and operation of smart products, these
Internet intermediaries will gain control over new
sources of data.
36
Hence, whereas the digital transformation of the
industry decreases existing entry barriers and may
even force industrial incumbents out of the market,
control over data enables rms originating in the
Internet sector, such as Google, to enter into and
37 This has recently been highlighted by a study conducted
by Roland Berger Strategy Consultants on behalf of
Bundesverband der Deutschen Industrie (BDI). See Roland
Berger Strategy Consultants and BDI, ‘Analysen zur Studie
“Die digitale Transformation der Industrie”’ (February
2015) 4-8, available at:
Digitale_Transformation.pdf> (accessed 10 September
2016).
38 Ibid, at 8.
gain considerable market power in a large variety of
different markets for the production and operation of
smart products. Recognition of data ownership may
therefore have the unwanted effect of strengthening
the market power of these rms even more, while,
from a competition perspective, it would be wiser
to promote access to data that is needed by other
market players to operate in such markets.
IV. The interests of different
stakeholders
37
The preceding analysis already provides some
important insights into the interests of different
stakeholders. This analysis underlines the
observation in the introduction (at A. above) that
industrial players who have already started to invest
in the Internet of Things are reluctant to advocate
data ownership.
38 The major technological challenges of the Internet
of Things relate to big data analytics. This is the
area where most investment is needed for tackling
the technological obstacles to handling rapidly
growing dynamic datasets and solving the problem
of analysing a large variety of different kinds of
data. However, such innovation is more likely to
be fostered through copyright protection for the
software solutions employed in the framework of
big data analyses rather than through ownership in
the data analysed.39
39
Moreover, it is to be acknowledged that the non-
economic interests of natural persons in the use of
their personal data deserve to be safeguarded, also
in the data economy. While personal data protection
needs to be taken into account, it does not argue
as such against the recognition of an economic
ownership right of a rm that collects data about
the use of a smart product by a natural person.
Both rights can coexist. This has the important
consequence that rules on the protection of personal
data can prevent a data owner from commercialising
that data. The industrial holder of personal data can
also respect data protection rules by making the data
collected from individual natural persons available
to third persons in an aggregated and anonymised
form in larger datasets. To the extent that big data
analytics manages to reproduce personal data, data
39 Another kind of protection would consist in patent
protection for algorithms. However, this is rejected by Josef
Drexl, Reto M. Hilty, Luc Desaunettes, Franziska Greiner,
Daria Kim, Heiko Richter and Gintarė Surblytė, ‘Data
Ownership and Access to Data—Position Statement of the
Max Planck Institute for Innovation and Competition of 16
August 2016 on the Current European Debate’, paras 12-17,
available at: (accessed
12 September 2016).
Designing Competitive Markets for Industrial Data
2017
267
4
protection rules may apply again as regards the re-
use of that data.
40
As regards personal data, it is important to note that
the fact that a natural person is and will often be
the source of specic data does not automatically
argue in favour of allocating data ownership as an
economic right to commercially exploit that data to
this person. Protection of personal data is neither
vested in the natural person for economic purposes,
nor is it an absolute right.
40
Personal data protection
does not allocate economic value.41 Hence, there is
room to grant economic rights of exploitation of data
originating from natural persons to other persons
or rms.
41
The same applies as regards the property of the
purchaser of a smart product. The property in the car
as a physical object does not automatically extend
to the commercial exploitation of the data that are
produced by the sensors of that car. The question
of whether data ownership should be recognised,
and for whom and with which scope of protection,
should only be decided against the backdrop of
economic welfare considerations.
C. Existing protection regimes as
a basis for ‘data ownership’
42
Already at the end of the preceding part, it was
claried that at least two rights that are recognised
by law do not provide a sufcient basis for data
ownership; namely, personal data protection and
real property in a smart product that produces the
relevant data. However, there are other legal regimes
that could provide protection in favour of the rm
that controls data. Most obvious candidates are
database rights and trade secrets protection. Beyond
this, in certain circumstances, the question may arise
whether patent protection extends to data that is
generated through a patented process. Moreover,
one could also contemplate unfair competition rules
and the like, as well as a generalisation of property in
tangibles as a civil law concept. In sum, none of these
regimes provides a convincing or comprehensive
basis for data ownership. In contrast, it will be shown
that factual control over data can enable the data
holder to commercialise that data without additional
legal protection by relying on contract law.
40 See Recital 4 of the General Data Protection Regulation
(supra n 17). See also Pamela Samuelson, ‘Privacy as
Intellectual Property’, (2000) 52 Stanford L Rev 1125.
41 Zech (supra n 24) at 60.
I. Database protection
43
At rst glance, database rights present a most
obvious property regime for controlling access
to data.42 However, this kind of protection has
particular limitations that explain why it will often
fail to provide protection to data for the new business
models of the data industry.43
44
The EU legal regime for database protection provides
for a two-tier system: Copyright protection is
granted to creative databases;
44
sui generis protection
is granted to databases based on ‘substantial
investment’.45
45
The availability of copyright protection can be
excluded from the outset. Article 3(1) of the Database
Directive claries that the character of a creative
work dened as ‘the author’s own intellectual
creation’ has to relate either to the selection or to the
arrangement of the database’s contents. According
to the CJEU this originality requirement is satised
if ‘through the selection or arrangement of the data
which it contains, its author expresses his creative
ability in an original manner by making free and
creative choices … and thus stamps his “personal
touch”’.46 Already this denition explains that
the individual data as such will not be copyright
protected. This is also explicitly conrmed by
Article 3(2) of the Directive, which states that
copyright protection for databases will not extend
to the contents as such. Hence, even if data were
included in a copyrightable database, such copyright
protection would not extend to that data.
46
Sui generis database protection may at rst glance
provide a better basis for protecting data generated
in a world of the Internet of Things.47 However, this
form of protection also has its limitations. They arise
from both the subject-matter of protection and the
scope of protection.
47
As regards the subject-matter of protection, a
‘database’ is uniformly dened as a ‘collection of
42 Arts 7-10 Database Directive (supra n 19).
43 See also Wiebe (supra n 25).
44 Art 3 Database Directive.
45 Art 7(1) Database Directive. Note that both forms of
protection may also coincide. A given database may be both
creative and based on substantial investment.
46 Judgment in Football Dataco v Yahoo! UK, C-604/10,
ECLI:EU:C:2012:115, para 38 (adopting the general originality
concept of EU copyright law as developed by the Court for
other categories of works to databases).
47 It is even argued that the sui generis database right will
often protect big data databases; see Giulio Corragio, ‘Big
data and IoT–a great match with troubles…’ (19 June 2015),
available at:
a-great-match-with-troubles/> (accessed 10 September
2016).
2017
Josef Drexl
268
4
independent works, data or other materials arranged
in a systematic or methodical way and individually
accessible by electronic or other means’.48 Protection
will also be granted if the arrangement and storage
is accomplished by ‘electronic, electromagnetic or
electro-optical processes’.49 Hence, collections of
digital data can usually be considered as databases
in the sense of the Directive.50 However, a sui generis
database right only subsists if ‘there has been
qualitatively and/or quantitatively a substantial
investment in either the obtaining, verication or
presentation of the contents’.51 The CJEU has interpreted
these requirements in a very restrictive way. It
claried that the investment has ‘to refer to the
resources used to seek out existing independent
materials and collect them in the database, and not
to the resources used for the creation as such of
independent material.’52 The CJEU explained this with
the objective of the Directive to create incentives for
the making of databases and not for the creation
of the data that goes into the database.
53
Hence, a
distinction is to be made between the ‘creation’ of
the materials contained in the database and the
‘obtaining’ of these materials.54 This leads to the
conclusion that the creation of smart products with
sensors that collect data should not be considered
for the assessment of whether the investment in the
database was ‘substantial’.55 The same applies to big
data analyses. These may well require substantial
investment. However, such analyses only lead to
the creation of new data in the form of knowledge,
which may then be included in databases. For the
protection of these databases, the investment in the
big data analyses is not to be taken into account.
48
As regards the scope of protection, it is important
to note that the sui generis database right only
protects the database as a collection of data and not
the individual data. The Directive thereby aims to
keep the (semantic) information that can be derived
from the data in the public domain.56 Extraction and
re-utilisation of individual data only fall within the
scope of protection of the database if these data form
48 Art 1(2) Database Directive (supra n 19).
49 Database Directive, Recital 13.
50 Zech (supra n 24) at 70.
51 Art 7(1) Database Directive (supra n 19) (emphasis added).
This means at the outset that there may be databases
fullling the denition of a ‘database’ in the sense of the
Directive that, however, are not protected since they
meet the requirements neither for copyright-protected
databases, nor for sui generis databases. Conrmed by the
CJEU in its Judgment in Ryanair v PR Aviation, C-30/14,
ECLI:EU:C:2015:10, paras 35-40.
52 Judgement in British Horseracing Board, C-203/02,
ECLI:EU:C:2004:128, [2004] ECR I-2195, para 31.
53 Ibid.
54 Ibid, para 32.
55 See also Żdanowiecki (supra n 10) at 21.
56 See Zech (supra n 24) at 71.
a ‘substantial part, evaluated qualitatively and/or
quantitatively, of the contents of that database’.57 The
concepts of ‘extraction’ and ‘re-utilisation’ further
restrict the scope of protection. In particular, big
data analyses, whereby the ‘code comes to the data’
in order to generate new information, will not lead
to any ‘extraction’ since there will be no ‘permanent
or temporary transfer of all or a substantial part of
the contents of a database to another medium’.58
49
In sum, it is quite obvious that the Database Directive
is based on a database technology that no longer
corresponds to the use of data in an era of ‘Industry
4.0’ or the Internet of Things. In particular, by
protecting a collection of materials for a given
period of time (15 years as of the completion of
the database),59 the concept of a database is much
too static to adequately respond to the features of
constantly changing datasets and real-time data
services.
50 This latter point may raise the question of whether
the Database Directive is in need of a reform.
However, the fact that the Directive does not
respond to the needs of the modern data industry
in a technologically appropriate manner cannot by
itself justify reforming the Directive by introducing
a right of data ownership. Rather, such reform is in
need of an economic justication, which is part of
the analysis further below (section D. below).
II. Trade secrets protection
51
Trade secrets protection is another protection
regime that inevitably comes to mind as regards the
protection of data.
52
The EU has recently adopted a directive for
harmonising the national rules on trade secrets
protection.60 As regards the modern data industry,
this Directive may already be considered as
technologically out-dated, since at the time of
the preparation of the Commission Proposal, the
implications of the new data economy were not yet
fully perceived or understood.61 As a consequence,
57 Art 7(1) Database Directive (supra n 19).
58 Article 7(2)(a) Database Directive.
59 Article 10(3) Database Directive only takes changes to
contents of databases into account to the extent that such
changes amount to a new substantial investment, which
leads to a revival of protection for 15 years.
60 Directive (EU) 2016/943 of the European Parliament and
the Council of 8 June 2016 on the protection of undisclosed
know-how and business information (trade secrets) against
their unlawful acquisition, use or disclosure, [2016] OJ L
157/1.
61 See Proposal of the Commission of 28 November 2013
for a Directive of the European Parliament and of the
Designing Competitive Markets for Industrial Data
2017
269
4
the text of the Directive is rather unclear as to
what extent, for instance, data produced by smart
products benet from trade secrets protection.
53 In comparison to database protection, trade secrets
protection has the obvious advantage of protecting
the specic information. However, there are other
shortcomings:
54
Most importantly, trade secrets protection relies
on rather narrowly dened requirements for the
subject-matter of protection. According to Article
2(1) of the Directive, the know-how or business
information (1) needs to be ‘secret’ in the sense
that it is not ‘generally known among or readily
accessible to persons within the circles that normally
deal with the kind of information in question’; (2)
the information must have ‘commercial value’
because of its secrecy; and (3) it has to be subject
to ‘reasonable steps …, by the person lawfully in
control of the information, to keep it secret’. None
of these three requirements can be easily applied in
the context of data produced by sensors attached
to smart products. First, while the secrecy could be
conrmed for data that is produced by the machines
inside a factory, data collected by smart cars on
freely accessible roads could be collected by the cars
of many manufacturers and, hence, will not full this
requirement.
62
Second, while data may nowadays
have great commercial value, it is quite questionable
whether it will always be possible to establish a
causal link between the secrecy of the information
and its commercial value. In the context of big data
analyses, an individual piece of information may
appear quite trivial, but particular value may arise
from correlations with other data.63 Third, it is very
unclear which steps will be required of the person
in control to keep the information secret.64 Fourth,
where data is generated in a network of different
entities connected through a value network, it will
be particularly difcult to allocate protection to a
single person controlling the secret.65
55 Yet another question is whether the subject-matter
of protection needs to be interpreted narrowly in the
Council on the protection of undisclosed know-how and
business information (trade secrets) against their unlawful
acquisition, use and disclosure, COM(2013) 813 nal. See
also Wiebe (supra n 25) at 880 (pointing out that the drafters
of the Directive did not have big data in mind).
62 In this context, it is important to note that independent
discovery of the same information will not lead to unlawful
acquisition of the information. See Article 1(3)(a) Trade
Secrets Directive.
63 See Zech (supra n 24) at 63 (therefore criticising Recital
8 Trade Secrets Directive according to which trivial
information should not be protected).
64 On the difculties to keep information secret in a network
environment, see Wiebe (supra n 25) at 880.
65 See also Wiebe (supra n 25) at 880.
light of the objectives of the Directive. The Directive
pursues the goal of promoting the competitiveness
and innovative strength of businesses through
protecting secret information.66 However, data
are nowadays largely produced as a by-product of
smart machines and goods, whereas these data can
be commercialised in completely different markets
and for completely different purposes (not least in
the public interest). Here, data is largely used by the
data holder as an asset for generating additional
income. In addition, protection of the data as trade
secrets will not always promote innovation through
the holder of that data. Rather, the challenge will
often consist in promoting access to that data for
other rms and public entities that may generate
additional knowledge from that data through big
data analyses. This argues for making a distinction
between information that serves the core business
of the holder of data, such as personal data held by
Internet platform operators, as the backbone of the
underlying business model, as well as data generated
through machine sensors that are designed to be
immediately used for the production process on
the one hand, and other data, which are rather a
by-product of the rm’s core business, on the other
hand.
