Propiedad Intelectual 2019

World Intellectual Property
Report 2019
The Geography
of Innovation:
Local Hotspots,
Global Networks
World Intellectual Property
Report 2019
The Geography
of Innovation:
Local Hotspots,
Global Networks
Except where otherwise indicated, this work is licensed under
the Creative Commons Attribution 3.0 IGO License.
The user is allowed to reproduce, distribute, adapt, translate

and publicly perform this publication, including for commercial
purposes, without explicit permission, provided that the content
is accompanied by an acknowledgement that WIPO is the
source and that it is clearly indicated if changes were made to
the original content.
Suggested citation: WIPO (2019). World Intellectual Property

Report 2019: The geography of innovation: Local hotspots, global
networks. Geneva: World Intellectual Property Organization.
Adaptation/translation/derivatives should not carry any official

emblem or logo, unless they have been approved and validated
by WIPO. Please contact us via the WIPO website to obtain
permission.
For any derivative work, please include the following disclaimer:

“The Secretariat of WIPO assumes no liability or responsi
bility with regard to the transformation or translation of the
original content.”
When content published by WIPO, such as images, graphics,

trademarks or logos, is attributed to a third party, the user of
such content is solely responsible for clearing the rights with
the right holder(s).
To view a copy of this license, please visit:

https://creativecommons.org/licenses/by/3.0/igo/
The designations employed and the presentation of material

throughout this publication do not imply the expression of any © WIPO, 2019
opinion whatsoever on the part of WIPO concerning the legal
status of any country, territory or area or of its authorities, or World Intellectual Property Organization
concerning the delimitation of its frontiers or boundaries. 34, chemin des Colombettes, P.O. Box 18
CH-1211 Geneva 20, Switzerland
This publication is not intended to reflect the views of the
Member States or the WIPO Secretariat. ISBN: 978-92-805-3095-7
The mention of specific companies or products of manufacturers Attribution 3.0 IGO

does not imply that they are endorsed or recommended (CC BY 3.0 IGO)
by WIPO in preference to others of a similar nature that are
not mentioned. Printed in Switzerland
Table of contents
Foreword 4 Chapter 1
The changing global geography of innovation 15
Acknowledgements 7 1.1 Concentration of innovation in
urban hotspots 16
Executive summary 8 1.2 Networks and the global spread
of innovation 21
Technical notes 123 1.3 Conclusions 25
Acronyms 128 Chapter 2

Global networks of innovation hotspots 31
2.1 The two sides of global
knowledge production 33
2.2 Global networks of collaboration
and sourcing 41
2.3 Local innovation and global networks
of innovative hubs 51
2.4 Conclusions 57
Chapter 3
Auto and tech companies – the drive for
autonomous vehicles 61
3.1 Definitions 62
3.2 Technological evolution of the
automotive industry 62
3.3 Technological shift 66
3.4 Competition and cooperation in AV 66
3.5 Role of geography in AV technology 69
3.6 AV innovation, countries and cities 69
3.7 Is AV technology changing the
geography of innovation in the
automotive industry? 76
3.8 Potential positive and
negative impacts of AVs 76
Chapter 4
Plant biotechnology – connecting urban
innovation and rural application 87
4.1 The rising importance of
plant biotechnology 88
4.2 The innovation landscape of
4.3 The innovation network of
4.4 Future of plant biotechnology 105
Chapter 5
Policy perspectives: the case for openness 113
5.1 The economics of openness 113
5.2 Openness in an age of falling
R&D productivity 117
Foreword
Geography has always played a

central role in the organization of
economic activity. Cities first formed
as trading hubs for agricultural
and manufactured goods. Many of
them emerged where trading routes
intersected or where goods moved
from one mode of transportation to
another. With the onset of the industrial
revolution, cities became centers of
large-scale industrial production. As
industrialization advanced, some
expanded into megacities, while
others saw their fortunes decline.
4
Foreword
In the innovation-driven economy of the 21st century, in some detail how innovators from around the world
cities continue to play a pivotal role. However, the collaborate with each other. The emerging picture is a
forces shaping the geography of economic activity have complex one, with a limited number of global innova
changed. Companies want to be in urban hotspots, tion hotspots in a few countries accounting for most
because that is where the most skilled and talented innovative activity. Collaboration is widespread, taking
workers live. Well-paying and rewarding jobs as well place in increasingly larger teams and – for most but not
as the buzz of city life, in turn, attract yet more highly all countries – it is increasingly cross-border in nature.
skilled individuals to those hotspots. Innovation also
relies crucially on the exchange of ideas among people. In addition to this economy-wide perspective, the
Such exchanges usually happen best when people live report includes two case studies that explore in detail
and work in close proximity to each other. the evolving geography of innovation for two fields of
technology undergoing rapid change. One case study
Yet, the economic geography of the 21st century has focuses on autonomous vehicles technology. It details
another important dimension. Technology has facili how innovation is reshaping the car industry, with
tated new ways of collaborating and sharing knowl information technology (IT) companies challenging
edge, connecting skilled individuals located far apart. established carmakers. This transformation is broaden
The emerging global innovation landscape is thus one ing the innovation landscape, with several IT-focused
of geographically concentrated centers of excellence hotspots – which traditionally were not at the center of
around the world, which are embedded in a global automotive innovation – gaining prominence.
network carrying knowledge in many directions.
The other case study focuses on agricultural biotech
The evolving geography of innovation matters. nology. Scientific and inventive activity in crop bio-
Governments around the world strive to promote a technology is concentrated in a few high-income
policy environment conducive to innovation. Doing economies and China, and – within those economies
so requires an understanding of the local dynamics – mostly in large metropolitan areas. Relative to other
of innovation ecosystems. For example, where can areas of innovation, however, it is more geographically
government-funded research best enhance nascent widespread, spanning many countries in Africa, Latin
technological capabilities? How can smart city plan America and Asia. This partly reflects a need to adapt
ning encourage opportunities for knowledge sharing innovations to local conditions.
and collaboration? More broadly, the spread of innova
tive activity within economies increasingly affects the The evidence presented in this report highlights how
regional distribution of incomes. Understanding the globally intertwined innovation has become. Crucially,
driving forces behind this trend, in turn, enables better the ability of companies and researchers to collab
policy responses. orate across borders has relied on policies largely
favoring openness and international cooperation. The
Our World Intellectual Property Report 2019 offers report makes the case for maintaining policy openness
an empirical perspective on the global geography of and further strengthening international cooperation.
innovation. It does so by following the geographical Solving increasingly complex technological problems
footprint that innovators have left in millions of patent will require ever larger and more specialized teams of
and scientific publication records over the past few researchers. International collaboration helps form such
decades. WIPO’s Global Innovation Index has already teams and will therefore be indispensable in continu
embraced this big data approach in identifying the ously pushing the global technology frontier.
world’s largest science and technology clusters. This
report goes further. It employs more data going back While offering original insights, the analysis presented
several decades, analyzes time trends and explores in this report also comes with certain caveats. Patent
5
World Intellectual Property Report 2019
The evidence
presented in this
report highlights
how globally and scientific publication data provide rich and interna
tionally comparable information on innovative activity.
intertwined However, they neither capture all such activity, nor

do they fully portray the rich interactions taking place
innovation among innovators. In addition, the dynamic forces

that shape the direction of global innovation networks
has become. are manifold and interact with one another in intricate
ways. Further research that offers empirical guidance
on these forces would be of much value.
We hope that this report contributes to a greater appre

ciation of the importance that geography exerts on
innovative activity and, in doing so, that it helps in
refining policies to promote innovation and ensuring
that its benefits are widely shared.
Francis GURRY
Director General
6
Acknowledgements
This report was developed under the general direc The report team benefited greatly from external reviews
tion of Francis Gurry (Director General) and super and comments of the draft chapters and background
vised by Carsten Fink (Chief Economist). The report papers by Cristina Chaminade (Lund University),
was prepared by a team led by Julio Raffo (Head of Frédérique Sachwald (Science and Technology
Innovation Economics), Intan Hamdan-Livramento Observatory, Hcéres), Maryann P. Feldman (University
(Economist), Maryam Zehtabchi (Economist) and Deyun of North Carolina), Kazuyuki Motohashi (University of
Yin (Fellow), all from WIPO’s Economics and Statistics Tokyo), Luciana Marques Vieira (School of Business
Division (ESD). Administration of Sao Paulo, FGV EAESP), José Maria
da Silveira (Universidade Estadual de Campinas,
The report draws on various background papers UNICAMP) and Can Huang (Zhejiang University).
commissioned for this report. In particular, Chapter 1
is based on a literature review prepared by Riccardo Additional input, comments and data were kindly
Crescenzi (London School of Economics, LSE), Simona provided by Lesya Baudoin, Daniel Benoliel, Shakeel
Iammarino (LSE), Carolin Ioramashvili (LSE), Andrés Bhatti, Maurice Blount, Lee Branstetter, Richard
Rodríguez-Pose (LSE) and Michael Storper (LSE and Corken, Alica Daly, Gaétan de Rassenfosse, Philipp
University of California, Los Angeles). Großkurth, Christopher Harrison, Irene Kitsara, Agénor
Lahatte, Orion Penner, Leontino Rezende Taveira, David
Ernest Miguelez (Groupe de Recherche en Économie Sapinho, Florian Seliger and Usui Yoshiaki.
Théorique et Appliquée, GREThA, Bordeaux),
Francesco Lissoni (GREThA Bordeaux and Bocconi Hao Zhou and Kyle Bergquist assisted in putting togeth
University), Christian Chacua (GREThA Bordeaux), er data used in this report.
Massimiliano Coda-Zabetta (GREThA Bordeaux) and
Gianluca Tarasconi contributed the background report Samiah Do Carmo Figueiredo, Caterina Valles Galmès
and assisted in the data preparation for Chapter 2. and Cécile Roure provided valuable administra
tive support.
Chapter 3 relies on background research conducted
by Kristin Dziczek (Center for Automotive Research, Finally, gratitude is due to editorial and design
CAR group), Eric Dennis (CAR group), Qiang Hong colleagues in the Publications Division for leading the
(CAR group), Diana Douglas (CAR group), Yen Chen production of the report and to Richard Waddington
(CAR group), Valerie Sathe-Brugeman (CAR group) for his editing work. The WIPO Library provided helpful
and Edwin Marples (CAR group). research support throughout the report’s development
and the Printing Plant provided high-quality printing
Finally, Gregory D. Graff (Colorado State University) services. All worked hard to meet tight deadlines.
contributed to the background report for Chapter 4.
7
Executive summary
The geography of innovation can

seem paradoxical: the generation of
scientific knowledge and innovation is
both increasingly global and intensely
concentrated in a few local hotspots.
New players, particularly Asian

countries, are responsible for more
and more scientific research and
inventions, which were once the almost
exclusive domain of a handful of rich
economies. At the same time, this
greater international spread has been
accompanied – at country level – by
increased concentration of innovative
activity within a few densely populated
areas. These urban areas are vibrant
innovation ecosystems, such as
Silicon Valley outside San Francisco
in the U.S or China’s Shenzhen–Hong
Kong, a more recent hotspot.
8
Executive summary
This paradox is more apparent than real, however; The increasingly dispersed and interconnected flow of
the world’s most innovative urban agglomerations are knowledge and innovation has mirrored the develop
also the most open to the outside world. Sometimes, ment of complex global networks, or value chains, for
they are better connected internationally than they the production and delivery of goods and services.
are to their national hinterlands. Together, they form In particular, multinational companies have located
what economists have come to refer to as global knowledge-intensive production stages – most impor
innovation networks. Skilled individuals and innovative tantly research and development (R&D) – in urban
companies are at the center of these networks. Highly agglomerations that offer specialized knowledge
skilled workers gravitate toward innovative urban areas and skills. More generally, it is the need for greater
because they want to interact with one another and collaboration in the face of growing technologi
enjoy the amenities of metropolitan life. Large cities cal complexity that has driven both the increasing
offer companies a large local market, specialized concentration of innovation in certain urban areas
suppliers and academic institutions that allow them and its global spread.
to reap economies of scale and scope. Knowledge, in
turn, flows more fluidly among company and university
researchers when they work in close proximity, fueling Innovation is increasingly local
the innovation engine.
On the basis of geocoded inventor and scientific author
This WIPO report analyses these dual trends, exploiting data, this report explores the geography of innova
a rich data set of millions of patent applications and tion within countries and identifies the world’s main
scientific publications. Its findings argue for increasing agglomerations of scientific and technological activity.
openness and support for collaboration if innovation is It looks at two types: global innovation hotspots, which
to continue to flourish. show the highest density of scientific publication or
patent activity; and specialized niche clusters, where
the density of inventors and scientific authors is high
Knowledge creation is spreading in a given field but not high enough generally to be a
to more and more countries global hotspot.
For most of the period from 1970 to 2000 only three

countries – the United States of America (U.S.), Japan Innovation is geographically concentrated
and Germany – accounted for two thirds of all patent in a limited number of areas
ing activity worldwide. When the remaining Western
European economies are included the share reached The emerging landscape of global hotspots and niche
some 90 percent. But in the years since, the rest of clusters shows that inventive and scientific activity
the world has come from almost nowhere to account within each country is persistently concentrated in a
for almost one third of all patenting activity. Published few large, cosmopolitan and prosperous urban areas.
scientific data have spread even more widely, with the In the U.S., hotspots around New York, San Francisco
rest of the world going from less than a quarter of all and Boston accumulated roughly a quarter of all U.S.
such publication to around half over the last 20 years. patents filed from 2011 to 2015. In China, those around
Beijing, Shanghai and Shenzhen increased their share
China and the Republic of Korea are largely responsible from 36 percent to 52 percent of all Chinese patents
for the rising share of new areas in knowledge produc during the same period.
tion and innovation; together, they account for over 20
percent of patents registered in the years 2015–2017, Less than 19 percent of all inventive and scientific output
compared to under 3 percent in 1990–1999. Other worldwide is generated by inventors or researchers
countries, notably Australia, Canada, India and Israel, located outside hotspots and niche clusters. Despite
have also contributed to the global spread of innova this big change in the global innovation picture, more
tion. Many middle-income countries, however, and all than 160 countries – the vast majority – still generate
lower-income countries, continue to have substantially little innovation activity and do not host any hotspot
lower levels of patenting activity. or niche cluster.
9
Big cities are not necessarily the U.S. and Western Europe accounted for 68 and
hubs of innovation 62 percent, respectively, of all international inventive
and scientific collaboration. Most of the collabora
Not all large metropolitan areas are innovation dense. For tion happens among inventors and researchers from
example, North America hosts most hotspots in dense these countries. New entrants to these collaboration
urban areas along the east and west coasts, while many networks, from countries such as China, India, Australia
dense inland urban areas do not have an equivalent and Brazil, still mostly collaborate with the aforemen
density of innovation. Asia, Latin America and Africa host tioned economies rather than with each other.
many dense urban areas with no corresponding innova
tion density. Despite high populations, top metropoles –
for example, Bangkok, Cairo, Cape Town, Kuala Lumpur Hotspots and clusters drive international
and Santiago de Chile – only have a modest degree of collaboration and global networks
innovation density in some specialized fields.
Most global innovation hotspots have increased their
And less dense urban areas can sometimes host international collaboration over the last two decades.
niche clusters. Some examples are Ithaca in the U.S., This collaboration – whether national or international,
Stavanger in Norway and Bern in Switzerland, which patents or publication – forms a thick web of ties that
are highly innovative cities due to the strong innovation constitutes the global innovation networks. The shape
footprint of local academic institutions, industries or, of these networks has evolved, typically with more
sometimes, the presence of a key company. nodes and ties being added over time.
Inventors and scientists within hotspots and niche

Collaboration is increasingly the norm clusters collaborate internationally more than those
outside, particularly in scientific articles. Over the last
Data show that teams are involved in an increasing two decades, the share of scientific publication involv
majority of scientific papers and patents. In the early ing international collaboration between scientists from
2000s, teams already produced 64 percent of all inside hotspots has been more than triple that between
scientific papers and 54 percent of all patents. By the scientists outside of them.
second half of the 2010s, these figures had grown to
almost 80 percent and 70 percent, respectively.
Collaboration is concentrated
Most high-income economies also show rising inter
national collaboration. The forces pushing academia Notwithstanding new network nodes and their ties,
and companies to cross borders seeking partners for the hotspots in the U.S., Europe and Asia remain the
innovation are manifold. The scientific community has core of the global networks both in terms of output and
a long tradition of engaging in international collabora connectivity. Overall, larger hotspots collaborate both
tion, while multinational companies seek efficiency nationally and internationally, whereas niche clusters
gains from the international division of their R&D and and smaller hotspots collaborate predominantly at the
through international collaboration. national level. For instance, many French and United
Kingdom (U.K.) hotspots connect with the rest of the
The main exceptions to the trend toward international world mostly through Paris and London respectively.
ization are East Asia’s top economies where Japan, the In China, Shanghai, Beijing and Shenzhen act as the
Republic of Korea and, more recently, China have seen top gatekeepers.
falling shares of international collaboration, though not
in absolute number terms. Yet, not all hotspots have the same relevance in terms
of connections. U.S. hotspots are among the most
connected nodes. Beijing, London, Paris, Seoul,
A few countries account for Shanghai and Tokyo are also highly connected, but
most of the international ties much less so. Interestingly, the high volume of inven
tive and scientific activity of the U.S. hotspots does
Most international collaboration is concentrated among not fully explain their higher connectivity. Many other
a few main countries. In the period from 2011 to 2015, hotspots – for example, Tokyo or Seoul – show larger
10
Executive summary
or similar scientific or inventive output but are not so America and Africa rely intensively on the ingenuity of
highly connected. U.S., Western European and Chinese inventors.
The intensity of international collaboration varies