56
Finally, it should be noted that trade secrets
protection is much narrower in scope than an
exclusive data use right. It does not protect against
any use of the data, but requires ‘unlawful’ conduct
which, to summarise the different provisions of the
Directive, can be regarded as contrary to honest
commercial practices.67 Hence, the Trade Secrets
Directive only establishes a system of liability for
specic tortious conduct and not a property rights
system.68 However, such further limited protection
can be considered as better suited to serve the
purposes of the data economy, by focussing on
the particular way in which a third party has
specically acquired access to the data instead of
granting exclusive protection against the use of data.
Such exclusive property protection would easily
conict with the fundamental right of freedom of
information.
III. Patent law
57
In limited sets of cases one could even consider
protection based on patent law. The reason for this
is that the scope of process patents also extends to
‘products’ that are obtained through that process.
For instance, in the European Union, Article 25(c)
of the—yet not effective—Agreement on the Unied
66 See Trade Secrets Directive, Recital 1.
67 See, in particular, Art 4(2)(b) Trade Secrets Directive.
68 See also Drexl et al (supra n 39) at paras 18-20.
2017
Josef Drexl
270
4
Patent Court stipulates that a process patent also
provides the right to prevent a third party from
‘offering, placing on the market, using, or importing
or storing for those purposes a product obtained
directly by a process which is the subject-matter of
the patent’.
58
The question in this regard is whether ‘data’ can be a
‘product’ that is obtained by using a process patent.
69
This question would become relevant for instance
where data is produced in a factory in applying
a patented production method or, maybe more
relevant, in the context of a process patent applied
in medical diagnostics. In the latter case, the patent
owner would also ‘own’ the ‘result’ of the diagnosis.
59 However, at the outset, such protection would only
become relevant where the patent is used without
the consent of the right-holder. Only if the patented
process is used without a licence does the patent
holder have a right to prohibit the commercialisation
of the product as the offspring of the process.
60
The reason why the legislature extends the protection
of process patents to the commercialisation of
products is that process patents are much weaker
than product patents. The owner of a product patent
enjoys full protection against price competition from
imitators in the product market. In contrast, the
holder of a process patent runs the risk of having
to compete with rms that offer essentially the
same product manufactured with an alternative
process. Extending protection to the products that
are produced with the process assimilates process
patents to product patents regarding the economic
incentives arising from the patents. It also addresses
the problem that third parties could otherwise
legally serve the market with products produced
abroad by applying the process if the process patent
is only protected in the importing country.
61
However, already as a matter of principle, it does
not seem appropriate to extend patent protection
to information as the product of a process patent.
Moreover, German courts seem to deny protection
for information that is derived from a process
patent. An interesting decision in this regard is the
one by the District Court of Düsseldorf in the Hunde-
Gentest case.70 In this case, the process patent for a
69 On the similar issue whether patent protection for a
computer-based process for producing aesthetic creations
extend to these creations see Jean-Marc Delthorn,
‘Councours de droits sur les œuvres numériques—Le cas des
creations issues de procédés brevetés’, (2016) 60 Propriétés
intellectuelles 285.
70 Landgericht Düsseldorf of 16 February 2010, Case 4b 0
247/09—Hunde-Gentest, available at:
de/duesseldorfer-archiv/?p=813> (accessed 10 September
2016). See also Oberlandesgericht München (Higher District
Court Munich) of 22 October 2015, Case 6 U 4891/14, (2015)
Beck-RS 18783.
genetic test for dogs was protected in Germany,
but not in Slovakia. The defendant, who previously
applied the test in Germany, moved the testing to
Slovakia to avoid a patent infringement. Therefore,
the Court was only requested to decide whether the
plaintiff can rely on a process patent to prevent the
defendant from communicating the test results to
Germany. The Court denied such protection, arguing
that the test results as pure information cannot be
considered the product of the process. The Court
noted that, since information is directly accessible
for humans without any further technical process,
information as such lacks technicity and therefore
cannot be patented. Yet the Court refrained from
arguing that the ‘product’ of a process needs to
be patentable by itself in order to be protected
within the scope of the process patent.71 Rather, the
Court showed great sensitivity for the free ow of
information. It rejected protection so as to avoid
using patent law as a kind of trade secrets protection.
In particular, the Court stressed that patent law
should not support a claim to ban communication
of the test result to anybody in Germany, which, in
the last resort, would even include denying a person
who knows about the test result entrance to German
territory.
IV. Unfair competition law and
similar protection regimes
62
In many jurisdictions, unfair competition laws
and similar protection regimes, such as the tort of
misappropriation in common law countries, may
provide subsidiary means of protection against free-
riding where other protection mechanisms are not
available.
63 However, whether such a role should be attributed
to these general principles or laws as regards the
holding of data, is again a policy issue which should
only be answered in the afrmative if there is
sufcient economic justication for protection
against free-riding (see section D. below). Free-riding
as such should not be considered a violation of the
law unless it undermines incentives for investment
in the production of the asset that is copied.
71 This is also the view of the EPO. See EPO, Decision of the
Enlarged Board of Appeals, G 1/98, Transgenic plant/
NOVARTIS-II, [2000] OCJ EPO 111, at 138. The Enlarged
Board of Appeals conrmed the availability of process
patents, including protection of the products deriving
from the process according to Art 64(2) EPC, even in a case
where the product would be a plant, which is excluded from
patentability under Art 53(b) EPC.
Designing Competitive Markets for Industrial Data
2017
271
4
V. ‘Digitisation’ of the civil law
concept of property?
64
Civil law countries are not unlikely to discuss
nowadays whether the concept of property found
in the national Civil Code, which is usually limited
to the ownership of tangible items and land, should
be opened, namely, in a move to ‘digitise the Civil
Code’, to also include data. For instance, in 2016, the
Deutsche Juristentag,
72
which is the most important
private discussion forum for legal reform in
Germany, bringing together law professionals from
all different sectors, considered whether German
civil law is in need of a ‘digital up-date’.73
65
Yet, to equate data with tangible objects as a subject-
matter of property is a rather risky undertaking. The
risk is that, as an expression of general enthusiasm
and striving for modernisation, the legislature or
courts will not give sufcient consideration to the
different economic characteristics that distinguish
markets for non-tangible objects from those for
tangible objects.
66
Hence, the question of whether civil law is in need of
being ‘updated’, should be considered carefully and
within the specic context of protection. To transfer
the principles of contractual liability developed
for the sale of tangible goods to defects of digital
goods, is one thing;74 to recognise a property right
for holders of data with exclusionary effects on third
parties is another thing. In Germany, the debate is
mostly triggered by certain limitations of tort law.
Under Section 823(1) German Civil Code, there is
only a claim for damages if somebody injures the
‘life, body, health, freedom, property or another
right’ of someone else.
75
Courts have continuously
extended the range of ‘other rights’, to include, for
instance, the general personality right, but they have
also limited those rights to ‘absolute rights’. This is
why it is now also discussed whether courts should
recognise ‘data ownership’ as another absolute right
to protect the integrity of datasets against injuries
72 The Deutsche Juristentag convened in Essen on 13-16
September 2016.
73 The debates at the Deutsche Juristentag revolve around
Gutachten (expert reports), which are usually prepared by
law professors. The ‘digital update’ of the German Civil
Code is assessed in the Gutachten by Florian Faust, Digitale
Wirtschaft—Analoges Recht—Braucht das BGB ein Update?
Gutachten A zum 71. Deutschen Juristentag (Munich: C.H. Beck,
2016), also available at:
ic/f/1376130/26847040/1455040340113/Faust+Digitale+Wir
tschaft+Analoges+Recht+Gutachten+fur+den+71.+DJT.PDF?t
oken=73St8IVwwV4tYnJQSVMQJmH3F8c%3D> (accessed 10
September 2016).
74 See the Commission Proposal for a Directive (supra n 32).
75 English translation of the Bürgerliches Gesetzbuch available
at:
(accessed 10 September 2016).
committed by third parties. For instance, the need
for such protection is quite evident when computer
viruses delete large and valuable datasets, while the
physical carrier and its functions remain intact. The
downside of this is that recognition of such a right
in the framework of Section 823(1) Civil Code would
also provide for injunctive relief to prevent injury.
For that purpose, German courts rely on an analogy
to Section 1004 Civil Code, the basis for injunctive
relief in case of unlawful interference with property.
67
Injunctive relief raises the more important question
regarding the extent to which the scope of protection
of such data ownership is to be assimilated to
property in tangible objects. Property in tangibles
basically provides two sub-rights, a right of integrity
and a right to exclude others from any use.
76
The
debate on data ownership is inspired by the lack of
protection as regards the integrity of data, whereas
the recognition of a right to exclude other persons
from any use of the data would amount to a very
powerful intellectual property right that would
have the potential of undermining the free ow of
information.77 Also, from an economic standpoint,
a right to exclude others from the use of data is less
needed than in the case of tangibles. Data are not
rivalrous; hence, someone else’s use of the same
data does not prevent the ‘owner’ from using these
data. Accordingly, from an economic perspective, it
is easier to justify protection of the integrity of data
than to provide full protection, including injunctive
relief, as regards the use of data.
68
This debate on extending the property concept
to digital data was more recently also inspired by
the UsedSoft decision of the CJEU.78 In this case, the
Court explicitly recognised ‘ownership’ of the person
legally downloading a computer program from the
Internet. However, this holding was limited to the
application of the exhaustion rule in the Computer
Programs Directive.
79
Exhaustion of the distribution
right under copyright law requires a rst ‘sale’ of
a copy of the work through the right-holder or
with her consent. The CJEU dened a ‘sale’ as ‘an
agreement by which a person, in return for payment,
transfers to another person his rights of ownership
76 As regards the right to exclude under German law, see Sec
903 Civil Code. On the distinction between the three different
rights of property regarding data ownership, including (1)
possessing data—with the possibility to exclude access—, (2)
using data, and (3) destroying data (right of integrity), see
Zech (supra n 24) 56-57.
77 See also Wiebe (supra n 25) at 882. (considering whether
recognition of data ownership would lead to a paradigm
shift in protecting information).
78 Judgment in UsedSoft (supra n 35).
79 See Art 4(2) Directive 2009/24/EC of the European
Parliament and of the Council of 23 April 2009 on the legal
protection of computer programs, [2009] OJ L 111/16.
2017
Josef Drexl
272
4
in an item of tangible or intangible property’.80 By
recognising ownership in the digital copy of the
program, which is provided to a customer on a
permanent basis,81 the Court managed to transfer
the concept of exhaustion to the digital eld. Hence,
in UsedSoft, the CJEU did not recognise any general
concept of data ownership.82 Rather, the Court only
relied on ownership in a digital download to limit
the exclusivity of the copyright as another property
right. We can learn from this judgment that limited
recognition of property rights can also have a
liberalising effect and thereby promote the free
movement of data in the digital economy. However,
such recognition should not be generalised by
arguing in favour of allocating ownership involving
third-party effects wherever persons are legally in
permanent control over the use of any data. This
may well have the opposite effect of hampering
the free ow of data and information in the data
economy.
VI. Factual exclusivity
and contract law
69
Despite the uncertainties and shortcomings of the
different protection regimes, the players of the
data economy do not seem to suffer from the lack
of recognition of general data ownership. The reason
is that markets can also develop with relatively little
legal exclusivity where access can effectively be
controlled by technical means.83 Factual exclusivity
has the potential of forcing parties into negotiations
and can trigger transactions in very similar ways as
in the case of intellectual property.
70 Such data contracts based on the factual holding of
data are therefore meant to grant access to these
data.
84
However, this does not exclude agreement on
certain limitations of the use of data. Accordingly,
contract law may exercise even stronger restrictions
on the use of data than a new ownership agreement
that could provide for mandatory exceptions and
limitations.85
71
A very prominent example of an area where markets
for immaterial exploitation emerge with very
little legal exclusivity is the marketing of sports
rights. There are only few jurisdictions which
80 UsedSoft (supra n 35) at para 42.
81 Ibid, at para 45.
82 This is also conrmed by authors who rely on this judgment
to argue in favour of a concept of general data ownership.
See De Franceschi and Lehmann (supra n 9) at 60-63.
83 See also Żdanowiecki (supra n 10) at 25.
84 See Zech (supra n 24) at 59.
85 On the question whether promoting access may hence
justify introduction of a data ownership see at D.V. below.
provide special intellectual property rights for the
audiovisual exploitation of sporting events.86 Other
jurisdictions manage to provide the same conditions
for markets for sports rights with comparable value
streams without such legislation. The reason for this
is that the organisers of such events can control
access to the premises of the sporting events and
thereby charge a price from the broadcaster that is
allowed to produce the broadcast.87 Of course, there
is a risk that third parties will use the broadcasts
without authorisation. However, it sufces in this
regard that the broadcasting corporation that
was granted access to the event is protected by
its investment by copyright, or at least its original
related right, in the broadcast.
72
As regards the data economy, this example of
the sports rights may explain that, even where
misappropriation by third parties is a concern, there
is no need to recognise ownership of the data holder
as long as the investor in access to the data—such
as the big data analyst—disposes of an intellectual
property right that prevents third-party use, such as
the copyright in the software tools for analysing big
data. The data holder itself will regularly be able to
exclude others from access through technical means,
including technical protection measures. Rules of
criminal law that make unauthorised access to
data a crime, such as data or computer espionage,
can further strengthen factual exclusivity without
recognition of ownership in the sense of private law.
D. Potential justifications for
recognising data ownership
73
Against the backdrop of the uncertainties and
shortcomings of existing protection regimes, we now
turn to the question of whether there is an economic
justication for the recognition of data ownership.
In this regard, the analysis can rely on insights from
intellectual property scholarship.
86 The most prominent example is French law. Arts L333-
1 through L333-5 Sports Code (Code du sport) vest the
sports associations with an exclusive right of audiovisual
exploitation. Original French text of the Code du sport
available at:
do?cidTexte=LEGITEXT000006071318> (accessed 10
September 2016).
87 See also Thomas Margoni, ‘The Protection of Sports Events in
the EU: Property, Intellectual Property, Unfair Competition
and Special Forms of Protection’ (2016) 47 IIC 386 (arguing
that, in principle, the combination of the exclusivity of the
sports venue and contract law is capable of making markets
for sports rights work).
Designing Competitive Markets for Industrial Data
2017
273
4
I. Incentives for generating
and collecting data
74
The standard argument in favour of recognising
intellectual property rights is based on a utilitarian
incentive theory. Intellectual property is designed
to promote innovation. However, the subject-matter
of protection of these rights, such as inventions and
works of creativity, is characterised by the features
of public goods. Without the recognition of legal
exclusivity, everybody else would be able to free-
ride by copying and, consequently, nobody would
be willing to invest in the production of such public
goods.88
75 As demonstrated further above, the generation and
collection of data allows for very new and innovative
business models that lead to large gains in allocative
efciency in manufacturing and maintenance, as
well as far-reaching social benets based on big data
analyses. Hence, there is a case for also fostering
incentives for generating and collecting the
underlying data. However, it is less clear whether,
for that purpose, data ownership is required. In this
regard, the incentives of different players need to
be analysed.
76 There is always some human act that can be found
at the very beginning of the generation of data
and the commercial exploitation of these data. A
manufacturer may decide to employ machines and
robots that are equipped with sensors to control and
steer the production process. The owner of a smart
car decides where to go with this car and where the
car will register data about the density of trafc or the
physical conditions of the road. A patient provides
the blood for a blood test, the result of which may
go into datasets that are subsequently analysed. In
all of these cases, the relevant person would and
should certainly know about the generation of the
digital data, and may even have to give her consent
based on the rules on data protection. However,
additional ownership in the data is not necessary
as an incentive to generate such data. Hence, in
principle, it is possible to conclude that there is no
need to vest the person at the beginning of the value
chain with exclusive rights to exploit that data as
a means to create incentives for the generation of
that data.
77
The same holds true for the next step of exploitation.
The data produced by a smart car will be transferred
to the manufacturer of that car. The car manufacturer
will be sufciently motivated to generate data that
88 On the public goods theory for intellectual property, see,
in general, William M. Landes and Richard A Posner, The
Economic Structure of Intellectual Property Law (Cambridge, MA
and London, UK: The Belknap Press of Harvard University
Press, 2003) 12-16.
will guarantee smooth operation and maintenance
of the car. Generation of that data is very much
part of the rm’s business model. Furthermore, the
potential of follow-on markets creates sufcient
incentives for collecting the data, whether database
rights are available or not, even in cases where the
main business model does not require the data to be
stored on a permanent basis.
78 Nor are additional incentives needed as regards the
business model of Internet platform operators (e.g.,
search engines, social media etc.), for which the
collection of personal data is the very core of the
success of the underlying business model. Yet the
fact that rms nowadays know that, in an emerging
data economy, any data may become interesting and
that they may be able to commercialise that data
based on factual exclusivity, it cannot be argued
that there is suboptimal generation and collection
of digital data. In general, data are not a scarce
resource.89 The sheer volume and variety of big data
constitute the basis but also the particular challenge
of big data analytics.
79
Hence, there is not sufcient evidence of the need of
data ownership as justied by the incentive theory
concerning the generation and collection of data.
However, there could be a need for more incentives
to invest in tools for technologically challenging big
data analyses. Within the value stream of exploiting
data, data analyses generate major social value by
producing new knowledge and thereby optimise
decision-making in many elds. However, although
the evolving business models of big data analyses
may still be in need of further research, it seems
that data ownership will not be the appropriate
mechanism for protecting the interest of big data
analysts. Access to data held by others should be
more of a concern to big data analysts than acquiring
ownership in data. It is more important for big data
analysts that the data they have access to respond to
the challenges of velocity and veracity than having
claims against third parties for unauthorised use
of the data they produce. Since, in many instances,
real-time data is key, data analysts do not have to
be so much afraid of competitors’ free-riding. What
counts more is getting access to the various datasets
from which they can gather new knowledge. As
regards the other side of the market, namely, the
rms and public entities to which big data analysts
provide new knowledge for optimising decision-
making, data ownership will not be needed either.
Such relationships will often be based on contracts
for services through which customers are supplied
with accurate knowledge at a given point in time.
From a competition perspective, the core question
is whether data analysts need to rely on data
ownership in competition with other data analysts.
89 See also Becker (supra n 25) at 7.
2017
Josef Drexl
274
4
This question has to be answered in the negative.
Data analysts will not gain a competitive edge by
‘owning data’ at the expense of their competitors.
Rather, they will prevail in competition if they
manage to have better access to the various sources
of big data, for which they will not rely on ownership
but contractual business relationships with the
holders of such datasets, on the one hand, and
the effectiveness and accuracy of their big data
analyses, on the other hand. As regards the latter,
it is more important that big data analysts control
the technology for big data analysis. For this, they
will rely on copyright protection in the software
infrastructure and possibly technical know-how
rather than data ownership.90 The same holds true
for rms that deliver—typically software-based—
tools for big data analysis of other rms.
80
At the last stage, the customers to whom information
is delivered based on big data analyses are not in
need of data ownership either. To the extent that
these data are kept secret and the data analysts
are under a contractual obligation to keep that
information secret, this information may enjoy trade
secrets protection. Public entities as customers of
big data analysis services will be less likely to have
an interest in keeping the result of big data analysis
secret. In the framework of emerging laws on open
data, public institutions may even be under an
obligation to provide access to the data both to the
public and, pursuant to public-sector-information
(PSI) laws, for commercial re-use by private actors.
II. Incentives for the
commercialisation of data
81
Another and more modern justication for
property rules is the goal of creating incentives
for the commercialisation of the subject-matter
of protection. In the context of patent law, this
is often called the ‘prospect theory’—in contrast
to the traditional incentive theory, whereby the
latter is designed to reward those who invest
in the generation of the subject-matter for that
investment.91
82
In general, innovation does not end with the
generation of the subject-matter of protection
and the acquisition of the IP right. Innovation will
90 Against a justication of patent protection for the algorithm,
see Josef Drexl et al. (supra n 39), paras 12-17.
91 The foundations of the prospect theory were laid by
Edmund Kitch, ‘The Nature and the Function of the Patent
System’ (1977) 20 J L & Econ 265. On a more modern market-
related patent theory that departs from the classical reward
theory, see also Daniel F Spulber, ‘How Patents Provide the
Foundation of the Market for Inventions’ (2015) 11 J Comp L
& Econ 271.
only serve society if it reaches the market. And
quite often more investment will be needed for
the commercialisation of the subject-matter of
protection than for its generation.
83
A good example of this can be observed in the
pharmaceutical sector. The major investment
that goes into the development of drugs relates
to the nancing of the lengthy and risky clinical
trials, which typically take place after the ling
of patents. Indeed, in order to protect investment
in the clinical trials against free-riding by others,
the pharmaceutical company is in need of patent
protection prior to making that investment. In most
cases, the patent holder will also be the rm that
conducts the clinical trials and brings the product
to the market. However, the patent holder may also
decide to license the patent to another company
that, based on that licence and with the prospect
of having a secured market later on, will make the
investment in developing the drug.
84
Similarly, investment in the commercialisation of
copyrighted works is not typically effectuated by the
creator, but by the representatives of the copyright
industries, such as publishers and producers. Only in
countries that follow a work-made-for-hire doctrine
will the latter be considered initial copyright
owners, whereas in other countries they can rely
on exclusive copyright licences or, at best, related
(neighbouring) rights.
85
These examples show that the original right does
not necessarily have to be vested in the person who
makes the investment in the commercialisation.
The licensing system, based on contract law
and exclusive licences, can provide for the same
incentives. Granting the original right at the stage
of the creation of the content, however, may produce
additional distributional effects. The copyright
protected in favour of the creator may generate
additional income for the creator, at least if there
are additional rules in place that guarantee fair
remuneration.
86 As regards the data economy, however, no case for
recognising data ownership can be identied based
on the goal of producing additional incentives for the
commercialisation of the data. The major argument
is that the holders of data do not have to be afraid
that competitors will free-ride on investment in the
commercialisation of their data. Likewise, there is
not any particular risk that the data will be copied
by competitors for the purpose of substituting the
data holder’s offer, nor does the grant of access to
the data to others, such as big data analysts, involve
particular investment by the data holders.
87 Nor are the big data analysts unable to recoup their
investment in the commercialisation of their data
Designing Competitive Markets for Industrial Data
2017
275
4
without data ownership. They are much more likely
to rely on the control of their software solutions
to protect their innovation under competitive
conditions.
88 The situation is likely to be different as regards data
brokers. Data brokers can play an important role
in the enabling of big data analyses in particular.92
Data brokers may also act as aggregators of datasets.
Property rights have the potential of stabilising their
activities. However, these brokers can also rely on
factual exclusivity regarding the control of datasets
that are transferred to them. Concerning situations
where real-time data is key, data brokers are less
likely to act as intermediaries that buy and resell
identiable datasets. They are more likely to act as
agents that bring together providers of large and
dynamic datasets with customers that are interested
in services that build on big data analyses. Such
brokers will enable direct transactions between data
providers, on the one hand, and big data analysts
and their customers, on the other hand. To do this,
they are not in need of property rights in the data.
III. Data ownership as a means
to stabilise transactions
89
Property rights can also stabilise and, thereby,
facilitate transactions. Conversely, this is an effect
which cannot be provided in the framework of trade
secrets protection. Transactions on trade secrets
suffer from major instability. Every sharing of
trade secrets increases the risk that the information
will ultimately become publicly available with no
possibility for the holder of the trade secret to act
against the re-use of that information.
93
Accordingly,
recognition of data ownership is advanced as a
means to facilitate trading with data as a commodity.
The argument is that, even where there is factual
92 See Federal Trade Commission, ‘Data Brokers—A Call for
Transparency and Accountability’ (2014), available at:
reports/data-brokers-call-transparency-
accountability-report-federal-trade-commission-may-
2014/140527databrokerreport.pdf> (accessed 12 September
2016). The business models of data brokers were however
heavily criticised in the US in particular, where those
brokers have contributed to the spread of personal data
and provided uncontrolled access of the government to
personal data. See Chris J. Hoofnagle, ‘How ChoicePoint and
Other Commercial Data Brokers Collect and Package Your
Data for Enforcement’ (2004) 29 NC J Int’l L & Com Reg 595.
93 According to Art 3(3) of the new EU Trade Secrets
Directive the ‘use’ of trade secrets is only unlawful under
rather restrictive conditions, namely, when the user has
acquired the information unlawfully or is in breach of a
condentiality agreement or any other agreement on how
to use the information. Once the trade secret has become
known to third persons, these persons can lawfully use the
information.
exclusivity, without ownership there are no direct
remedies against unauthorised use by third persons
once the data has been disclosed.94
90
Yet considering the risk that business models will
be undermined by unwanted free-riding in an
environment in which the availability of real-time
data is key, this argument of stabilising transactions
will hardly ever be very convincing.
IV. Legal certainty
91
Another argument relates to legal certainty. Clear
attribution of ownership can enhance legal certainty
by informing the stakeholders about their rights and
obligations.
92
This, however, is not very convincing as regards data
ownership either. On the one hand, new property
rights will always give rise to additional conicts and
litigation. At the same time, allocation of property
rights may not be so clear at all. As regards data
ownership that is recognised independently of
factual control over data in an environment where
individual data may constantly be integrated and
arranged in different datasets, data ownership is
more likely to reduce transparency and increase the
risk of unintentional infringement of rights.
V. Ownership as a means
to enhance access
93
A nal potential justication for data ownership may
look counterintuitive at rst glance, but in particular
deserves closer attention.
94
As has already been explained above in the context of
the discussion of the UsedSoft decision of the CJEU,95
property rights regimes can also be used as a means
to enhance the free ow of data. In this decision, a
limitation of copyright protection regarding digital
downloads was used as a means to promote free
circulation of digital copies of computer programs.
95
This example shows that general recognition of
property rights can also make sense where factual
exclusivity is already particularly strong. Adoption
of a fully-edged rights regime can include far-
reaching mandatory exceptions and limitations that
cannot be set aside by contractual restrictions.96 For
94 See, in particular, Zech (supra n 24) at 60.
95 At C.V. above.
96 See also Becker (supra n 25) at 9 (assuming that the industry
may even refuse to claim new legislation on data ownership
since such legislation could provide more access than they
2017
Josef Drexl
276
4
instance, such exceptions and limitations can also be
found in the French legislation on the exclusive right
of sports associations as regards the audiovisual
exploitation of sporting events.97 Hence, such
ownership systems could provide better guarantees
for access than reliance on general contract law
based on the unrestricted principle of freedom of
contract.
96 However, this approach is not without alternatives.
Access can also be guaranteed by special legislation
on access that takes precedence over contractual
restrictions. As regards the commercial exploitation
of sporting events, such access rules can be included
in the general media law. Current EU law also
enhances access to information held by public
bodies. Thereby, the European rules on public sector
information do not have to recognise ownership of
public bodies in the information they hold in order
to regulate the principles that apply to the licensing
of the commercial re-use of such information.98
97 An interesting case is also presented by the current
proposal of the Commission to introduce an un-
waivable exception to copyright protection for
carrying out text and data mining for the purpose
of scientic research.99 This proposal seems to prove
the case that exceptions promoting access to data can
easily be drafted within existing ownership systems.
However, separate access legislation on data mining
could also be drafted by building on the model of the
proposal with application beyond copyright and with
regard to other interests whenever the data holder
has granted access to somebody in the framework of
a contractual agreement. To do this there is no need
to recognise data ownership up front.
98
An additional argument against adopting ownership
as a means to enhance access arises from challenges
regarding the form of regulation of such exceptions
or limitations. There are two approaches, both of
which are problematic. The rst approach consists
in a general clause similar to the fair use exception of
US copyright law.
100
This approach has the advantage
of general applicability but the disadvantage of lack
of precision. It would hence cause legal uncertainty,
give rise to legal disputes and potentially favour the
interests of those parties that have less of a problem
to nance litigation. As regards data ownership in
particular, this approach has the additional drawback
that it would have to be formulated in an extremely
general way in order to be adaptable to the very
currently are willing to provide under contract law).
97 See Arts L333-6 through L333-9 Code du sport.
98 See PSI Directive (supra n 21).
99 Art 3 Commission Proposal of 14 September 2016 for a
Directive of the European Parliament and of the Council
copyright in the Digital Single Market, COM(2016) 593 nal.