considerably across countries. For example, hotspots Innovation hubs move
in India and Switzerland are highly connected interna and can disperse over time
tionally, while those in the Republic of Korea and Japan
are weakly connected. In many hotspots, internation Multinationals can have very different needs and strat
alization often goes hand-in-hand with an increase in egies as to where to source for talent, and these can
the share of local-only interactions. In many Chinese change over time. For example, Google and Siemens
hotspots the number of co-inventions within hotspots have concentrated inventive activities within their top
has grown remarkably, leading to a decline in share of hubs. In the 2010s, San Jose–San Francisco accounted
both national and international collaboration outside for 54 percent of Google’s patents, up from 36 percent
these hotspots. in the 2000s. Similarly, Nuremberg – the most important
source of patents for German manufacturing company
Siemens – accounted for 32 percent during the 2010s
Multinational companies lie against 27 percent in the 2000s.
at the center of the web
The concentration is even greater in Asian companies,
Patent data shed light on the corporate R&D networks at though it has slightly decreased over time. Tokyo
the center of global innovation networks. Multinational and Shenzhen–Hong Kong were the most important
companies from around the world increasingly list inventive sources for Sony and Huawei in the 2010s,
foreign inventors in their patent applications and accounting for 71 percent and 81 percent of patents,
those foreign inventors come from a widening set of respectively. However, this is down from a respective
countries. In the 1970s and 1980s, only 9 percent of 83 percent and 88 percent in the previous decade,
patents filed by U.S. companies had foreign inventors; suggesting a relative dispersion of innovation.
by the 2010s, this share had risen to 38 percent. Western
European companies saw a similarly sharp increase,
from 9 percent to 27 percent in the same period. Innovation is reshaping
the car industry
Such international patent sourcing still mostly happens
between companies and inventors from high-income The report delves deeper into the evolving geography of
economies. In the 1970s and 1980s, 86 percent of the innovation by studying two industries seeing profound
international patent sourcing was between multinational change. One is the automotive sector, which is in the
companies and inventors from the U.S., Japan and early phases of technological disruption. New entrants
Western European countries. However, this share fell – from within the automotive industry and from the
to 56 percent in the 2010s. information technology (IT) industry – are challenging
established players.
Middle-income economies are Fully autonomous vehicles (AV) have yet to reach the
new players in MNC networks market. Nevertheless, artificial intelligence (AI) data
analytics and the interconnectivity of devices and
Two main developments explain this fall. On the one components are reformulating the industry’s business
hand, multinationals from these countries increasingly model toward services and the so-called “platform
outsourced R&D activities to middle-income econo economy.” Traditional automakers fear being displaced
mies, particularly China, India and, to a lesser extent, in their core business of making and marketing cars.
Eastern Europe. For instance, in the 2010s, more than a
quarter of all international patent sourcing by U.S. multi Patent data suggest that traditional automakers and
national companies had an inventor from China or India. their suppliers are at the forefront of AV innovation. Ford,
On the other hand, multinationals from middle-income Toyota and Bosch – accounting for 357, 320 and 277
economies also actively participate in global innovation of AV patent families, respectively – are the top three
networks. Companies from Asia, Eastern Europe, Latin AV patent applicants. However, non-automakers also
11
feature in the list of top patent applicants. Google, and The landscape of plant biotech innovation
its AV subsidiary Waymo, are in eighth position with 156
patents, ahead of traditional automakers like Nissan, A handful of countries accounts for the bulk of biotech
BMW and Hyundai. Uber and Delphi each have 62 AV nology inventive and scientific output. The U.S.,
patents and are ranked joint 31st. Germany, China, Japan and the Republic of Korea
accumulate more than 55 percent and 80 percent of all
crop biotechnology articles and patents, respectively.
Incumbents and new entrants collaborate Only Argentina, Australia, India, Israel, Mexico and
among and between themselves Singapore join them in the list of countries hosting plant
biotechnology clusters; and, except for Australia, they
Neither the incumbents nor the new entrants currently all have only one.
have all the required competencies for producing AVs.
They either need to join forces or else develop inter There is a geographic divide between where plant
nally the respective skills they lack. AV innovation is biotech innovation occurs and where transgenic crops
a long-term and costly endeavor. Stakeholders have are farmed. In most cases, crop biotechnology hotspots
high incentives to collaborate and share risks and costs are located in large metropolitan areas, either in global
with different type of partners. Three types of collabo innovation hotspots or in specialized niche clusters
ration are forming: between incumbent automakers; with strong biotech competences. This also holds for
between tech firms, and between automakers and developing countries, where national crop biotechnology
tech firms. The emerging collaboration network is an clusters are typically located in large urban areas, such
amalgam of all the above: none is mutually exclusive, as São Paulo in Brazil and Cape Town.
and they coexist.
Some clusters are close to rural areas – for example,
Viçosa in Brazil or Irapuato in Mexico. Wherever they
Automotive and IT firms stay tied are, their presence is usually associated with influential
to their traditional clusters public institutions, such as universities, international
agricultural research centers and/or national agriculture
The top automakers and top IT giants still strongly favor research systems.
home-based sites for their inventive activities. There is
some shift in geography at the margins, so it might be
too early to give a definitive answer as to whether AV Increasing private–public collaboration
technology will change the geography of innovation in
the automotive industry. Private firms, particularly the four major agrobusi
ness companies – Bayer and BASF from Germany,
ChemChina and Corteva Agriscience from the U.S. –
Innovation is sown in biotech labs undertake a large part of the R&D investment in plant
and harvested in agricultural clusters biotechnology. The need for access to proprietary
technologies has spurred collaboration within the
Crop biotechnology is an industry where innovation has industry through cross-licensing, licensing-in, joint
to be adapted to local agro-ecological conditions. While research ventures and even mergers and acquisitions.
most plant biotechnology inventions may come from
high-income countries – for example, the U.S., Western Nevertheless, there is an increasing need for collabora
Europe and East Asian countries – they still need adapt tion with the public sector to access, for example, pools
ing to different climate and soil conditions. Most of the of germplasms and cultivars – crop varieties with desir
transgenic crops used in emerging middle-income able traits – often held by public research institutions.
countries during the late 1990s were locally adapted For public institutions the high costs of commercializing
germplasms of their North American counterparts. As crop biotechnology products almost always require
a result, plant biotechnology innovation clusters exist collaboration with large multinational companies. Since
in many parts of the world. However, the data show the 2000s, co-patenting between private firms and
that crop biotechnology innovation in many countries public institutions has overtaken co-patenting among
in Africa, Latin America and the Caribbean, and Asia private firms to become the main type of collabora
is geographically concentrated. tion. In fact, since the 2010s, co-patenting between
12
Executive summary
private firms has slipped to third in importance behind Making openness work requires
co-patenting between public institutions. international cooperation…
International cooperation in relation to innovation has

Openness in pursuit of innovation many dimensions. It is needed to promote incentives
brings mutual gains for investments in innovation that reflect the demands
and size of the global economy. It can also play an
What does the global geography of innovation, as important role in making it easier for innovators to do
portrayed in this report, imply for policymaking? The business internationally. Lastly, governments can pool
growth of global innovation networks has relied on poli resources and fund large-scale scientific projects that
cies favoring openness and international cooperation, exceed national budgets or require technical knowledge
but this should not be taken for granted – especially available in different countries.
as public perceptions have become more skeptical of
the benefits of globalization.
…and policies that address
Economic theory offers good reasons why the free growing regional divergence
exchange of knowledge is beneficial: it promotes the
specialization of different innovation clusters around One worrying trend of the past few decades is the
the world, leading to more efficient and diverse knowl increasing inter-regional polarization of incomes, inno
edge production. The public good nature of knowledge vative activity and high-skilled employment and wages
reinforces the benefits of openness: if knowledge flows within countries. Openness strengthens the gravita
generate economic benefits abroad without diminish tional pull toward champion regions. As this report
ing those at home, there are bound to be mutual gains demonstrates, the most vibrant innovation hotspots,
from openness. which are embedded in global innovation networks,
tend to be located in what are already the richest
Theoretically, there may well be circumstances in which metropolitan agglomerations within countries.
strategic restrictions on trade and knowledge flows
could benefit the growth paths of economies. However, Regional support and development policies can play an
the experience of high-income economies over the past important role in helping regions that have fallen behind.
decades suggests an overall positive impact from the While not reversing the gravitational pull of successful
flow of knowledge about new technologies. regions, they can promote innovation-driven growth
that benefits economies as a whole.
Falling R&D productivity reinforces

the case for openness
Continuously pushing the technological frontier is

becoming exceedingly difficult. Evidence suggests that
achieving the same level of technological progress as
in the past requires more and more R&D effort.
Falling R&D productivity calls for constantly increasing

investments in innovation. It also calls for collabora
tion and openness. Finding solutions to increasingly
complex technological problems requires larger teams
of researchers and greater specialization in research,
both of which can be promoted by openness and
international collaboration.
13
Innovation is increasingly
concentrated in urban
“hotspots.” At the same
time, these hotspots are
connecting and
collaborating across
the world.
Chapter 1
The changing global

geography of innovation
Technological innovation is the engine that The first industrial revolution – spurred in the
propels economic growth and fosters higher late 18 th century by new manufacturing process
living standards. As described in WIPO’s World es powered by steam – concentrated the world’s
Intellectual Property Report 2015, the growth industrial output in Western Europe, particularly,
record of the past 200 years has been historically the United Kingdom (U.K.). 2 It changed the face of
unprecedented. A series of technological the world economy, generating a different global
breakthroughs have greatly improved the quality development hierarchy. Equally important, it also
of life and generated widespread material led to persistent regional divergences within Europe,
prosperity. Even so, some national economies with a select set of regions and cities – such as
have seen faster and more sustained growth Manchester and London in the U.K., Normandy, Paris
than others. The geographical distribution and and Lyon in France, the Ruhrgebiet in Germany, Liège
the spread of innovation activities – whether in Belgium or the Franco-German region that included
technological or knowledge-generating – go a Lorraine, Saarland and Luxembourg – constituting
long way to explaining why some economies the “European core.”3
have developed faster than others. New
technologies, in turn, have shaped where and The second industrial revolution – driven by a broad
how innovation has taken place.1 array of electro-mechanical inventions in the second
half of the 19th century – witnessed the entry of North
America into the high-income club of the world, while
broadening the industrialized regions of Europe. There
was not a perfect overlap between the countries, cities
and regions at the core of the two revolutions. Some
previously core regions declined while others thrived.
In Europe, the waves of industrialization expanded
concentrically, taking in, among others, southwest
France, northeastern Spain, the Milan–Venice corridor
in northern Italy, Berlin, Vienna, Krakow and Prague, as
well as moving north to Oslo and Sweden’s Gothenburg.
In the United States of America (U.S.), northeastern
cities – such as Boston, New York and Baltimore –
remained industrially important, but industrialization
expanded to several Midwest cities, such as Chicago,
Detroit, Minneapolis and Cleveland.
Starting in the 1970s and 1980s, a third industrial revo

lution has occurred, broadly involving digital technolo
gies, life-science and biological technologies, financial
engineering and significant breakthroughs in transport
and logistics. It has coincided with major increases
in global trade and investment flows. Innovation and
economic development has spread to northeast Asia,
moving from Japan to the Republic of Korea and later
to China. Tokyo, Seoul, Shenzhen and Beijing have
grown into megacities shaping the direction of techno
logical progress today. The “incumbent” high-income
15
economies in Europe and North America continue to entrepreneurs – which are linking the main innovation-
be at the forefront of innovation, but again with a new creation centers around the globe. The final section
geographical landscape within them. explores some of the consequences arising from
today’s global network of highly concentrated innova
What forces can explain why innovation has concen tion hotspots.
trated in certain geographical areas and has spread
only unevenly? Going beyond the broad patterns
outlined above, how exactly is the global geography 1.1 Concentration of
of innovation changing? How do companies in today’s innovation in urban hotspots
globalized age organize their innovation activities
across the world? Framing the geography of global innovation requires
understanding both the forces driving innovation’s
This report endeavors to provide a perspective on concentration and its spread. One prominent feature
these questions. It does so in three parts. First, it of the geography of economic development is common
reviews economic thinking and empirical evidence to both established high-income economies and
explaining the geographical distribution of innovative successful emerging middle-income ones: the geog
activity – a task performed in this opening chapter. raphy of high incomes is increasingly metropolitan,
Second, it draws on patent and scientific publication reflecting renewed inter-regional divergence within
data with geocoded inventors and scientific authors countries. These metropolitan areas are also hotbeds
from around the world to show how this geography for the formation of ecosystems of innovation. In
of innovation has evolved over the past decades. the U.S., two prominent examples are the south
The discussion of the emerging trends – presented ern part of the San Francisco Bay Area in Northern
in Chapter 2 – will portray this geography in terms California and the Greater Boston metropolitan region
of global innovation networks (GINs), geographically in Massachusetts, often dubbed Silicon Valley and
concentrated innovation hotspots and niche innova Route 128, respectively.
tion clusters connected to one another, which are
increasingly leading the way. The report will also
illustrate the operation of such GINs through two case What economic forces explain the
studies – one on autonomous vehicles and the other agglomeration of innovation?
on agricultural biotechnologies. These case studies
will be presented in Chapters 3 and 4, respectively. One of the toughest questions for geography, econom
Finally, the report concludes – in Chapter 5 – with ics and development studies is to ascertain why such
policy perspectives on its main findings. They empha hotbeds, or agglomerations, of innovation arise and
size, in particular, the benefits of national innovation flourish where they do. This question goes from the
systems remaining open to the international exchange general factors that lie behind the clustering of inno
of knowledge. vation to the specific geographies of those agglom
erations.4
This opening chapter discusses the main economic
forces behind both the geographical concentration Several different economic theories address this ques
and spread of knowledge creation and diffusion. The tion. These most commonly consider the economic
following section reviews the main economic theo forces relating to pools of skilled labor, market scale
ries and existing evidence behind the geographical and knowledge spillover – where highly innovative
concentration of innovative activities. It shows that the firms are concentrated, knowledge can filter, or “spill,”
processes of knowledge creation and flow, investment from one to another. Historical accidents or deliberate
and appropriation are reinforcing global innovation and policy can affect all these, but none succeed entirely
economic hierarchies and how they are concentrating in answering the question. At the same time, there
innovation within geographical hotspots or clusters, are forces pushing in the opposite direction, toward
most of them metropolitan. Section 2.2 considers geographical dispersion, but all accounts indicate
how these processes are at the same time leading to these are not as a strong.
an increased dispersion of hotspots at a global scale.
It explores the main forces – operating principally Figure 1.1 provides a graphical summary of the main
through global networks of firms, researchers and forces driving concentration, discussed below.
16
1 The changing global geography of innovation
People, companies and Empirical evidence shows that regions that had a
ideas cluster together higher-than-average concentration of college-educated
workers in the past observe further growth in the share
Figure 1.1 Main economic forces driving
of college-educated workers, per capita incomes,
geographical concentration of innovation
patents and other direct and indirect proxies for inno
vation. The characteristics of the local labor supply
seem to influence the development trajectory of
Labor Market
innovation-generating agglomerations and impact
Skills Pool of organizations regional innovativeness, both in the U.S. and in the
Education Economies of scale
Migration European Union (EU).6
Agglomeration
of economic However, what explains the origin of a singularly skilled
activities workforce in a given region? At some points in history,
Urbanization skilled workers have changed their geographical distri
S&T specialization
Innovation hotspot bution, favoring spatial spread. Why has the behavior of
the skilled workforce spontaneously changed in favor
of geographical concentration? This was the case in
the maturing phase of the second industrial revolution,
Knowledge when there was a massive migration of skilled workers
Information spillovers
Tech capabilities and relatedness – together with unskilled ones – from all over the world
Recombinant nature to the leading developed countries from 1940 to 1980.7
Note: S&T = Science and technology. Historical accidents involving unusual individuals can
be partial explanations for the location of innovative
agglomerations, especially those involving first-mover
Does skill supply help drive regions in the key technologies of each industrial
innovative agglomeration? revolution. Thus, by some accounts, Silicon Valley
is where it is, because William Shockley – the inven
Mainstream economic theory offers a number of ways tor of the silicon-based semiconductor – decided to
in which the geographical concentration of innovation relocate from New Jersey to be near his aging mother.
can be an indirect outcome of labor supply – both Another anecdote concerning Shockley is that after he
quantity and quality.5 attracted top-quality associates to his first-mover firm,
his difficult management style caused them all to quit
These theories assume that workers with different skills on the same day. This event – known as “Shockley’s
gravitate toward different regions. In basic terms, highly massacre” – created a first and unexpected example of
skilled workers cluster together, because they want to the spin-offs that have become so typical in the Silicon
interact with one another, so the education and skills Valley process of development. The annals of innova
of the workforce in a given region can act as a force tion contain plenty of other such people-based stories.
of attraction. At the same time, migration can change
the skills base of the workforce of the receiving region, Nevertheless, the randomness of such “great inven
reinforcing the agglomeration effect. tor presence” stories raises some doubts. There are
so many famous individuals associated with Silicon
The preferences of highly skilled workers for vibrant Valley – from Shockley and Frederick Terman, one of
agglomerations and to work in innovation are part of the acknowledged “fathers” of Silicon Valley, to Apple’s
this picture. Innovation-generating occupations offer Steve Jobs and Google co-founders Sergey Brin and
career trajectories and life-long learning, which secure Larry Page – that it seems unlikely that all could be there
future employment opportunities at a time when auto by coincidence. Moreover, Saxenian (1994) powerfully
mation increasingly seems to threaten many traditional argues that the mere presence of early innovators
white-collar occupations. Such work also offers high is not enough. Plenty of the early great information
wages that compensate for rising costs of living and technologies (IT) innovators were located in Boston,
housing. Cost pressures also push less well-paid, but they did not stay. Facebook’s Mark Zuckerberg
unskilled workers to the urban periphery. left Boston for Silicon Valley, because Boston was not
17
the right place to transform a breakthrough invention productive the higher the productivity of the original
into a fully fledged innovation, just as had happened pool of firms. The agglomeration and innovativeness of
to New Jersey decades earlier, when William Shockley the car industry in Detroit in the first quarter of the last
left for the Bay Area. century was due to a great extent to spin-offs carry
ing on the technological and organizational practices
In addition, “skill” is not a uniform entity and different of their parents.
technologies may need different sets or combinations
of skills that may not always overlap. A finance worker Academic institutions – such as universities – are also
will be attracted to different locations than an IT engi important factors in concentration. The concentration
neer, and these locations and jobs are not interchange of university graduates and science, engineering and
able. But at the same time, different professions and technology workers mirrors the spatial concentra
skills may be complementary for the production of a tion of innovation activities. In the U.S., skilled work
given innovation. ers, particularly in the services sector, relocate to