100 See Sec 107 US Copyright Act.
different sectors of the data economy. Hence, it is
very doubtful whether such a ‘fair use’ clause would
really be able to enhance access in practice.
99 The second approach would consist in formulating
a precisely dened exhaustive catalogue of
exceptions and limitations that takes care of specic
countervailing interests. However, this would require
the legislature to fully anticipate the interests of a
large number of potential stakeholders in highly
diverse sectors of a data economy that is only just
emerging.101 There is a clear risk that legislation
on exceptions and limitations would largely be
postponed to the future, while the legislature would
immediately adopt a strong rights system that goes
beyond the restrictions data holders can implement
under contract law. In sum, this approach would
entail the risk of largely hampering the free ow
of information without sufcient remedies for
addressing problems of access.
100 In addition, balancing conicting interests is more
difcult for the legislature, where the question of
who should be the owner remains a most difcult
issue.102 Whomever the legislature singles out as the
right-holder, this will produce an additional negative
impact on the interests of other stakeholders and
may intensify a conict of interests. In contrast,
by choosing the alternative approach of balancing
factual control over data by access-only legislation,
the legislature will react to the conict as it arises
from the specic context of the market without
intervention.
101
In sum, it seems more advisable to prefer an approach
of progressive adoption of access regimes as part of
sector-specic regulation. Such an approach could
still develop principles and guidelines that emerge
over time and ultimately rely on general models of
regulation.103
102
It can be thus concluded that no reasons can be
identied that would argue in favour of introducing
data ownership in favour of any of the stakeholders.104
E. Problems related to the design
of the rules on data ownership
103
Since there is no clear case for introducing
legislation on data ownership, the question of how
to design such legislation is not even relevant. Yet,
101 On the many and very context-dependent stakeholders in
the data economy see at B.IV. above.
102 See at E.I. below.
103 On this see at F.IV. below.
104 Also against adopting legislation on data ownership, Wiebe
(supra n 25) at 884.
Designing Competitive Markets for Industrial Data
2017
277
4
some challenges regarding such legislation should
nevertheless be addressed since, in the current
debate, it seems that these challenges are not
sufciently discussed105 and, consequently, largely
underestimated when the idea of data ownership is
advanced.
106
There are many reasons why the design
of such protections is enormously complex. Several
dimensions of this problem can be identied:
I. Complexity of the legal issues
104
For any intellectual property rights system, a
decision has to be made on what subject-matter is
to be protected, on who should own the right, and
on the scope of protection, including the exceptions
and limitations. As to the latter aspect, a decision is
to be made regarding the terms of protection.
105
As regards the subject-matter of protection, it has
already been mentioned that the law has to decide
whether data should only be protected on the
syntactic or also on the semantic level. The latter
should rather be avoided because of the risk of
obstructing the free ow of information.
107
However,
the question still remains whether data can be
protected as ‘raw data’ on the syntactic level. This
is questioned because data is in need of specication
on the semantic level in order to qualify as subject-
matter of protection beyond the encoding in the
form of bits and bytes.108 If, however, protection
was granted on the semantic level, the very practical
problem is to identify whether information is
‘new’.
109
Another issue is whether data ownership
should relate to individual data or datasets in their
entirety. The latter would follow the example of the
Database Directive with all its shortcomings, namely,
that it fails to protect the individual data. Yet, if each
and every individual piece of data were protected,
data ownership of individual persons in a world of
big data would disappear like drops of rain in the
sea. Such a system would present major challenges
in terms of its governance and of the enforcement
105 See, however, the discussion of a data producer right by
Zech (supra n 24) at 74-78.
106 This is also true for EU Commissioner Oettinger. His idea of
a ‘data use right’ does not explain what this right should
protect, who should be the owner and how far protection
should go.
107 See also Zech (supra n 24) at 74 (delineating his data
producer right only on the syntactic level). For a review of
different proposals see Wiebe (supra n 25) at 882.
108 Wiebe (supra n 25) at 883.
109 See Wiebe (supra n 25) at 882, highlighting that this
requires a showing that the same information has not been
stored before in form of 0s and 1s. In addition, it ought to be
remembered that the same information can be represented
differently on the syntactic level, for instance, in a different
language or a different form (eg, a video and not a text).
of myriad individual rights, not to mention the
challenges for users in the context of rights-clearing.
106
As regards potential owners, it has been shown in this
analysis that in a complex world of networks where a
considerable number of different players collaborate
in generating value, not least by contributing their
data, the allocation of data ownership is particularly
difcult.110 Furthermore, if everybody contributing
to the generation of data in a value network is
vested with ownership, this allocation could easily
run the risk of creating too many property rights,
which would block efcient exploitation of big data
in particular.111 The proposition to vest consumers
with the ownership of their personal data in order
to enhance trading with that data as a commodity112
does not explain why allocating the economic value
to consumers can be justied from an economic
perspective.113
107 Moreover, the denition of the scope of protection
also remains a difcult task. It is not clear at all
in which situations there is a particular risk that
the need for investment will be distorted by the
free-riding of third parties. The proposal to limit
protection to the copying of encoded information,
while allowing for the re-generation of the same
data,114 would only conrm that data should not be
protected on the semantic level of information.
108 The denition of the subject-matter of protection,
the identication of the owner of the right and the
scope of protection will be most relevant for nally
identifying the need for exceptions and limitations.
In the light of the large number of stakeholders, it
would be particularly difcult to clearly identify
the conicting interests and to design rules for
balancing these interests.
109 The interaction between all of these issues reaches
an enormous level of complexity, which argues in
favour of preferring legislation on access regimes
to the implementation of a fully-edged new
ownership system.115
110 This is considered a main counterargument against devising
a property right in data according to Wiebe (supra n 25) at
883.
111 See also Wiebe (supra n 25) at 883 (against co-ownership
because of the conicting interests).
112 See, in particular, Zech (supra n 24) at 60.
113 This is also conceded in principle by Zech (supra n 24) at 69.
114 In this sense Wiebe (supra n 25) at 882.
115 See also discussion on adopting an ownership regime as
a vehicle for promoting access through exceptions and
limitations at D.V. above.
2017
Josef Drexl
278
4
II. The one-size-fits-all issue
110
In addition, legislation on data ownership would
have to respond adequately to very diverse
circumstances in which data is generated and used
in the future. The data economy and the use of smart
products are predicted to enter all different elds of
modern life. However, data collection as regards the
operation of smart cars is very different from the
processing of data in the healthcare sector. Whether
it is possible ex ante to conceive uniform rules on
the subject-matter of protection, the person owning
the rights and the uses that will be covered by the
right, while the peculiarities of different sectors are
delegated to exceptions and limitations, remains
rather doubtful.116
III. The dynamic character
of the data economy
111 Several times it has been underlined in this analysis
that the data economy and big data in particular,
is not about stable datasets but about the ‘moving
target’ of highly dynamic data. ‘Velocity’ and
‘veracity’ are a fundamental concern in this economy.
112
This however questions the very appropriateness
of a property approach to regulating that economy.
IP systems are largely based on the paradigm of
protecting intangible assets, such as technologies in
particular, that play a role as input in the production
of physical goods. Such a paradigm does not seem
to t a world in which customers have to rely on
real-time and accurate information as an input.
This contradiction becomes most obvious if one
addresses the issue of the terms of protection. In an
environment where it is key to capture the moment
and where being late leads to wrong decisions, asking
the question of how long data should be protected
will simply miss the needs of this economy. Rather,
the starting point of any legislation should be a clear
analysis of the emerging new business models and
the question of what kind of protection rms need
in order to make their business models successful in
competition with other rms and in the overarching
interest of society.
113
As a matter of principle, contract law seems to
provide the better regime for such protection. It
allows the parties to specically design the rights
and obligations as needed for making new business
models work. Contract law provides the parties
with the possibility to experiment with different
arrangements over time and with the exibility to
adapt to different circumstances in very different
sectors of the data economy.
116 Similar doubts are expressed by Wiebe (supra n 25) at 884.
F. Regulating access to data
114 However, contract law cannot be expected to make
the data-driven economy work without frictions.
Contract law will only work in instances where the
holder of data has an economic interest in sharing
the data with others and where the bargaining
power of the contracting parties is equally strong.
Hence, the question arises whether government and
legislative action is needed to promote access.
115
From the outset, it has to be clear that a refusal
to grant access by itself is not sufcient to justify
intervention. In line with the rationale of trade
secrets protection, such refusal should not be
considered illegitimate where exclusive control over
data provides rms with a competitive edge over
others and, thereby, creates the necessary incentive
to invest in data-based business models. This also
means that the leading rms of the data economy
such as Google and Facebook should not blindly be
forced to share their user data, the most valuable
asset they have to conduct their business.
116
Striking the balance between access to and legitimate
control of data is hence a most difcult task. The
eld of law that rst comes to mind to tackle the
issue is competition law. In this regard, a more
thorough analysis of competition law is needed
in order to assess competition law’s potential to
provide a workable access regime. For this purpose
and as a preliminary clarication, it is important to
place competition law as a tool for enforcing access
to data in the context of the current competition
policy debate on big data (section F.I. below). This
will be followed by an analysis of the potential
application of rules of EU competition law to refusals
to grant access to data (section F.II. below). This
analysis will help in discussing additional actions
that could enhance access to data (sections F.III. and
F.IV. below).
I. The current competition
law debate on big data
117
The debate and literature on how and whether
competition policy should react to the advent of
big data has exploded within a remarkably short
period of time.117 The discussion is mostly driven by
the enormous success and expansion of rms in the
digital economy such as Google or Facebook, whose
business models are largely built on the control of
user data. There is in fact growing awareness that
control over big data should play a more prominent
117 Among the major and most recent contributions from
competition law scholars are Rubinfeld and Gal (supra
n 28); Maurice E Stucke and Allen P Grunes, Big Data and
Competition Policy (Oxford: Oxford University Press, 2016).
Designing Competitive Markets for Industrial Data
2017
279
4
role in assessing market power and dominance,
not least in the framework of mergers.118 The EU
merger cases of Google/DoubleClick
119
and Facebook/
WhatsApp120 are among the rst cases where control
over user data in terms of ‘data concentration’121
was taken into account for assessing the effects of
mergers on the online advertising market.122 Yet in
both cases the Commission held that the emerging
data concentration was not sufcient to signicantly
impede competition in this market.123 The growing
role of data in the digital economy has also convinced
competition law enforcers to further develop their
policies as regards the impact of control over data
on competition.124
118 Yet this discussion on how competition law should
react to the challenges of the data economy and big
data is based on a particular perspective. First, control
over data is considered to be a potential competition
problem. This corresponds to the general role of
competition law to ban anti-competitive conduct.
Second, the focus is very much on market structure,
market power and dominance,125 as well as on market
118 See, for instance, Inge Graef, ‘Market denition and market
power in data: the case of online platforms’ (2015) 38 World
Competition 473.
119 Commission Decision of 11 March 2008, Case No
COMP/M.4731—Google/DoubleClick, available at:
ec.europa.eu/competition/mergers/cases/decisions/
m4731_20080311_20682_en.pdf> (accessed 10 September
2016).
120 Commission Decision of 3 October 2014, Case No
COMP/M.7217—Facebook/WhatsApp, paras 164-67 and 181-91,
available at:
cases/decisions/m7217_20141003_20310_3962132_EN.pdf>
(accessed 10 September 2016).
121 Facebook/WhatsApp (supra n 120) para 164.
122 From the perspective of the data economy, the Commission
Decision of 4 September 2002, Case No COMP/M.6314—
Telefónica UK/Vodafone UK/Everything Everwhere/
JV, available at:
cases/decisions/m6314_20120904_20682_2898627_EN.pdf>
(accessed 10 September 2016) may even be more interesting.
In this case, the Commission assessed the impact of the
joint venture for the introduction of an electronic payment
system (‘mobile wallet’) on the market for data analyses.
123 In Facebook/WhatsApp, the Commission specically looked
at WhatsApp as a potential source of user data for better
targeting Facebook’s advertising activities. It nally
concluded that even if Facebook implemented such a policy
post-merger, it would only control a small share of user data
on the Internet as a resource for online advertising. See
Facebook/WhatsApp (supra n 120) paras 180-89.
124 See, in particular, the joint policy paper by of French and
German competition authority: Autorité de la concurrence
and Bundeskartellamt, ‘Competition Law and Data’ (10
May 2016), available at:
de/SharedDocs/Publikation/DE/Berichte/Big%20Data%20
Papier.pdf?__blob=publicationFile&v=2> (accessed 10
September 2016).
125 In their joint policy paper on data, the French and German
competition authorities devoted the whole second half to
the role of data for assessing market power. See Autorité de
entry barriers arising from the control of big data.
126
This is explained by the fact that anti-competitive
effect, especially in unilateral conduct cases,
depends on the ability to behave independently of
the competition.
119 Within the framework of the current ‘Free Flow of
Data’ initiative of the Commission, however, the role
attributed to government is a more proactive one
of industrial policy. The question is not only how
to protect the free market economy against anti-
competitive conduct of rms. Rather, the question
is what can be done in order to promote the digital
economy.
120
In this regard, competition law has certain
advantages but also shortcomings. On the positive
side, competition law is in principle applicable
to all sectors of the economy that are currently
undergoing a digital transformation. Competition
law can work as a platform on which legislatures
can build to formulate more targeted, sector-
specic rules whenever competition law does not
provide sufcient remedies. In addition, competition
policy and law can also prevent the legislature from
excessive intervention. In instances where there is
no identiable harm to competition, policy makers
will have to look for an alternative justication for
adopting access rules.
121 On the negative side, competition law is likely to be
too limited to provide sufcient remedies. As regards
its substantive criteria, competition law only reacts
to one particular kind of market failure. Intervention
is only justied where there is identiable harm to
competition. While the outer boundaries of what
can be considered such harm is not at all clear, there
are kinds of market failures that cannot specically
be addressed by competition law. For instance, in a
world of big data analytics involving techniques of
data mining by searching datasets for correlations,
negotiations about access to data may simply fail
because of information asymmetries regarding
the value of the data.127 From an institutional
la concurrence and Bundeskartellamt (supra n 124) at 25-52.