larger cities and away from small and medium-sized
ones. Academic research is also more productive and
Demand side: how market forces creative – i.e. more unconventional – in larger and more
generate innovation hubs diverse agglomerations.9
Market economic forces complement labor supply New theories of economic geography have extended
ones as drivers of the geographic concentration of and perfected the argument of the pool of organizations.
innovation activities. The main market economic forces Unlike most traditional spatial analysis, they identify
are generated by the “pool” of organizations – notably geographical concentration as a snowball process
private firms – within a market and the consequent whereby regions progressively draw in supplier firms
economies of transport, scale and scope. and human talent. In its simplest version, regional differ
ences in productivity or economies of scale alone can
This pool is at the heart of productivity differences across explain the divergence in geographical concentration
regions. As with the “unusual individual cases,” histori between two equivalent regions, or explain the reinforce
cal accidents involving innovation breakthroughs by a ment of the concentration in core regions vis-à-vis the
key firm within a local economy, a so-called “anchor” periphery. The basic mechanism is that any confirmed
firm, can be equally important in starting an innovation difference in productivity or innovation levels in a given
ecosystem that then organically grows as both skilled region accrues to generate or confirm the leading posi
labor and related activities relocate. At such moments, tion of the more innovative or productive region.10
key firms can have an idiosyncratic influence on agglom
eration. But this is not always the case. Motorola located According to these theories, agglomerative market
the largest early semiconductor facility in the world in forces are at work when both firms and consum
Phoenix, Arizona, in the 1950s, for example, but this did ers can take advantage of clustering in one location.
not establish that U.S. city as a subsequent center for Agglomerations with large local markets are preferred
IT industries.8 Motorola had believed that it could be a sites for the production of consumer goods due to
geographically isolated first-mover in a technologically the economies of transportation, scale and variety.
innovative industry. It turned out that only those first- Transport economies are in place when local firms
movers that did not isolate themselves from the open can serve a large local market faster and cheaper than
source networks of the emerging Silicon Valley – such distant ones. Similarly, firms supplying large markets
as semiconductor firm Fairchild and computer company benefit from economies of scale by splitting the sunk
Hewlett-Packard – were able to keep up with the rapidly investment costs over more units sold and optimiz
rising technology curve. ing production processes through several iterations.
Consumers in larger markets enjoy a higher variety of
Regions with industrial concentration benefit from goods. Not only can consumers find the exact variety
more complete local labor markets. Firms can find of product they are looking for in a larger market, firms
specialized skills more easily, reducing costs related can specialize in delivering it. These three mecha
to employees’ skill conversion or relocation. Similarly, a nisms – transport, scale and scope – also affect firms
higher concentration of firms is more likely to generate producing intermediate goods locally, which reinforces
new firms. These spin-offs are more likely to be more economies downstream in the local supply chain.11
18
Do knowledge spillovers and technological related technological capabilities and new manage
conditions attract concentration? ment approaches easily transposable to the nascent
IT industry. Princeton and other East Coast hubs had
But market scale and the availability of a skilled work a much narrower technological IT ecosystem based
force do not lead straightforwardly to a region master on few large companies.14
ing the next wave of innovation. Advantages obtained
from previous successful innovation processes do not In this sense, more diversified agglomerations have
assure future technological advantages. a greater probability of successfully transitioning to
a new technological capability than narrowly special
As with large markets and complete labor markets, ized ones.15 The literature abounds with stories of
so-called information and knowledge spillovers are also how narrowly specialized economies are locked into
positive externalities favoring co-location of innovative their technologies and do not transition after negative
firms, academic centers and talented human resourc demand shocks or technology shifts. It seems that
es.12 Knowledge is not restricted to the technological technological innovation is more likely to occur in
and organizational practices of an existing organiza regions with a broader portfolio of technical compe
tion or individual; it may spill from one to the other. tences, especially when it is easy to recombine these.
Firms more successfully exploit economies of scale Dominant industries tend to monopolize talent, supplies
and scope if they learn from the experience of other of economic factors of production, such as capital
firms. Skilled workers disseminate tacit knowledge or entrepreneurship, and attention. Such resource
when they interact with other skilled workers, change concentration potentially crowds out other activities and
organizations or migrate. can channel the evolution of regional economies down
different pathways. For instance, Detroit – the “Motor
Most empirical evidence points to knowledge spillovers City” – is held up as a case of over-specialization. And
being extremely geographically concentrated. This is yet there are highly specialized centers of mechani
mainly due to the high costs associated with codifying, cal engineering and automotive technology that have
exchanging and absorbing knowledge. While informa mastered subsequent waves of technology, such
tion, such as data, flows increasingly freely across orga as Stuttgart in Germany. Boston was once narrowly
nizations and regions, spillovers of knowledge – what is specialized in mill-based industries, but is now a high-
needed, for example, to interpret data – are “stickier.” tech center. The capacity for regional economic evolu
Firms, academic organizations and individuals have to tion is governed by possibilities for moving into related
actively interact, collaborate and, sometimes, move to varieties of technologies and technological capacity.16
make knowledge flow. The concentration of knowledge
spillovers can, therefore, be both a consequence and However, technological relatedness and complementar
a trigger of the agglomeration of innovation. Innovative ity are not the entire story. There are many examples
firms will move to where knowledge spillovers are higher, of regions that capture major new sectors with little
reinforcing spillovers in that region and crowding out technological relation to their previous activities. Los
non-innovative firms to the periphery.13 This joint innova Angeles was not a major mechanical engineering region
tion and spatial co-evolution can determine a regional in the 1920s and 1930s when it became the aircraft-
development path, which can be largely irreversible. engineering center of the U.S. and, by the 1940s, the
world’s biggest aerospace cluster. Los Angeles also
While previous regional technological endowments had no background in the entertainment industry when
are likely to shape subsequent creation of innovation, the movie studios were established there around 1915.
not all innovative regions follow the same trajectory. In Detroit had fewer antecedents in mechanical equipment
the 1930s, both Princeton, New Jersey – site of RCA than Illinois in the 1890s, but rapidly became the center
Laboratories – and Silicon Valley were home to close of U.S. car technology and manufacture.
technological antecedents of the IT industry; but they
developed very different innovative paths. Silicon In these, and many other examples, there were techno
Valley’s remarkable IT innovation trajectory grew out of logical windows of opportunity. These ruptures in tech
the pre-existing and mutually supporting manufactur nological relatedness largely obviate the advantages
ing industries of power grid tubes, microwave tubes of pre-existing agglomeration and create a relatively
and silicon components. These industries enriched flat playing field for a short time in the early days of a
the northern Californian IT innovation ecosystem with technology’s existence.
19
To sum up, the interaction between innovation and not equivalent to Boston or Silicon Valley, the Research
geography reflects the juxtaposition of individual, Triangle Park is known as a leader in a wide variety of
organizational and technological antecedents. high-tech fields and as a model of one of the first and
Saxenian’s (1994) seminal comparison of Boston’s most successful research parks.19 Public policy can
Route 128 and Silicon Valley shows that the types of also affect the geography of innovation more indirectly
entrepreneurship, production organization and system through the R&D system and, in particular, the role
coordination experienced by existing firms and actors of universities and public research laboratories and
in a region will shape how that region evolves economi organizations. In the U.S., from 1875 to 1975, the federal
cally and what kinds of new activities it can generate Land Grant Colleges system extended the geographical
and capture. spread of research universities, while federal funding for
universities reinforced the proliferation of universities.
The California system is perhaps the most successful of
Can policy mold the forces of all, with the public University of California system having
innovation agglomeration? six of the world’s top universities. The same applies to
the geographical distribution of public sector labora
There is little systematic, large-scale evidence for the tories, such as the national laboratories in the U.S. or
success of policies trying to create new local clus the Conseil National de Recherche Scientifique (CNRS)
ters. The last several decades are littered with failed labs in France.
“technopolis” or “the next Silicon Valley” policy initia
tives. Government subsidies might actually attract the Analogously, most of the former middle-income econo
“wrong” kind of firms that have low productivity and mies that are now high-income and highly innovative
depend on subsidies for survival or which are not in fact regions of the world – such as the Republic of Korea,
open to creating networks among local firms for fear of Singapore or Israel – made a successful effort to
leaking intellectual property (IP). Because of the path build top-ranked research universities.20 In China, it
dependency of industry growth and cluster creation, seems likely that the appearance of top world innova
it is questionable how much policy can achieve. As in tion clusters is related to the investments in top world
nature, firms form innovation ecosystems that are not research universities.
easily transplantable or reproducible, as they develop
embedded in territorially-specific institutional settings There are also selective examples of successful
and social fabrics.17 government intervention to generate clusters in
middle-income economies. For example, in 2008,
Nevertheless, the above does not mean that all policy the municipal government of Chongqing, China,
has failed in influencing cluster formation. Indeed, a successfully helped to transplant several smaller
common feature of any national innovation system coastal notebook computer manufacturing clusters
is that market forces of agglomeration are not the into the city. Their policies targeting investments in
only factor shaping the geography of innovation. The infrastructure, labor market organization and other
public sector, as well as the higher-education sector business-friendly actions incentivized new busi
and academic institutions are also key actors that nesses, initiatives and entrepreneurship. However,
shape the innovativeness of countries and regions. the policy moved existing clusters rather than grow
This is particularly true in developing economies ing new ones organically. Other initiatives in China
where public investment is the main driver of research have taken different approaches, depending on the
and development (R&D) expenditures.18 A variety of capabilities and powers of the local administration.
circumstances motivates public sector support for India’s investment in a space program located in
innovation. In some countries and regions, stagnat Bangalore incubated an IT cluster in the same area.
ing productivity has stimulated a revival of industrial Then, with the policy support provided by investment
policy. In many of the most successful former middle- in infrastructure and human capital, the new cluster
income economies, industrial policy with a strong grew organically. All of these clusters started as
innovation component was in evidence during their manufacturing hubs and became innovative to varying
economic ascent. degrees as the manufacturing phase matured. But
the process also involved a significant contribution
In the U.S., a notable successful policy story is the from multinational companies, whose role in global
Research Triangle Park in North Carolina. While perhaps innovation networks will be examined below.21
20
All the public investments required to carry out such agglomeration and concentration of innovation in
strategies are large and must be long-term and appro specific – often metropolitan – areas. Rather than
priately institutionally organized. However, there is an being antagonists, these trends are complementary,
inherent tension between inter-regional equity and the reinforcing each other. If there is any single image that
excellence that is present in virtually every large country can capture this emerging geography it is that of a
with a public higher education system today. As it is globalized hub-to-hub system. The world system of
not practical to have equally well-endowed research innovation links national systems of innovation and
universities in every locality, any successful innovation global firms through a spiky geography of knowledge
policy eventually leads to some internal concentration. creation. The result is a global network of these spikes
or hubs, many of which are better connected to one
In addition, in today’s agglomerated innovation envi another than they are to their national hinterlands in
ronment, certain public sector institutions – especially terms of knowledge creation and diffusion.
universities – are strongly reinforced by market forces
that make some more attractive to students, faculty and International openness is a distinctive trait of leading
funders. This reduces the efficiency of public sector poli agglomerations of geographically concentrated innova
cies for spreading innovation around different regions tion today. But long-distance exchanges of knowledge
and creates the risk that public entrepreneurship will are not a new feature of the economic system. In the first
follow the geographical patterns of the private sector and second industrial revolutions, knowledge and hard
and largely benefit those regions with strong institutions ware travelled, international imitation and rivalry were
and favorable conditions. Unfortunately, innovation active parts of the landscape, and there were always
or industrial policy frameworks have only rarely been networks of people who helped along such knowledge
able to both raise the national level of innovation and exchange. In the past, however, such exchanges often
distribute it relatively evenly within the national territory. involved the display and then possible imitation of what
was created in a rival agglomeration.
Other policies – or a lack of these – can also indirectly
affect the forces behind the agglomeration of innovation. Contemporary knowledge clusters have long-distance
The preferences of highly skilled workers, entrepreneurs ties which have become more organized and extensive
and innovative companies for certain agglomerations over time, and which often involve the co-development
may reflect deliberate cluster policies, but they are also of technologies across agglomerations, both within
not unconnected to tax, social security and education firms and between competing firms.
policies, among others. 22 For instance, successful
innovative regions with highly unregulated real-estate Thus, knowledge-generating agglomerations today are
markets will likely observe a housing price uptick, not self-contained local systems, but rather consist of
which will exclude low-skilled workers and drive them key nodes in dispersed and global networks of inno
to regions with lower-cost housing, as noted above. vation.23 Indeed, highly productive localized innovation
systems are also those that are most tied into long-
distance relationships of various sorts. New knowledge
Why do geographical concentration may be brought into a region through the establishment
and spread coexist? of these linkages. Innovators rely on collaborations both
inside and outside the organizations and the regions
Threaded throughout the discussion above there is where they work.
considerable reference to regional concentration within
a process of global dispersion. This is the other major These networks of geographical spread of innovation
defining characteristic of the contemporary geography will be discussed in the following section.
of innovation. While innovation activity is increasingly
concentrated in metropolitan areas, it is also gradually
spreading to hubs outside the traditional centers in the 1.2 Networks and the global
U.S. and Western Europe. spread of innovation
The empirical evidence simultaneously points both In recent decades, global networks for the production
to the importance of the increasing global nature and delivery of goods and services have expanded
of innovation and to the growing forces driving the greatly. In comparison to previous globalization waves,
21
the current globalization has a much higher proportion It is worth recapitulating that the geographical spread
of intra-industry exchange of both components and final of innovation to peripheral areas of a region or coun
goods within global value chains. Prior to 2000, most try is often limited, because the forces pushing for
of such intra-industry trade took place among a few concentration are too strong. Nevertheless, the same
countries, most notably in the northern hemisphere. But strong concentration forces operating in one urban
since then, it has more and more concerned the rela cluster are at work in others. This can lead to reciprocal
tionships between developing economies and the rest relationships that generate a further spread of innova
of the world. Global production networks often involve tion and knowledge. Regions spreading and receiving
multiple or circular trade flows, with exports wrapped into innovation are likely to remain connected, but these
subsequent outputs and ending up as imports, blurring knowledge and technological flows often skip the
the line between foreign and domestic production.24 peripheral regions of the world and link directly to the
main economic agglomerations.
The current globalization, in other words, involves
intricate forms of interdependency, not just between Figure 1.2 sums up the three main bidirectional econom
economies as a whole, but inside the most delicate ic forces forging the links of an international or inter-
plumbing of the economic system, within and between regional network – market, skilled labor and knowledge.
firms and industries. This is also true of the underlying
innovation networks and ecosystems, which are both
a consequence of global productive integration and, People, companies and ideas link
increasingly, a cause of it. beyond geographical borders
Figure 1.2 Main bidirectional economic forces
Mirroring the growing globalization and complexity of
spreading innovation
production systems is an increasingly dispersed and
complex plumbing system for knowledge produc
tion. The globalization of innovation is a result of Market
Market adaptation
an increased international integration of economic Agglomeration Specialized suppliers
Agglomeration
of economic of economic
activities and the rising importance of knowledge in activities Production networks activities
Sk
economic processes.25 ill e t
dl
ab r ke
or Ma
From the mid-20th century until the Great Recession, Agglomeration