126 See, in particular, the thorough analysis of potential
barriers to entry caused by big data by Rubinfeld and Gal
(supra n 28).
127 This is known as the ‘information paradox’. Contractual
negotiations on data as a commodity can easily fail because
the purchaser, not knowing which information can be
extracted from the data, will not be able to assess the
value of the data. If, however, the data is made accessible
to the prospective purchaser for solving the information
problem, the purchaser will no longer be willing to pay
for access. The ‘information paradox’ was rst framed by
Arrow in the context of patent law. See Kenneth J Arrow,
‘Economic Welfare and the Allocation of Resources for
Invention’ in: National Bureau of Economic Research (ed),
The Rate and Direction of Inventive Activity (1962) 609. But it
is also to be noted that markets can provide solutions to
2017
Josef Drexl
280
4
perspective, competition law enforcers are able
to ban identiable anti-competitive conduct, but
they are not well equipped for regulating markets
ex ante by imposing positive rules of conduct in the
form of behavioural remedies that require on-going
monitoring.
122
Hence, already based on these general observations,
it is very likely that actions will be needed that go
beyond competition law. But competition law should
be placed at the beginning of the following analysis
(section F.II. below). Competition law thinking as a
market-compliant approach will however also prove
important for devising additional pro-competitive
regimes that promote access to data (sections F.III.
and F.IV. below).
II. Addressing refusals to
grant access to data under
EU competition law
123 EU competition law has not yet developed specic
case-law on access to data in the data economy
that is only now about to emerge. However, as the
following analysis will show, the practice on refusals
to deal and, more concretely, refusals to license can
produce some indications on how to assess future
data-related cases. At the outset, it should be noted
that it is not important whether data to which access
is requested is protected by intellectual property (IP)
rights or not.
128
Even in cases in which neither IP
protection nor trade secrets protection is available,
but the holder of data can rely on factual exclusivity
provided particularly by technological protection
measures, a refusal to grant access can be captured
as a refusal to deal under competition law. For the
assessment of such cases, under Article 102 TFEU,
the question is whether the holder of data is market
dominant and whether the refusal to grant access
to data constitutes an abuse. These issues will be
addressed in the framework of the following review
of the existing case-law.
the information paradox. For instance, data analysts can be
appointed as trustees to do tests on the utility of datasets
for the purposes of a prospective customer to assess the
value of the dataset, without providing direct access to the
information contained in the datasets to this customer.
128 In the Microsoft case, which was on access to the
interoperability information contained in the Windows
program, both the Commission and the General Court
(GC, former Court of First Instance) left open whether
this information was IP-protected or not and applied the
test developed by the Court of Justice of the EU (CJEU) for
refusals to license an IP right. See Judgment in Microsoft v.
Commission, T-201/04, ECLI:EU:T:2007:289, [2007] ECR II-
3601.
124
The three major cases that established the
foundations for assessing refusals to license, namely,
Magill,129 IMS Health130 and Microsoft,131 are all, in one
way or another, ‘information-related’. Beyond these
three cases, the following analysis will also take into
account the more recent Huawei case, which dealt
with a refusal to license a standard-essential patent
(SEP).132
1. The requirement of dominance
125 For cases regarding access to data in the context of
the currently emerging data economy, Magill and
Microsoft are most suitable precedents. In both cases,
the holder of information that was indispensable
for entering a downstream market refused to grant
access to that information. In Magill, the TV stations
broadcasting in the Republic of Ireland and Northern
Ireland refused to grant a copyright licence for their
TV listings and thereby excluded a publisher from
the market who intended to offer comprehensive
TV guides to consumers. Microsoft is perhaps an
even better precedent for refusals to grant access
to data because, in this case, the interoperability
information for the Windows operating system as
such was not freely available to the competitors
in the market for work group server operating
systems.133 Yet Magill laid the foundations for dealing
with the issue of information-based dominance. The
Court convincingly stated that, due to copyright
protection, the TV stations were the only source
of the relevant information and that, therefore,
the three TV stations had to be considered as de
facto monopolists with regard to the information
contained in their respective TV listings.134 The
situation in Microsoft was very similar. However,
market dominance did not arise from an IP right,
but from the fact that Windows, based on network
effects, had emerged as a de facto standard in the
market for operating systems, which made the
129 Magill (supra n 27).
130 Judgment in IMS Health, C-218/01, ECLI:EU:C:2004:257, [2004]
ECR I-5039.
131 Microsoft (supra n 128).
132 Judgment in Huawei, Case C-170/13, ECLI:EU:C:2015:477.
133 Art 6 Computer Programs Directive (supra n 79) allows for
decompilation of an existing computer program where
this is necessary to obtain interoperability information
for the purpose of establishing interoperability for an
independently created computer program. However, this
exception and limitation is insufcient in a modern software
environment, where the interoperability information can
constantly be changed by updates. Hence, competition
law may still be needed to order the dominant holder of a
computer program to provide access to the interoperability
information. Recital 17 of the Computer Programs Directive
explicitly safeguards the applicability of EU competition
law in such instances.
134 Magill (supra n 27) para 47.
Designing Competitive Markets for Industrial Data
2017
281
4
interoperability information an indispensable input
for offering interoperable programs that would run
on Windows.
126
The two cases demonstrate that it is easiest to
show dominance in data-related cases where the
petitioner seeks access to concrete information that
is indispensable for doing business in a market.
127
More typical for the data-driven economy are
however cases in which somebody, such as a big data
analyst, seeks access to large datasets for purposes of
data mining. In the light of its utility, namely, to rely
on statistical correlations among different pieces of
information contained in larger sets of aggregated
data for generating new knowledge, such datasets
have to be considered a kind of resource which is
distinct from concrete semantic information such
as in the case of Magill. Yet the test of Magill, as an
expression of general competition law principles,
can be adapted to meet the challenges of cases
that deal with access to large datasets to enable
big data analyses. The test in both cases is whether
the respective dataset can be considered the ‘only
source’ of the resource.
128
This leads to the issue of substitutability of datasets.
The fact that data are non-rivalrous and, therefore,
individual data could be found in various datasets
seems to count against dominance. Whether datasets
are substitutable or not will depend on the concrete
circumstances, including the very nature of the
information contained in the data. If, for instance,
a supplier of parts wants to have access to the data
collected by the end manufacturer after the sale of
the nal product to control the quality of its parts,
the end producer’s datasets will indeed be the only
source of that data. However, if a city is in need of
information about the qualities of streets which is
collected by smart cars, different car manufacturers
may be able to provide access to that information
through their datasets. The reason is that the latter
kind of information is freely available in the public
in the rst place, and, hence, can be duplicated in
the datasets of any other data collector. Publicly
accessible information is by nature non-rivalrous135
and can therefore be registered by anybody in a
digital format.
129 Yet assessing dominance in a world of big datasets
by using the concept of substitutability remains a
most difcult task, since even the petitioner for
access, such as a big data analyst, will often only
have a vague understanding about the kind of data
contained in the dataset and about which data will
produce the most valuable new information based
135 The character of non-rivalry of data is also highlighted by
Autorité de la concurrence and Bundeskartellamt (supra n
124) 36-37.
on observable correlations.
130
However, larger collections of data will generally
guarantee a higher level of accuracy of the new
information, since such information derived from
correlations of data within such datasets is based
on statistical likelihood. Hence, just as in the case
of multisided platform markets, the collection of
datasets for the purpose of enabling big data analysis
may exercise particular network effects and enhance
market power of the rm that controls access to the
larger dataset.136 The same may occur in the case
of data-sharing platforms. An example of such a
platform is provided by the joint venture of the
three German car manufacturers, Daimler, BMW and
Audi, that acquired Nokia’s digital map HERE as an
important element of their systems for autonomous
driving. For instance, such digital platforms could
be used for collecting and exchanging real-time
information about the weather conditions of roads.
The quality and reliability of such an information-
sharing platform would obviously increase with
the number of cars contributing information to this
system. Accordingly, the three car manufacturers
should have a strong self-interest in convincing
other car manufacturers to join the system.
137
At the
same time, this may tip the market and give rise to
market dominance of the joint venture.
2. The four requirements for abuse
according to Magill and IMS Health
131
The two cases of Magill and IMS Health have
established the European test for assessing whether
a refusal to license constitutes an abuse. In IMS Health
this test was phrased as one with three cumulative
conditions, which, however, contained the additional
underlying condition that the resource to which
access is sought be indispensable for conducting a
business.
138
In Microsoft, the General Court rephrased
this test in a better and more structured way.139
According to the Court, the following four conditions
for a refusal to license need to be fullled in order to
present ‘exceptional circumstances’ for considering
the refusal an abuse:
136 See also Rubinfeld and Gal (supra n 28) at 42.
137 Indeed, when the Bundeskartellamt, the German
competition agency, cleared the acquisition under
German merger control law, it specically considered
that other car manufacturers would not be excluded
from participating in the system. See Bundeskartellamt,
BMW, Daimler and Audi can acquire Nokia’s HERE
mapping service’ (6 October 2015), available at:
www.bundeskartellamt.de/SharedDocs/Meldung/
EN/Pressemitteilungen/2015/06_10_2015_HERE.
html?nn=3591568> (accessed 10 September 2016).
138 IMS Health (supra n 130) para 38.
139 Microsoft (supra n 128) para 332.
2017
Josef Drexl
282
4
(1)
The refusal relates to a product or service that is
indispensable to the exercise of a particular business
in a related (secondary) market;
(2)
The refusal excludes effective competition in that
related market;
(3) The refusal prevents the emergence of a new product
for which there is consumer demand;
(4) The refusal is not objectively justied.
132
In applying these conditions to refusals to grant
access to data and larger datasets in particular,
several issues arise:
133
First, as regards the indispensability requirement,
a problem arises when data relate to information
that it is publicly accessible but can only be found in
a digital format in the datasets of one undertaking.
Since registration and digitisation makes the
information retrievable and treatable, including for
purposes of big data analysis, the digital data should
be considered a product with added value that differs
from the original, publicly accessible information.
Accordingly, the holder of the digital data in such
a situation can indeed be considered a monopolist
and, hence, a potential addressee of Article 102 TFEU.
However, this does not automatically mean that the
data is also ‘indispensable’ in the Magill/IMS Health
sense, since anybody else including the petitioner
could also register the same information in a digital
format.
134 For understanding the concept of indispensability,
the judgment in Bronner is most relevant; although
the case did not deal with access to data but access to
a nationwide home delivery scheme for newspapers.
According to the CJEU in this case, access to a
resource of a competitor cannot be considered
indispensable if there are no ‘technical, legal or even
economic obstacles capable of making it impossible,
or even unreasonably difcult’ to duplicate the
resource.
140
Thereby, the Court showed reluctance
to accept the argument of lack of economic viability
too easily. The Court stressed that it is not enough
to show that duplication of the resource would not
be economically viable against the benchmark of
the petitioner’s scope of business in the secondary
market.141 Rather, the question is whether it is
economically viable to create the resource ‘for
production on a scale comparable to that of the
undertaking which controls the existing product or
service’.142
140 Judgment in Bronner, C-7/97, ECLI:EU:C:1998:569, [1998] ECR
I-7791, para 44.
141 Ibid, para 45.
142 As rephrased in IMS Health (supra n 130) para 28, with
reference to Bronner (supra n 140) para 46.
135
This seems to indicate an objective standard for
indispensability that does not depend on the size
of the petitioner’s business and that imposes on the
petitioner the burden to make the same investment
as the one made by the dominant undertaking.
Regarding cases on refusal to grant access to data,
this may well mean that indispensability cannot
be argued where the information as such is freely
accessible and it is only a matter of registering
the data in a digital form. On the other hand, it
would be easier to argue indispensability where
data is generated through business models that are
characterised by strong network effects such as
search engines and Internet platforms like the HERE
data-sharing system described above. The possibility
to duplicate similarly large and valuable datasets is
excluded by the economic characteristics of these
markets.143
136 Second, the requirement of excluding competition
in a secondary market qualies the European rule
on refusal to licence as one, which is based on a
leveraging and exclusion theory. This presupposes
that the dominant rm is also active as a competitor in
the secondary market. This, however, will frequently
not be the case when rms refuse access to data. It
is a typical feature of the new data economy that
data is collected for one purpose, such as enabling
predictive maintenance services, but turns out to be
interesting for very different purposes pursued by
other rms of a very different sector and even the
government. In such instances, the European rule on
refusals to license and refusals to deal, as developed
in the abovementioned case-law, would not apply.
137
More recently, in the Huawei judgment, the CJEU
clearly indicated that the ‘cumulative’ Magill/IMS
Health conditions are not the only ‘exceptional
circumstances’ to make a refusal to license an abuse.
the CJEU accepted that exceptional circumstances
are also present in the case of a refusal to license an
SEP if (1) the standard was xed by a standardisation
body144 in return for which (2) the patent holder has
irrevocably committed to license on fair, reasonable
and non-discriminatory (FRAND) terms.145 Since the
Court did not reiterate the condition of exclusion
of competition in a secondary market as part of the
description of these exceptional circumstances, the
question may be asked whether a refusal to license
or a refusal to deal can also be considered abusive
if the dominant rm is not vertically integrated.
However, the Huawei decision itself presents many
uncertainties in this regard, because the Court in its
reasoning still indicates that harm to competition
143 This problem of ‘access to data’, though not in the context
of the indispensability test, is also addressed by Autorité de
la concurrence and Bundeskartellamt (supra n 124) at 38.
144 Huawei (supra n. 132) para. 49.
145 Ibid, para 51.
Designing Competitive Markets for Industrial Data
2017
283
4
is conceived as harm through exclusion of a
competitor in a downstream market. In particular,
the Court reasoned that ‘the fact that the patent
has obtained SEP status means that its proprietor
can prevent products manufactured by competitors
from appearing or remaining on the market
and, thereby, reserve to itself the manufacture
of the products in question’.146 From this, one
could conclude that exclusion of competitors in
a secondary market also remains a requirement
in SEP cases. This would indeed be correct if one
accepted the conservative approach to competition
law, according to which competition law can only
promote innovation indirectly, namely, only in cases
in which there is identiable harm to competition
through exclusion.147 In contrast, the Commission
also argued a violation of Article 102 TFEU in the
Rambus case against a non-vertically integrated SEP
holder who tried to extract excessive royalty rates
from the implementers in a case of patent ambush.148
138
This debate, however, may not be very relevant
for cases on access to data. The reasons for this are
twofold. First, those cases do not involve SEPs related
to standards adopted by a standardisation body.