Skilled labor of economic
beginning around 2008, technological activity was Diaspora networks activities
also steadily internationalizing, with new countries Migrant returnees
Knowledge
R&D offshoring Knowledge

emerging in the international system of innovation.
Value chains
More recently, there is some evidence of the selective Professional networks
reshoring of some key R&D and innovation activities Scientific networks
back to home countries. At the same time, however,

Agglomeration
the post-recession period has seen a growing articula of economic
activities
tion of value chains beyond national borders, involving
an increasing share of intra-firm trade flows, with the
knowledge flows they entail.26
The international and inter-regional mobility of skilled

What economic forces explain innovators is a key feature of the contemporary inno
the spread of innovation? vation environment, creating an interpersonal link
between hubs. This mobility may positively stimulate the
The economic forces driving the spread of innovation international dispersion of innovation by strengthening
are very similar to those stimulating its concentration in these innovation networks.27
specific clusters. Economic agents in a given innova
tion hotspot spread innovation to other regions of the The global diffusion of knowledge does not take place
world and vice-versa, which is why the spread of global through random mobility of people, of course, but by
innovation can be thought of as a bidirectional network their moving between places in which they are likely
of knowledge and technological flows. to find the right conditions and the right people to
22
unleash their innovativeness. As noted, these networks Market forces and role of MNCs in
serve not just as a means of dispersion and mobility, internationalization of networks
but as key points of attraction for the skilled. This
talented workforce can reap learning and experience The forces at work in market agglomeration and spread
premiums by being in geographical hotspots with create links within a firm, among different firms or
key network nodes and whose networks are deep. across different organization types. For example,
The ability to acquire more experience and improve market size can make a firm relocate production to
skills is considered one of the main reasons why the reduce transport costs and benefit from economies of
skilled continue to move into the most expensive scale. Technology transfers will necessarily be involved
cities, in spite of high costs of living, contributing to in putting in place the new production site, while inverse
the rapid growth in geographical differences in the knowledge flows will also happen when adapting the
wages of the skilled. Evidence that both international product to local taste or regulations.
and inter-regional “brain drain” is at a very high level
today complements this picture. A key agent in all of these long-distance technological
interactions is the multinational. In addition to access
Labor mobility can take various directions. After to lower costs and foreign talent, MNCs opt to inter
concentrating in a region, skilled migrants often gener nationalize their R&D activities to benefit from other
ate a diaspora network linking the origin and destination market externalities, such as shorter times to bring
regions. Moreover, many highly trained migrants return products to market and to tap into localized areas of
to their original region to apply their higher skills there technological excellence. Reversed flows of FDI can
as entrepreneurs. also benefit established innovation centers. MNCs
from middle-income economies are increasingly using
Saxenian (1999) explores the interaction of people and outward FDI to expand market reach and to capture
investment networks through the mobility of skilled strategic assets, such as technologies, skills, commer
Asian entrepreneurs from and to Silicon Valley. She cial knowledge and brands. Clearly, local technological
explains how skilled workers come to Silicon Valley and competence is only important for attracting FDI of this
acquire human capital and experience; they become kind if the prospective subsidiary is going to engage in
integrated into local networks while continuing to technology-intensive activities.
maintain links back home. For example, Chinese and
Indian U.S.-trained engineers coordinate activities Intra- and inter-firm offshoring of R&D increase
between Silicon Valley technology producers and the corporate innovation performance.29 Key knowledge-
manufacturing and design expertise in their home generating territories around the world are usually both
country regions. As skilled Asian entrepreneurs move home to key firms that construct and participate in
around, they engage in knowledge-sharing, leading these international networks and hosts to foreign firms
to what is termed “brain circulation.” By drawing on wishing to get access to their knowledge-generating
their networks, they also seem to be able to facilitate ecosystems, talent pool and researchers.
investments in new business ventures, highlighting
the parallel movement of networks and foreign direct The existence of a pool of specialized suppliers is also a
investment (FDI) channels. motivation to connect to another region. A given region
may specialize in a certain technology that can benefit
Equally, multinational companies (MNCs) relocating complementary industries, even if located in other
R&D sites offshore in order to benefit from an extraor regions. Bidirectional knowledge flows will be estab
dinary – or cheaper – research labor supply also gener lished between the buyer and specialized supplier in the
ate bidirectional knowledge flows, at least with the form of technical specifications and goods with embed
headquarters’ region. Access to talent and R&D cost ded technology. In industries with a complex supply
are among the main known MNC motivations to interna chain, these links can involve several hubs, building
tionalize R&D. Global patenting is increasingly the result complex and often international production networks.
of the collaboration of large teams operating within the
organizational boundaries of MNCs. For instance, a Arguably, the forces driving agglomeration success
significant share of Chinese and Indian patents at the fully attract MNCs and other firms – especially high
U.S. Patent and Trademark Office (USPTO) are the value-added ones – to particular locations in both
result of collaborations of this type.28 developed and developing economies. The resultant
23
clustering makes these destinations progressively less reasoning behind the previously discussed skilled
dependent on purely cost-based and relative techno labor- and market-spreading forces.
logical endowment considerations. Intangible location
advantages, such as knowledge spillovers, are highly As seen, the internationalization of corporate R&D
concentrated within specific regions, cities and local plays a key role in all of these long-distance techno
systems. The advantages arising from newly vibrant logical interactions, making MNCs one of the most
local innovation ecosystems may benefit the MNCs at important types of organization and network node in
this location, their headquarters and their entire value the international spillover of new knowledge. MNCs’
chain. The resulting knowledge flows are, therefore, international affiliates are gaining more autonomy and
eminently bi- or multidirectional between so-called becoming – where the right incentives are in place
home and host locations.30 – more embedded in regional and local innovation
systems. Increasing autonomy for international affiliates
also means that the choice of the specific subnational
Knowledge spread: organization location becomes more important and driven by a wider
and person-based interactions range of factors other than costs.32 Characteristics of
the regional innovation ecosystem – including its institu
Geographical proximity is not the only source of tions – are particularly important for attracting foreign
knowledge spillovers today.31 Knowledge externalities investment in innovation and technological operations
become forces promoting the spread of innovation and become relevant factors in attracting investments
through organization-level interactions and inter-person at more advanced and knowledge-intensive stages of
or professional-level networks. These organizational global value chains.
and professional connections may be made stronger
by geographical proximity, but the latter is not neces
sary for them to exist. The rise of global innovation networks
Long-distance organization-level interactions – for Knowledge and innovation production have been inter
example, interactions within global value chains or national phenomena for a long time, but only recently
across international scientific networks – lower trans have they become really global.33 Nowadays, actors
action costs within firms and research organizations. located in different countries can carry out innovation
These organized structures can facilitate deep knowl activities in a truly integrated form. Innovation has
edge interactions without the requisite of co-location. become increasingly the outcome of global networks
The effect is enhanced if actors operate within stan connecting dispersed knowledge centers.34
dardized rules or routines set by their organization or
pool of organizations. In 1981, Microsoft opened a It is against this general backdrop of globalization of
campus in Silicon Valley just to connect their Seattle knowledge and innovation – where their production is
operations in Bellevue (and later Redmond) with the done with a higher degree of functional integration – that
effervescence of the Bay Area. the concept of global innovation networks has emerged.
A global innovation network is a globally organized
Likewise, more people-based and less institutional web of collaboration between organizations – firms
ized relations – for example, professional and scien and others – engaged in knowledge production that
tific communities – can also effectively exchange results in innovation. The networks are characterized
knowledge with a commonly agreed set of rules and by (1) their really global spread – and not confined to
routines. This establishes a social proximity – ranging networks based in high-income countries; (2) their
from inter-personal to being part of the same culture networked nature; and (3) the outcome, i.e., innovation.35
or group – among innovators, which lowers interac
tion costs, eases verification and increases trust to Their formation is the result of a knowledge-seeking
exchange and produce new knowledge. strategy by the organizations involved, which makes
global innovation networks different from global
These economic forces overlap and intertwine to the production networks that follow more efficiency- and
extreme point of being inseparable. Similar to the market-seeking strategies. The focus of global innova
forces driving knowledge agglomeration, knowledge- tion networks is, therefore, knowledge exchange and
spreading forces are an intrinsic part of the underlying integration, and subsequent innovation, but not in
24
production or simple manufacturing.36 Global innovation organization-based interaction is often the framework
networks happen largely through the internationaliza where this people-based collaboration can take place.
tion of corporate R&D. Traditionally, multinationals’ internal networks have
been a suitable means to partially overcome barriers
From this perspective, the MNCs can exert a strong related to both geographical distance and differ
influence on the geographical concentration and global ing national cultures. But the more recent decline in
spread – the shape – of global innovation networks by the costs of travel and communication has certainly
deciding where to locate investment, production and favored the rise of person-based international link
knowledge sourcing.37 Geographical conditions and ages without necessarily an accompanying organi
the existing sectoral innovation system are particularly zational structure.43
important as drivers for the most sophisticated and high
value-added stages of supply chains, such as R&D,
design or advanced business services.38 The offshor 1.3 Conclusions
ing of R&D activities has created new interconnected
architectures of innovation and research as well as Innovation has always had geographical concentra
new co-location patterns with production activities. tions or hotspots: Manchester was to the first industrial
This has offered new opportunities for regions and revolution what San Francisco is to the third. However,
cities to link up to different parts or functions of global for a long period between these two revolutions, it
supply chains in ways that promote economic upgrad appeared that within the advanced economies the
ing and innovation. spread of the capacity for innovation was incremental.
The strong concentration of innovation since the end of
At the same time, global participation is a challenge the 20th century thus requires additional consideration.
for weaker regions, given the risk of being locked into
low value-added and low-innovation activities. The Firms used to cluster along supply chains. In the first
geographic unevenness in participation and embed and second industrial revolutions, innovation activity
dedness in global production networks and value clustered together with leading production activities,
chains generates new core–periphery patterns in the making for large industrial cities, some of which also
global geography of innovation. concentrated R&D and product development. Over
the last century, these patterns of agglomeration have
Most of the related literature in the field of international slowly changed. Location choices have become more
business indicates that organization-based linkages determined by shared skill requirements – especially
– both intra and inter firms – are behind the forma in the services sector – such as labor market pooling
tion of global innovation networks.39 International co- across different but related innovation sectors.44 In
inventions – the archetypical indicator for global inno the third industrial revolution, many industries are not
vation networks – have expanded greatly to India and heavily capital-intensive in their production activities,
China since the 2000s, but a large proportion remains and global supply and value chains are far longer and
under the control of firms in the U.S., Japan and a more complex. As a result, leading urban agglom
few Western European countries. This suggests that erations of innovation today specialize in the abstract,
companies can and do slice the R&D process in multiple cognitive and conceptual tasks of R&D and innovation.
stages/segments – as they do for goods – allowing These highly specialized sectors and complementary
new countries to participate in the different segments broad service sectors have displaced the tradition
according to their comparative advantage.40 This facili ally co-located routine production tasks observed in
tates the transformation of existing global value chains the past.45
or production networks into global innovation networks.
The consequences of this new concentration of innova
Furthermore, an increasing number of studies suggest tive activity are far reaching. Ultimately, the geographi
personal relationships, aside from organization-based cal distribution of innovation shapes the economic
ones, are also critical drivers of the formation of global development path of cities and broader regions. One
innovation networks.41 These relationships range from prominent feature of the geography of economic devel
direct person-to-person innovation-related interna opment in recent times is the inter-regional divergence
tional collaborations, to the international mobility of of incomes within countries (see Chapter 5). Generally
scientists, innovators and entrepreneurs.42 However, speaking, large metropolitan areas – the hotbeds of
25
agglomerated ecosystems of innovation – are increas multinational corporations and the true beneficia
ingly outpacing other regions in terms of income growth. ries of globalization, being centers of corporate
decision-making and control, knowledge generation
Yet, there is also divergence within these large metro and exchange, skills and jobs. But their prosperity is
politan areas. Jobs in innovation-related activities tend accompanied by high levels of income inequality and
to pay higher wages than others. Rapid growth in a spatial segregation, leading some to speak of a new
concentrated geographical area and within a particular “urban crisis.” 48
sector may have further effects on the local economy.
While high-skilled jobs create a larger number of low- There is still limited evidence on which to draw firm
skilled jobs, inflows of high-earners, combined with conclusions as to the causes and consequences of the
limited housing supply, often result in growing inequality concentration and spread of innovative activity. This
and falling disposable income for low-earning house may partially reflect the complex nature of innovation
holds.46 Ultimately, this may lead to increased sorting processes and their uncertain impact. However, the
by skill groups into innovative, high-earning areas and long-term consequences deserve careful consider
non-innovative, low-earning areas, excluding the low- ation, even with partial knowledge.
skilled from the opportunities and amenities of living
and working in an innovative environment.47
These patterns seem to prevail in the largest global