Hence, the alternative ‘exceptional circumstances’
accepted in Huawei will not apply. Second, the
alternative, dealing with refusals to grant access
to data by non-vertically integrated data holders
as a pure case of exploitative abuse in the form of
excessive pricing under Article 102 lit. a) TFEU, would
turn competition law enforcers into general price
regulators. Fullling such a role would particularly
be difcult in cases on access to data in which the
parties also encounter major information problems
as regards the economic value of data contained
in large datasets. Accordingly, it is very unlikely
that a claim of abuse of market dominance will be
successful in a case where access to data is sought
and the holder of those data is not a competitor of
the petitioner in the secondary market in which
the petitioner wants to use those data. This would
exclude reliance on competition law in two very
important sets of cases. The rst case concerns big
data analysts who seek access to data for purposes of
146 Ibid, para 52.
147 This is indeed the approach advocated by Pablo Ibáñez
Colomo, ‘Restrictions on Innovation in EU Competition Law’
= LSE Law, Society and Economy Working Papers 22/2015
(2015), available at:
(accessed 14 May 2016).
148 Commitments Decision of the Commission 9 December
2009, Case COMP/38.636—Rambus, available at:
ec.europa.eu/competition/antitrust/cases/dec_
docs/38636/38636_1203_1.pdf> (accessed 10 September
2016). The Commission’s approach is supported by Josef
Drexl, ‘Innovation as a Parameter of Innovation and its
Implication for Competition Law Application’, Paper
presented at the 11th ASCOLA conference (30 June 2016)
(forthcoming) (in favour of protecting dynamic innovation
competition beyond cases involving exclusion).
data mining. The holders of such data will typically
not be active as competitors in the market of
providing new information generated through big
data analyses. The second case regards cases where
the government seeks access to data in the public
interest. In such cases, a secondary market is missing
in the rst place, since the government will not make
use of that data as an undertaking in the sense of EU
competition law.
139
Third, the question is whether the requirement of the
prevention of a new product (so-called ‘new product’
rule) also applies to cases of a refusal to grant access
to data. According to the General Court in Microsoft,
this is an additional requirement that only applies to
cases involving the refusal to license an intellectual
property right, but not to general refusal-to-
deal cases.149 As demonstrated further above,150
it is very unlikely that data are already protected
by intellectual property rights. The judgment in
Magill, where access to the relevant information was
controlled by a copyright, can only be explained by
the very low standards of copyrightability under the
British and Irish copyright case-law of that time,
which most likely can no longer be maintained
against the backdrop of more recent copyright
decisions of the CJEU.151 To the extent that there
is trade secrets protection, the question is still left
unanswered by the European Courts whether the
test on refusals to license an IP right would also
apply.152 Yet if the European legislature decided to
create a new intellectual property right in data, this
may well make it more difcult to control access
to data based on European competition law since,
then, there should be less doubt as to whether the
additional requirement of the prevention of a new
product applies.
149 Microsoft (supra n 128) para 334.
150 At C. above.
151 The CJEU requires that there be scope for the author to
make ‘free and creative choices’, by way of which the author
‘stamps the work created with his personal touch’. See
Judgment in Football Association Premier League and Others,
C403/08 and C429/08, ECLI:EU:C:2011:631, [2011] ECR I9083,
para. 98; Judgment in Painer, ECLI:EU:C:2011:798, [2011] ECR
I12533, paras 89 and 92; Judgment in Football Dataco v Yahoo!
UK (supra n 46) para 38.
152 In 2005, under the impression of the Microsoft case, the
Commission argued that applying the standard developed
for refusals to license an IP right ‘may not be appropriate’
in cases on a refusal to grant access to interoperability
information that is protected as a trade secret. See
Commission, ‘DG Competition discussion paper on the
application of Article 82 of the Treaty to exclusionary
abuses’ (December 2005), available at:
eu/competition/antitrust/art82/discpaper2005.pdf>
(accessed 10 September 2016). For arguments in favour of
such a distinction see Gintarė Surblytė, The Refusal to Disclose
Trade Secrets as an Abuse of Market Dominance—Microsoft and
Beyond (Berne: Staempi, 2011) 173-210.
2017
Josef Drexl
284
4
140 More specically, in the context of the data-driven
economy, many complex issues would arise in
applying the new-product rule. From the outset, it
is to be remembered that this rule presupposes that
both the data holder and the petitioner for access
are competitors in the same secondary market. Only
under this condition does the question make sense
whether the petitioner for access would offer a ‘new’
product as compared to the product of the dominant
rm. In cases on access to data, the product offered
by the entity that seeks access to data can be
enormously diverse. If it is about use of the data
by big data analysts, the new product will consist
of new knowledge or information, which may then
be offered in a secondary information market. How
to apply the concept of a ‘new product’ in relation
to different information is rather unclear. To argue
that the information produced by the petitioner
differs from that produced by the data holder may
seem convincing at rst glance. However, this is
less clear in the light of the rationale of the new-
product rule, which is based on a balancing of the
interest in protecting competition with the interest
in protecting the intellectual property right.
Accordingly, the new-product rule was devised
to guarantee that the IP right, which is meant to
promote innovation, can only be restricted if the
petitioner for the licence is also an innovator.153
However, whether the generation of (any) new
information can be considered innovation, remains
rather doubtful. Of course, data may also be used
to offer diverse goods and services in secondary
markets. Access to data may especially lead to the
improvement of goods and services. Yet it is not
settled whether any improvement of a product
can be considered a ‘new’ product. In Microsoft, the
General Court seemed to argue this way by pointing
out that, according to Art 102 lit. b) TFEU, there is
not only an abuse when the dominant undertaking
limits production, but also in the case of a limitation
of ‘technical development’ to the prejudice of
consumers.154 It is to be noted that the new-product
rule would also exclude application of competition
law to public entities that seek access to data in the
public interest where these entities do not engage in
any economic activity in the sense of the concept of
an undertaking under EU competition law.
141 Fourth, as regards potential justications, it is still
very unclear whether and what kind of efciencies
can be considered in the framework of an efciency
defence in cases of a refusal to grant access to data.155
153 See IMS Health (supra n 130) paras 48-49.
154 See Microsoft (supra n 128) para 648.
155 See only Stucke and Grunes (supra n 117) ch 19 (at 302-12) on
‘data-driven efciency claims’ (however with a particular
focus on the efciency defence in merger control law).
142
In sum, the analysis of the case-law on refusals to
licence under EU competition law produces a number
of limitations und uncertainties.The requirement to
show market dominance based on control over larger
datasets presents particular challenges for assessing
whether different datasets can be considered as
substitutes. The case-law so far can only be applied
with certainty to vertically integrated data holders,
while, in many instances, the petitioners for access
and the data holder will not be competitors in any
markets. The case-law will not provide any remedy
when government bodies seek access to data in
the public interest. The rule on exploitative abuse
(Article 102 lit. a) TFEU) will hardly ll the gap since
it would require competition law enforcers to act as
price regulators where it is extremely difcult for
the parties themselves to assess the value of data.
Hence, this analysis highlights the shortcomings
and uncertainties of the current state of competition
law to provide adequate remedies against refusals
to grant access to data in the data-driven economy.
3. Access to indispensable
tools for data treatment
143
The analysis so far has concentrated on access where
data or whole datasets are an indispensable input.
However, the European case-law on refusals to
license has more to offer.
144 In IMS Health, the CJEU used the Magill judgment as
a template for assessing a case that nevertheless
presented very distinct features.
156
The reason for
doing this was that an intellectual property right,
namely, a copyright protecting a database, was at
stake and this made IMS Health a refusal-to-license
case similar to Magill.
145 As a precedent for cases relating to the data-driven
economy, it should however be noted that the
subject-matter of copyright protection in IMS Health
was characterised by a particular functionality. The
so-called 1860-brick structure, representing a map of
Germany subdivided into 1860 geographical sectors,
was used as a tool for collecting and treating data
on the sale of drugs. IMS Health was dominant in
the service market for the collection of sales data to
assist the pharmaceutical companies in designing
their marketing activities. A smaller competitor
encountered major problems entering the market
with its own ‘structure’ since the pharmaceutical
companies refused to work with a different structure.
The reason for this was that IMS Health’s brick
structure had emerged as a de facto standard in the
industry, which led the smaller competitor to simply
use the 1860-brick structure; this competitor was
156 IMS Health (supra n 130).
Designing Competitive Markets for Industrial Data
2017
285
4
then sued by IMS Health for copyright infringement
in Germany. In this context, the question arose
whether the defendant could rely on a competition-
law defence.
146 By only looking at the fact that the brick structure
was protected by copyright law, the CJEU missed
the point that the case was indeed one on de facto
standardisation regarding the tools used in the
relevant service market for collecting data. Therefore,
the distinction between two related markets, the
upstream licensing market and the downstream
product market, as well as the application of the
leveraging theory based on an extension of market
dominance from the upstream to the downstream
market, appears rather formalistic.
147
As regards cases on access to data, IMS Health
produces the particular insight that the tools for
treating data have a tendency to emerge as de facto
standards since they allow data to be communicated
between the different market participants involved
at the different levels of the value chain of treating
and analysing data. Use of the same tools in the
industry will produce positive network effects.
On the downside, de facto standardisation will
create access problems regarding the use of these
tools. These tools will regularly be software-based
and hence protected by copyright law. Market
participants that are not allowed to use these tools
will encounter difculties to enter the market for
the treatment of such data.
148
The IMS Health judgment would directly apply to
such cases. From a competition policy perspective,
the CJEU should have given more weight to the fact
that the IP right controlled access to a standard
with a foreclosure effect on competitors. This
places cases such as IMS Health in between Magill and
Huawei.157 The question in such cases is whether the
new-product requirement makes sense in the rst
place.158 Also in Huawei, the CJEU did not require the
prevention of a new product for considering the
refusal an abuse.
149
Of course, the better option would be to promote
standard-setting through standard-setting bodies
and licensing of such standards regarding the
tools for data treatment on FRAND terms. To the
extent that such standards will emerge, the Huawei
judgment would become directly relevant.
157 Huawei (supra n. 132).
158 This has already been questioned by Josef Drexl,
‘Intellectual Property and Antitrust Law. IMS Health and
Trinko—Antitrust Placebo for Consumers Instead of Sound
Economics in Refusal-to-Deal Cases’ (2004) 35 IIC 788.
4. Learning from the judgment in Huawei
150
Indeed, the judgment in Huawei can also provide
inspiration for dealing with cases on access to
data. As regards SEPs, the problem is that patent
holders enter into a FRAND commitment vis-à-
vis the standard-setting organisation (SSO) when
the patents are notied as standard-essential, but
later no agreement can be reached between the
patent owner and the standard implementer on the
concrete royalty rate. Such disputes are prone to
being affected by strategic behaviour by either party
of the licensing negotiations. Since rights-clearing
is enormously difcult in the telecommunications
industry, which is characterised by several thousands
of declared SEPs held by multiple right-holders, to
require users to wait with implementation until they
have cleared all rights would considerably delay
implementation of the standard in the industry.
At the same time, the FRAND declaration creates
a legitimate expectation that the licence will be
granted. However, once the user has started to
implement the standard by producing standard-
compliant goods, the SEP holder may try to
extract excessive royalty rates by challenging the
implementers with claims for injunctive relief (so-
called ‘patent hold-up’). Conversely, if injunctive
relief is not granted at all, implementers can be
tempted to reject any licence offer as non-FRAND-
compliant so as to avoid any payment (so-called
‘patent hold-out’). In order to strike a balance of
interest, in Huawei, the CJEU devised a framework for
negotiations that includes duties of both parties,159
and this may help the parties reach an agreement
without having to call upon the courts or arbitration
tribunals to make a decision on the appropriate
royalty rate.
151
In a data-related access dispute, one of the major
difculties may be that the parties are not easily
able to agree on price. Hence, devising a negotiation
framework for the parties similar to Huawei may
assist the parties to reach an agreement. Such
schemes could be implemented through private
institutions—by way of private ordering—or through
state regulation. This leads the analysis to the design
of additional legislative measures to promote access.
III. Access regimes for existing
contractual relations
152
As regards access regulation, a distinction can be
made as to whether the parties already entertain
a contractual relationship or not. Problems of
access to data may also arise within existing
contracts. The typical justication for legislative
159 Huawei (supra n. 132) paras 60-68.
2017
Josef Drexl
286
4
intervention in contractual relations beyond the
realm of competition law is unequal distribution of
bargaining power.
153 Unequal bargaining power is addressed by different
parts of the law. In particular, the EU has adopted
such rules on consumer contract law in the form
of the Directive on Unfair Contract Terms.160 The
Directive’s scope of application is broad enough
to also control standard contract terms on the
treatment of data. However, there are also particular
shortcomings. First, the Directive’s general clause on
unfair terms does not provide any guidance on how
to assess clauses that relate to the collection and use
of data. The indicative list of unfair contract terms in
the Annex to the Directive does not respond at all to
the modern challenges of a data economy. Second,
since the application of the Directive is limited to
consumer contracts, it fails to create a European
legal framework for addressing the regulation of
access to data in B2B cases.
154
However, as regards both B2C and B2B relations,
there are alternative ways to address cases of
unequal distribution of bargaining power.
155
As regards consumers, there is a considerable overlap
of consumer law with data protection law. The rule
on data portability in Article 20 of the General Data
Protection Regulation161 can rather be considered as
one of consumer protection than of data protection.
While the relevant data covered by Article 20 is
personal data as protected by the Regulation in
general, the purpose of the data portability provision
is not to protect the individual’s moral interests.
Rather, the rule is designed as an access rule that will
enable the individual to switch to other suppliers
where access to the data is crucial for competition
to work.
162
The German Monopolkommission, which,
as a commission of competition experts, fulls an
advisory role to the German government, supported
the right to data portability by stressing that it has
the potential to help the individual overcome a
lock-in effect163 and to react to the problem that
businesses, without ownership regulation in place,
often claim control over personal data as part of
160 Council Directive 93/13/EEC of 5 April 1993 on unfair
contract terms in consumer contracts, [1993] OJ L 95/29.
161 General Data Protection Regulation (supra n 17).
162 The pro-competitive character of this provision was
specically highlighted and praised prior to the adoption
of the Regulation by the German Monopolkommisson
(Monopolies Commission) in its Special Report of 2015. See
Monopolkommission, ‘Competition Policy: The Challenge of
Digital Markets’, Special Report No. 68 (2015) paras S15, S37
and S105, available at:
de/images/PDF/SG/s68_fulltext_eng.pdf> (accessed 10
September 2016).
163 Ibid, at para S105.
their contractual arrangements.164
156 This rule was inspired by the situation of platforms,
including social platforms that rely on user data.
Yet it will prove particularly effective in the
context of new data-driven business models built
on the collection of data. For instance, car insurance
companies have already begun to lower premiums
of customers who accept digital registration of their
driving habits.
165
The possibility to switch to another
insurance company will be considerably enhanced
by the possibility to use such data to prove that the
customer is indeed a careful driver.
157
Since this rule on data portability constitutes a most
suitable form of pro-competitive regulation, there is
no reason why the right to data portability should
be limited to personal data.166 The lock-in effect is
not necessarily restricted to such data.167 Beyond
consumer contracts, a lock-in problem can also arise
with regard to industrial data where suppliers want
to take data with them concerning the quality and
longevity of their parts after the termination of the
supply contract with the manufacturer of the nal
product. Hence, data portability rules should also be
considered for industrial relations.
158 Yet use of access to data as regards the relationship
between suppliers and an end producer could
also be addressed as part of specic competition
164 Ibid, at para S106.
165 On this see, for instance, Adam Tanner, ‘Data Monitoring
Saves Some People Money On Car Insurance, But Some
Will Pay More’ (2 September 2013), available at:
www.forbes.com/sites/adamtanner/2013/08/14/data-
monitoring-saves-some-people-money-on-car-insurance-
but-some-will-pay-more/#7bc2c423264a> (accessed 10
September 2016).
166 The French Parliament has just adopted a provision on data
portability that builds on Art 20 General Data Protection
Regulation (supra n 16) in Art L 224-42 of the Code de la
consommation (Consumer Act) through the so-called Loi
Lemaire (Loi pour une République numérique; Law for a digital
Republic). The law was adopted by the Assemblée nationale
on 20 July 2016 and nally approved by the French Senate
on 28 September 2016; available at:
fr/leg/tas15-131.html> (accessed 30 September 2016).
See comments on Art 12 in the English Explanatory
Memorandum, available at:
numerique.fr/pages/digital-republic-bill-rationale>
(accessed 10 September 2016).
167 Indeed, the new French portability rule is not limited to
personal data. The new Art L 244-43-3 of the Code de la
Consommation (Consumer Code), as amended by the Loi
pour une République numérique, seems to apply to any data
provided by a consumer. However, the rule is also more
restricted than the General Data Protection Regulation in
that it only applies where data are provided to an online
service communication service provider (fournisseur d’un
service de communication au public en ligne). This rule seems to
apply to social platforms in particular, but not necessarily
to a car insurance company, as in the example mentioned
above.
Designing Competitive Markets for Industrial Data
2017
287
4
law regulation. Regulation of supply traditionally
forms part of the Block Exemption Regulation in
the Motor Vehicle Sector.168 In times of the advent
of autonomous driving, a modernised regulation
could also address the treatment of data on the
functioning of the vehicle between the supplier of
parts and the manufacturer of the vehicle. There is a
particular risk that the latter, by relying on superior
purchaser power, will implement contract terms on
data treatment concerning parts that disadvantage
the supplier. The question will be how to implement
such rules within the framework of the Regulation.
While the Regulation will continue to build on the
market-share approach as a basis for the block
exemption, restrictions regarding the access of
data to the disadvantage of the supplier, including
a restriction on data portability, could be included
in the black list of hard-core restrictions. However,
for formulating such a rule, precision is needed in
order to clearly delimit the non-exempted clauses
from those that can be exempted. In particular, one
could imagine a rule that a supply contract cannot
be exempted if it does not include a rule on free-of-
charge data-sharing with the supplier concerning
the functioning of the parts delivered by the supplier.
Such a rule is justied by the fact that both parties
belong to the same network that contributes to the
generation of economic value.169
159 Of course, the issue of access to data by a supplier of
parts is not specic to the motor vehicle industry.
Hence, the Commission should consider creating
a generally applicable access regime in favour of
suppliers in the framework of its block exemption
regulations.
160
Finally, the legislature is free to draft targeted
rules that would ban contractual restrictions on
the use of data under particular circumstances.
The already mentioned Commission’s proposal for
an un-waivable copyright exception for text and
data mining for purposes of scientic research
provides such an example, which could be extended
beyond the realm of copyright and applied for other
purposes.170 In this regard, Article 3(1) Commission
Proposal for a Directive on Copyright in the Digital
Single Market requires that a research organisation
wanting to conduct text or data mining have legal
access—typically based on a copyright licence—to
the relevant subject-matter.
168 Commission Regulation (EU) No. 461/2010 of 27 May 2010
on the application of Article 101(3) of the Treaty on the
Functioning of the European Union to categories of vertical
agreements and concerted Practices in the motor vehicle
sector, [2010] OJ 129/52.
169 On the new paradigm of ‘value networks’ see at B.III. above.
170 See at n 99 above.
IV. Access regimes outside of
existing contractual relations
161
Regimes for access to data outside of existing
contractual relations are more difcult to devise.
In this eld, a more cautious approach is needed in
order to avoid excessive intervention in the market
economy. In addition, the particularities of very
different sectors where data is currently starting to
play a major role in generating economic value from
the outset seems to argue against a regime of general
applicability. On the other hand, designing regimes
for access to data is not an unprecedented exercise
for legislatures. Existing models can be considered
and discussed for cautious generalisations and
potential transfer to other sets of cases.
162
In any event, devising access regimes outside of
existing contractual relations depends on using
certain criteria to balance the interests involved
between exclusivity and access. Such criteria can
be discussed as the kind of information contained in
data, the identity of the data holder and the business
model through which it generates data and, nally,
the person or entity seeking access and the kind of
use this petitioner is intending.
1. Kinds of information
163
As regards the kind of information contained in
data, a rst distinction could be made between
information access to which is in the public interest—
such as information that helps to ght infectious
diseases—and other information in which there
is only a commercial interest. Such a distinction,
however, is very difcult to make, since information
that seems commercial at rst glance may still help
the state to make decisions in the public interest.
Hence, as regards ‘public interest data’, it is better
to address this issue further below in the framework
of the discussion of who is seeking access to data and
for which purpose the data will be used.
164
Yet there are examples where access to information
is promoted by specic legislative means based
on the nature of the information. This is the case
in particular as regards scientic information
contained in publications. Access to such information
is often controlled by academic publishers who
seek an exclusive licence also with regard to the
digital exploitation of the publications. In contrast,
governments increasingly promote open-access
publications. The tools used in this regard can be
very diverse.171 One approach consists of setting
171 As regards the European open access policy see Commission
Recommendation of 17 July 2012 on access to and
2017
Josef Drexl
288
4
nancial incentives. In instances where the scientic
information is the result of government-funded
research, a commitment to open-access publication
of the recipient can be made a requirement for the
grant decision.172
165
Furthermore, open-access regimes can also be
promoted through copyright law. In Germany, the
legislature recently adopted a so-called ‘secondary
publication right’, which vests the author with
an un-waivable right to make the work available
online after an embargo period of 12 months if the
publication is the result of research activity that is
at least 50 per cent publicly funded and provided
that the second publication does not serve any
commercial purpose.
173
The French legislature has
just introduced similar legislation as part of its ‘Loi
Lemaire’ (Loi pour une République numérique).174
166 Such a secondary publication right is characterised
by making use of the interest—namely, in
reputation—of one stakeholder, namely, the author,
to promote open access against the interests of
another stakeholder, namely, the publisher. In doing
so it indirectly benets users, who get unrestricted
benets. Hence, this model has the advantage of
promoting open access much more effectively than
by requiring each and every user to claim access.
This model could be transferred to other sets of
cases where there is conict of interest between two
parties contributing to the information and where
one party in contrast to the other is interested in
open access. One such case regards libraries and
other cultural heritage institutions that cooperate
with private businesses such as Google in the
digitisation of their public domain materials and
works. While the private partner would usually be
interested in exclusive exploitation, the cultural
heritage institution will typically prefer open
access.175
preservation of scientic information, C(2012) 4890 nal.
172 This is also the policy applied by the EU within its
Horizon 2020 research funding programme. See European
Commission, ‘H2020 Programme—Guidelines on Open
Access to Scientic Publications and Research Data in
Horizon 2020’, Version 3.1 (25 August 2016), available at:
h2020/grants_manual/hi/oa_pilot/h2020-hi-oa-pilot-
guide_en.pdf> (accessed 10 September 2016).
173 Sec 48(3) German Copyright Act (entry into effect on 1
January 2014).
174 See Art 17 Loi pour une République numérique (supra n. 166);
see also comments on Art 9 in the English Explanatory
Memorandum (supra n. 166). The French provision however
provides for an embargo period of 24 months, instead of 12
months, for publications in the human and social sciences.
175 In this context, see also Art 11(2a) of the PSI Directive (supra
n 21). As regards public-private partnerships of cultural
institutions with private entities for the digitisation of
cultural resources, this provision limits the grant of an
exclusive license for the re-use of the digitised version to 10
167 It is also to be noted that particular access features
of the secondary publication right are also shared
by the data portability rule of Article 20 Basic
Data Protection Regulation (see section F.III.
above). Moreover, in the latter case, two persons
contributing to the collection and generation of
digital data have opposing views on access of third
parties to the data. In both cases, the law strengthens
the rights of the person in favour of access, which
will indirectly benet third parties. From this
perspective, these rules can be qualied as enacting
partial, pro-access property rights. The legislature
refrains from creating an exclusive ownership
right relating to personal data under the Basic Data
Protection Regulation that would allow the owner
to prevent third parties from using those data,
176
but
still promotes access of third parties based on the
rights of the person from which the data originate.
The un-waivable right is limited to the right to make
the data available to third parties. In this context,
also the recognition of copyright exhaustion for
downloads of computer programs by the CJEU in the
UsedSoft case comes to mind.
177
In this case, ‘access’ in
form of tradability of the programs was enhanced by
recognising ownership in the digital of the program
downloaded by the licensee.
2. The data holder and its business model
168
Another distinction can be made concerning who
holds the data and what business models they use.
Access can be promoted by legal regimes that focus
on particular groups of data holders.
169
Legislatures can in particular promote access to data
where data is held by public institutions as part of an
open-data policy. At the EU level, the Public Sector
Information (PSI) Directive of 2003, in its revised
version of 2013,
178
provides an evolving approach for
the EU to overcome resistance among public bodies
in the Member States to make data more accessible
to the private sector.
170 As part of the Loi pour une République numérique, the
French legislature has just taken further steps to
make data more broadly available by going beyond
public institutions. The Law adopts the concept of
‘data in the general interest’ to expand the open-
data policy to private entities such as public service
concession holders or entities that receive state
years.
176 Similarly, the un-waivable secondary publication right does
not prevent the author from granting an exclusive licence
covering the publication right to the publisher.
177 UsedSoft (supra n 35).
178 PSI Directive (supra n 21).
Designing Competitive Markets for Industrial Data
2017
289
4
subsidies.179 In the rst case, the concession holder
is under an obligation to provide all data collected
in the framework of the concession to the public
authority in a digital format. In the second case, the
recipient of the subsidy is under an obligation to
provide all essential data as stipulated by the grant
agreement in a digital, reusable and exploitable
format to the authority.
171 In all of these instances, the state appears either as
the source, or as an intermediary for making data
available to the public. However, the more difcult
question is whether such access rights can also be
devised with regard to fully independent private
data holders. In this instance, for any access regime,
a fundamental distinction could be made according
to the features of the business model the data holder
applies. In the rst case, the creation of a dataset
is only a by-product, and the commercialisation of
the data in downstream data markets is not part of
the main business of that entity. This is the case,
for example, where a car manufacturer collects
geographic data through the cars’ sensors for the
purpose of predictive maintenance, but other rms
or the state would be interested in getting access
to that data. In such cases, the private entity may
anyhow be willing to grant access in order to
generate additional income, but the parties may
still be unable to agree on access due to information
problems. Intervention in the form of access regimes
that provide for a framework of negotiations,
mediation and arbitration will not reduce in any way
the data holder’s incentives to generate the data.
172
The situation is however very different in the second
case, where the collection of the data constitutes a
key element of the business model in competition
with other rms. Examples are in particular
the business models of search engines or social
platforms, such as Facebook, which build on the
control of user data to compete more effectively in
the market for online advertising. Access regimes
should not facilitate access of weaker competitors
to data where control over such data constitutes the
most valuable asset for competition.
173
The same argument applies to the tools for
collecting and processing information, in particular
as regards big data analytics, since these tools are
of crucial importance for the commercial success
of big data analysts. However, where such tools
become the standard for collecting and processing
information, as explained above,180 access regimes
may be justiable also from the perspective of sound
competition policy.
179 Arts 10 and 11 Loi pour une République numérique (supra n
166).
180 At F.II.3 above.
3. The person seeking access and
the intended use of the data
174
In particular, access to data is justiable where
public entities seek access for the fullment of
tasks in the public interest. In the light of the large
benets deriving from big data analytics, which
could help optimise public policies and decisions
of the state in many regards, this sub-category
for which access regimes could be implemented
seems most important.181 Such regimes could be
implemented at the different levels of government
through sector-specic regulation. Sector-specic
regulation appears as the road to take, since the
security interests of the state will most likely need
different rules than the prevention of infectious
diseases, the protection of the environment or the
functioning of smart cities or trafc control systems.
175
As explained above,182 this is a eld in which the
competition rules on refusal to deal will hardly be
able to promote access.
176
Going yet a step further, access based on public
interest does not have to be limited to public entities
as petitioners of access. An example of an access
regime in the public interest providing for access
to data in favour of even competitors is provided by
the REACH Regulation.183
177
This Regulation has the objective of ensuring ‘a
high level of protection of human health and the
environment, including the promotion of alternative
methods for assessment of hazards of substances,
as well as the free circulation of substances on the
internal market (...)’.184 To enable the assessment
of these hazards, the Regulation’s registration
provisions require manufacturers and importers to
generate data on the substances they manufacture or
import. To meet these obligations the manufacturers
and importers have to submit a dossier that contains
the relevant information to the European Chemicals
Agency (ECHA). Registered substances are allowed to
circulate within the internal market.185
181 See in this context in particular the study of OECD (supra n
5).
182 At F.II.2 above.
183 Regulation (EC) No 1907/2006 of the European Parliament
and of the Council of 18 December 2006 concerning the
Registration, Evaluation, Authorisation and Restriction
of Chemicals (REACH), establishing a European Chemicals
Agency, amending Directive 1999/45/EC and repealing
Council Regulation (EEC) No 793/93 and Commission
Regulation (EC) No 1488/94 as well as Council Directive
76/769/EEC and Commission Directives 91/155/EEC,
93/67/EEC, 93/105/EC and 2000/21/EC, [2007] OJ L 304/1;
consolidated version available at:
eu/legal-content/EN/TXT/PDF/?uri=CELEX:02006R1907-
20150601&from=EN> (accessed 10 September 2016).
184 Art 1(1) REACH Regulation.
185 Recital 19 REACH Regulation.
2017
Josef Drexl
290
4
178
Such assessment of hazards may also require
the manufacturers or importers to conduct new
tests.186 Tests may include animal testing.187 But
the REACH Regulation tries to avoid testing with
vertebrate animals by recourse to alternative
test methods wherever possible.188 As part of the
regulatory framework for preparing and submitting
a registration, Articles 27 and 30 REACH Regulation
implement a scheme for information sharing that
pursues the particular objective of avoiding animal
testing.189 More concretely, the potential registrant
is under an obligation to request a sharing of
information from previous registrants as holders
of studies, whether these studies include tests with
vertebrate animals or not. Thereby, the Regulation
also takes into account the interest of the previous
registrant in fair compensation for the testing it
has already undertaken.190 For that latter purpose,
the owner of the existing study has to determine
the costs of sharing the information in a ‘fair,
transparent and non-discriminatory way’.191 Under
this scheme, the parties are expected to enter into
an information-sharing agreement.192 In case such an
agreement cannot be reached, the REACH Regulation
provides for default rules. The potential registrant
can inform the ECHA about the failure to reach an
agreement.193 Then, within one month, the ECHA
gives the potential registrant permission to refer to
the information requested in its dossier, provided
that it has paid the previous registrant a share of
the cost incurred. At the same time, the Regulation
conrms the right of the previous registrant to claim
a proportionate share of the cost. This amounts to
an equal share of the cost if the previous registrant
makes the full study report available to the potential
registrant. This right of equal cost sharing is
enforceable before the national courts.194
186 Recital 26 REACH Regulation.
187 Such testing has to be conducted in conformity with
Council Directive 86/609/EEC of 24 November 1986 on the
approximation of laws, regulations and administrative
provisions of the Member States regarding the protection
of animals used for experimental and other scientic
purposes, [1986] OJ L 358/1.
188 Recital 47 REACH Regulation.
189 See also Recital 49 REACH Regulation.
190 Recital 50 and 51 REACH Regulation.
191 Arts 27(3) and 30(1)(2) REACH Regulation.
192 More concrete rules on the standards of negotiations are
contained in the Commission Implementing Regulation (EU)
2016/9 of 5 January 2016 on joint submission of data and
data-sharing in accordance with Regulation (EC) 1907/2006
of the European Parliament and of the Council concerning
the Registration, Evaluation, Authorisation and Restriction
of Chemicals (REACH), [2016] OJ L 3/41.
193 Art 27(5) REACH Regulation.
194 Arts 27(6) and 30(3) REACH Regulation.
179 In sum, the REACH Regulation builds on particular
features that could be used as guidance for similar
legislation in other elds. First, a duty to share
information is formulated against the backdrop of a
particular public interest in avoiding the duplication
of the generation of information. In this context,
it is important to remember that, in contrast, the
rules on refusal to deal under EU competition law
following the CJEU’s Bronner judgment do not exempt
the petitioner from making the same investment
as the holder of the essential facility.195 Hence, the
REACH Regulation facilitates access to information
beyond the remedies available under competition
law. Second, the subject-matter of access consists in
identiable information similar to the competition
law cases in Magill or Microsoft. However, it is to be
discussed whether this model could also be applied
to cases where somebody seeks access to large
datasets for the purpose of undertaking big data
analyses or engaging in data mining. It seems that,
to the extent that there is a particular public interest
in obtaining access, such broader access regimes
are also justiable. Third, the REACH Regulation
relies on a framework of contractual negotiations.
It thereby favours a pro-market solution over direct
government intervention. The detailed rules of the
REACH Regulation are very context-specic; but the
negotiation framework could be adapted to other
sector-specic circumstances. Fourth, the data-
sharing agreement also requires agreement on the
price or compensation to be paid for the sharing of
information. The REACH Regulation thereby relies
on concepts that resemble the FRAND concept as
used in particular by standard-setting organisations
in their IP policies concerning SEPs.196 However,
the REACH Regulation is more concrete about the
base for calculating the compensation, relying on
the cost for undertaking the relevant study.197 Fifth,
a negotiation-based access regime will only work
where the law offers a default rule that enables the
public interest to prevail and that provides sufcient
legal certainty for the parties when they assess
whether it makes sense to depart from that rule. This
default rule also has to include procedures of judicial
enforcement through state courts or arbitration
tribunals in case no agreement can be reached.198
195 See at F.II.2. above.
196 FRAND licensing is considered as a general solution to
overcome barriers to entry by Rubinfeld and Gal (supra n
28) at 37.
197 In contrast, R&D costs are not an appropriate standard for
calculating the value of a patent. There is agreement to the
extent that the royalty base should relate to price of the
product in which the technology is implemented. However,
there is disagreement as to whether the royalty should be
calculated as a percentage of the often very complex end
product, or as a percentage of the smallest salable unit.
198 Note that the default rule is very weak in the case of SEPs for
which the patent holder has committed to FRAND licensing.
The problem here is that the default rule is not based on
statutory rules but private ordering through the IP polices
Designing Competitive Markets for Industrial Data
2017
291
4
180
The question may still be whether and to what
extent an access regime like the one contained in
the REACH Regulation could also be implemented
for cases in which there is no additional public
interest. Indeed, such an access regime would make
sense if it is devised as a non-mandatory procedural
framework for negotiations on access to information.
For designing such a general framework, it would
be wise to assess the effectiveness of models
such as the REACH Regulation or the most recent
experience with the negotiation framework devised
by the CJEU in Huawei for the case of SEPs. Such
schemes could especially be applicable for cases in
which the holder of information publicly commits
to grant access to data on FRAND terms. It is yet
another question whether such a scheme should be
implemented by the EU or national legislatures, or
by way of private ordering, in particular through
industry associations that provide for commercial
arbitration. The European Union could cooperate
with the latter institutions to promote such non-
mandatory arbitration on access disputes.
G. Conclusion
181 This article shows that existing EU rules, as regards
both protection of data holders and access to data
based on EU competition law, are applicable in
principle to the data economy. However, in particular
the rules of the Database Directive, the brand-new
Trade Secrets Directive, and EU competition law,
present considerable uncertainties as regards their
application to the data economy. These uncertainties
cannot be expected to be claried quickly by the
European Courts.
182
Yet, although the Trade Secrets Directive was not
drafted to meet the needs of the data economy, trade
secrets protection can provide a sound approach to
protecting rms in the data economy to some extent.
Rather than recognising exclusive control over any
use of protected information, as would be typical for
intellectual property regimes, EU trade secrets law
implements a tort law approach that bans specic
conduct related to the acquisition, dissemination and
use of trade secrets that can be considered as unfair.
It is thereby better suited to balance the interest in
protection and in free ow of information than the
property approach.
of standard-setting organisations. To bring more precision
to the concept of what FRAND actually means may raise
competition concerns in the sense of an anti-competitive
price agreement. Hence, the default rule is ultimately in
need of judicial interpretation of the FRAND concept by
courts. Hence, FRAND licensing of SEPs does not provide a
perfect model for regimes to enhance access to data.
183
While a clarication of the scope of trade secrets
protection regarding data as it is collected and
used in the data economy would certainly be
welcomed, the analysis shows that there is no case
for creating a new system of data ownership. Apart
from the fact that the key issues to be addressed—
namely, regarding the subject-matter of protection,
the identity of the data owner, and the scope of
protection—are of enormous complexity, the
analysis does not produce any evidence for a need
or an economic justication for such legislation. In
principle, in the data economy, no incentives are
needed for generating and commercialising data.
Data holders are able to charge a price for making
data available to third parties based on factual
control over data, supported by technical protection
measures.
184
Hence, the question remains as to whether there
is a need for legislation on access. In principle,
the legislature could also promote access through
un-waivable exceptions and limitation as part of
a comprehensive legislation of data ownership.
However, this article favours stand-alone access
regimes. This latter approach better suits the
dynamic development of the data economy, which
most likely will only gradually inform the legislature
about impediments to access while business models
develop. In contrast, immediate adoption of an
integrated ownership system would result in general
recognition of exclusive control, whereas unfounded
trust in adequate operation of a fair-use provision
or postponing legislation on targeted exceptions
and limitations would fail to address the additional
limitations on the free ow of information generated
by new data ownership.
185
In principle, access can also be sought under
EU competition law. However, this law shows
considerable shortcomings as regards the data
economy: rst, the requirement of market
dominance in Article 102 TFEU considerably limits
the scope of application of this rule and requires an
often burdensome assessment. Second, it is quite
uncertain to what extent Article 102 TFEU can be
applied in cases in which, as will be frequently be
the case, the data holder is not competing with
potential customers in downstream data-related
markets. Of course, Article 102 TFEU can also be
relied upon to remedy excessive pricing. However,
competition law enforcers can hardly be expected to
act as price regulators in the data economy, which
is characterised by information problems and huge
uncertainties regarding the value of data. This puts
the state as a frequent end user of data services in
a particularly uncomfortable situation. Where the
state has to rely on access to privately held data
and big data analyses to optimise its decisions for
fullling tasks in the public interest, it does not act as
an undertaking in the sense of competition law and,
2017
Josef Drexl
292
4
hence, the rules on refusals to deal based on theories
of exclusion and leveraging of market dominance
by vertically integrated rms will not apply from
the outset.
186
Yet the state, including the legislature, could promote
access to data in a pro-active and pro-competitive
way. Where different stake-holders contribute to the
generation of data and information and only some
of these contributors are interested in promoting
access, the legislature can decide to particularly vest
these persons with rights to enforce access against
the interests of the other stakeholders. Examples of
this are the secondary publication right of authors
of scientic publications and data portability rights.
The latter can enhance competition where factual
control of other parties creates a lock-in effect. Block
exemption regulations can take care of conicts over
access to data between suppliers and end producers.
The state can promote access as part of its funding
policy and even when granting subsidies. More
importantly, there is a case for implementing sector-
specic access regimes in the public interest. While
it is hard to conceive a general legal framework for
access of the state to data in the public interest,
progressive sector-specic legislation in diverse
elds of law, including environmental law, public
health law, medicinal law or road trafc law, can
develop models for access regimes over time.
187
Public-interest considerations can also play a role
where private parties seek access to information.
European competition law sets a rather high
threshold for a duty of a dominant rm to share an
essential resource by requiring the person seeking
access to make at least the same investment in
duplicating the resource that was made by the
holder of the facility. There is a case for access
regimes below this threshold where additional
public interests, such as in the case of producing
data through animal testing or clinical trials with
human beings, or the interest in promoting scientic
research, argues against duplication of already
available data.
188
A main barrier of access is uncertainty about the
information contained in large datasets, the new
information that can be drawn from existing data
through data mining and big data analytics and,
hence, the value of data and the appropriate price to
be paid for access. The so-called information paradox
makes it particularly difcult to agree on the price
of access to information in contractual negotiations.
Access regimes should address this issue by favouring
a consensus-based approach to regulating prices.
Where pubic interest or competition law justies
access, a cost-based approach to assessing the
royalty rates seems most appropriate.
189
As regards access negotiations between private
parties, the Commission could support schemes of
private ordering that enable private initiatives to
pool data of multiple data holders.199 The Commission
could also cooperate with institutions that have
experience with arbitration to build up schemes for
mediating negotiations on data licensing.
190
The functioning of the data economy will also
depend on the interoperability of digital formats
and the tools of data collecting and processing.
200
The relevant tools have to rely on interoperability
and, hence, the markets for such tools will
typically be characterised by network effects. In
this regard, the Commission can cooperate and
support industry initiatives for standardisation
of these tools, whereby those initiatives should
also develop disciplines that promote access to
the standardised tools. Accordingly, these needs
of the data economy should also be taken into
account as part of the Commission’s competition
policy regarding standardisation agreements. The
Guidelines on Horizontal Co-operation Agreements
already recognise the principle that standard-
setting organisations should require participants
to commit to license their IP rights in the standard
on FRAND terms in order to make the standard
broadly accessible.201 This approach is superior to
de facto standardisation, not only because it will
enhance quick and general data sharing based on
interoperability of data across borders and across
sectors,
202
but also in the light of the fact that EU
competition law has so far not developed appropriate
disciplines through its case-law on refusals to license
regarding the access problems arising from de facto
standards.
199 This also has a competition law connotation, as demonstrated
by the rules on information sharing in the Communication
from the Commission—Guidelines on the application of
Article 101 of the Treaty on the Functioning of the European
Union to horizontal co-operation agreements, [2011] OJ C
11/1, paras 55-110.
200 See the standardisation issues regarding data and big
data analysis mentioned in Communication from the
Commission—ICT Standardisation Priorities for the Digital
Single Market (19 April 2016), COM(2016) 176 nal, p. 9.
201 Horizontal Cooperation Guidelines (supra n 199) para 285.
202 Commission Communication on ICT Standardisation
Priorities (supra n 200) at 9.

To continue reading

Request your trial

VLEX uses login cookies to provide you with a better browsing experience. If you click on 'Accept' or continue browsing this site we consider that you accept our cookie policy. ACCEPT