innovation hotspots around the world. They are the
primary homes and hosts of major knowledge-based
26
Notes
1 This chapter draws on Crescenzi 16 Frenken et al. (2007). 33 Chaminade et al. (2016).
et al. (2019b). 17 Chatterji et al. (2013). 34 Cano-Kollmann et al. (2016).
2 Acemoglu et al. (2005), Mokyr 18 Mazzucato (2015). 35 Barnard and Chaminade (2011).
(2005) and WIPO (2015). 19 Hardin (2008). 36 Chaminade et al. (2016)
3 Crafts and Venables (2003). 20 See, e.g., Hershberg et al. (2007). 37 Crescenzi et al. (2019a).
4 Storper (2018). 21 Crescenzi et al. (2019b). 38 Alcácer and Chung (2007) and
5 Glaeser and Maré (2001). 22 Davis and Dingel (2019) and Chidlow et al. (2009).
6 Crescenzi et al. (2007). Feldman et al. (2005). 39 Bathelt et al. (2004).
7 Kemeny and Storper (2019). 23 Bathelt et al. (2004), Boschma 40 Branstetter et al. (2014).
8 Scott and Storper (1987). (2005) and Frenken et al. (2007). 41 Lorenzen and Mudambi (2013).
9 Crescenzi et al. (2019b). 24 WIPO (2017). 42 Breschi et al. (2017) and Saxenian
10 Krugman (1991). 25 Archibugi and Iammarino (2002). (1994, 1999).
11 Boschma and Frenken (2006). 26 Crescenzi et al. (2019b). 43 Cano-Kollman et al. (2016).
12 These three externalities are 27 Breschi et al. (2017) 44 Diodato et al. (2018).
referred to as Marshallian 28 Branstetter et al. (2014). See also 45 Crescenzi and Iammarino (2017)
externalities (Krugman, 1991). Chapter 2. and Duranton and Puga (2005).
13 Boschma and Frenken (2006). 29 Nieto and Rodríguez (2011). 46 Moretti (2012).
14 Lécuyer (2006). 30 Iammarino and McCann (2018). 47 Diamond (2016).
15 These are known as Jacobsian 31 Boschma (2005). 48 Florida (2017) and Rodriguez-
externalities (Jacobs, 1961). 32 Cantwell (1995). Pose (2018).
27
References
Acemoglu, D., S. Johnson and J.A. Breschi, S., F. Lissoni and E. Miguelez Crescenzi, R. and S. Iammarino
Robinson (2005). Institutions as (2017). Foreign-origin inventors in the (2017). Global investments and
a fundamental cause of long-run USA: testing for diaspora and brain regional development trajectories:
growth. In Aghion, P. and S.N. Durlauf gain effects. Journal of Economic the missing links. Regional Studies,
(eds), Handbook of Economic Growth, Geography, 17, 1009–1038. 51(1), 97–115.
Vol. 1. Amsterdam: Elsevier, 385–472.
Cano-Kollmann, M., J. Cantwell, T.J. Crescenzi, R., O. Harman and D.
Alcácer, J. and W. Chung (2007). Hannigan, R. Mudambi and J. Song Arnold (2019a). Move On Up! Building,
Location strategies and knowledge (2016). Knowledge connectivity: An Embedding and Reshaping Global
spillovers. Management Science, agenda for innovation research in Value Chains Through Investment
53(5), 760–776. international business. Journal of Flows. Insights for Regional Innovation
International Business Studies, 47(3), Policies. Working Paper. Paris: OECD.
Archibugi, D. and S. Iammarino (2002). 255–262. doi.org/10.1057/jibs.2016.8
The globalization of technological Crescenzi, R., S. Iammarino, C.
innovation: definition and evidence. Cantwell, J. (1995). The globalisation Ioramashvili, A. Rodríguez-Pose and
Review of International Political of technology: what remains of the M. Storper (2019b). The Geography of
Economy, 9(1), 98–122. product cycle model? Cambridge Innovation: Local Hotspots and Global
Journal of Economics, 19(1), 155–174. Innovation Networks. WIPO Economic
Barnard, H. and C. Chaminade (2011). Research Working Paper No. 57.
Global Innovation Networks: Towards Chaminade, C., C. De Fuentes, G. Geneva: WIPO.
a Taxonomy. Paper No. 2011/04. Harirchi and M. Plechero (2016). The
Lund, Sweden: University of Lund, geography and structure of global Crescenzi, R., A. Rodríguez-
CIRCLE laboratory. innovation networks: global scope and Pose and M. Storper (2007). On
regional embeddedness. In Shearmur the geographical determinants
Bathelt, H., A. Malmberg and R., C. Carrincazeaux and D. Doloreux of innovation in Europe and the
P. Maskell (2004). Clusters and (eds), Handbook on the Geographies United States. Journal of Economic
knowledge: local buzz, global of Innovation. Cheltenham: Geography, 7(6), 673–709.
pipelines and the process of Edward Elgar.
knowledge creation. Progress in Davis, D.R. and J.I. Dingel (2019).
Human Geography, 28(1), 31–56. Chatterji, A., E. Glaeser and W. Kerr A spatial knowledge economy.
(2013). Clusters of Entrepreneurship American Economic Review,
Boschma, R. (2005). Proximity and and Innovation. NBER Working Paper 109(1), 153–170.
innovation: a critical assessment. 19013. Cambridge, MA: National
Regional Studies, 39(1), 61–74. Bureau of Economic Research. Diamond, R. (2016). The determinants
and welfare implications of US
Boschma, R. and K. Frenken (2006). Chidlow, A., L. Salciuviene and S. workers’ diverging location choices by
Why is economic geography not an Young (2009). Regional determinants skill: 1980–2000. American Economic
evolutionary science? Towards an of inward FDI distribution in Poland. Review, 106(3), 479–524.
evolutionary economic geography. International Business Review,
Journal of Economic Geography, 6, 18(2), 119–133. Diodato, D., F. Neffke and N.
273–302. doi.org/10.1093/jeg/lbi022. O’Clery (2018). Why do industries
Crafts, N. and T. Venables (2003). coagglomerate? How Marshallian
Branstetter, L., G. Li and F. Veloso Globalization in history: A geographi- externalities differ by industry and
(2014). The rise of international cal perspective. In Bordo, M.D., A.M. have evolved over time. Journal of
co-invention. In Jaffe, A.B. and B.F. Taylor and J.G. Williamson (eds), Urban Economics, 106, 1–26.
Jones (eds), The Changing Frontier: Globalization in Historical Perspective.
Rethinking Science and Innovation Chicago, IL: University of Chicago Duranton, G. and D. Puga (2005).
Policy. Chicago, IL: University of Press, 323–370. From sectoral to functional urban
Chicago Press, 35–168. specialisation. Journal of Urban
Economics, 57(2), 343–370.
28
Feldman, M., J. Francis and J. Jacobs, J. (1961). The Death and Life Rodriguez-Pose, A. (2018). The
Bercovitz (2005). Creating a cluster of Great American Cities. New York: revenge of the places that don’t
while building a firm: entrepreneurs Random House. matter (and what to do about it).
and the formation of industrial Cambridge Journal of Regions,
clusters. Regional Studies, Kemeny, T. and M. Storper (2019). Economy and Society 11(1), 189–209.
39(1), 129–141. Superstar Cities and Left Behind
Places: Disruptive Innovation, Labor Saxenian, A. (1994). Regional
Florida, R. (2017). The New Urban Demand and Interregional Inequality. networks: industrial adaptation
Crisis. New York: Basic Books. Paper presented at the 40th Annual in Silicon Valley and route 128.
Meeting of the Italian Regional Cityscape, 2(2), 41–60.
Frenken, K., F. Van Oort and T. Science Association, L’Aquila,
Verburg (2007). Related variety, Italy, September. Saxenian, A. (1999). Silicon Valley’s
unrelated variety and regional New Immigrant Entrepreneurs. San
economic growth. Regional Studies, Krugman, P. (1991). Increasing returns Francisco, CA: Public Policy Institute
41(5), 685–697. and economic geography. Journal of of California.
Political Economy, 99(3), 483–499.
Glaeser, E.L. and D.C. Maré (2001). Scott, A.J. and M. Storper (1987).
Cities and skills. Journal of Labor Lécuyer, C. (2006). Making Silicon High technology industry and regional
Economics, 19(2), 316–342. Valley: Innovation and the Growth of development: a theoretical critique
High Tech, 1930–1970. Cambridge, and reconstruction. International
Hardin, J.W. (2008). North Carolina’s MA: MIT Press. Social Science Journal, 112, 215–232.
Research Triangle Park. Overview,
history, success factors and lessons Lorenzen, M. and R. Mudambi (2013). Storper, M. (2018). Regional
learned. In Hulsink, W. and H. Dons Clusters, connectivity and catch- innovation transitions. In Glückler,
(eds), Pathways to High-Tech Valleys up: Bollywood and Bangalore in the J., R. Suddaby and R. Lenz (eds),
and Research Triangles. Wageningen global economy. Journal of Economic Knowledge and Institutions. Frankfurt:
UR Frontis Series, 24. Dordrecht: Geography, 13, 501–534. doi.org/ Springer, 197–225.
Springer, 27–51. 10.1093/jeg/lbs017
WIPO (2015). World Intellectual
Hershberg, E., K. Nabeshima and Mazzucato, M. (2015). The Property Report 2015. Breakthrough
S. Yusuf (2007). Opening the ivory Entrepreneurial State: Debunking Innovation and Economic Growth.
tower to business: university–industry Public vs. Private Sector Myths. Geneva: WIPO.
linkages and the development of London: Anthem Press.
knowledge-intensive clusters in Asian WIPO (2017). World Intellectual
cities. World Development, 35(6), Mokyr, J. (2005). The intellectual Property Report 2017. Intangible
931–940. origins of modern economic growth. Capital in Global Value Chains.
The Journal of Economic History, Geneva: WIPO.
Iammarino, S. and P. McCann 65(2), 285–351.
(2018). Network geographies
and geographical networks: Moretti, E. (2012). The New Geography
co-dependence and co-evolution of Jobs. Boston, MA: Houghton
of multinational enterprises and Mifflin Harcourt.
space. In Clark, G.L., M.P. Feldman,
M.S. Gertler and D. Wójcik (eds), Nieto, M.J. and A. Rodríguez (2011).
The New Oxford Handbook of Offshoring of R&D: looking abroad
Economic Geography. Oxford: Oxford to improve innovation performance.
University Press. Journal of International Business
Studies, 42, 345–361.
29
The top 10
collaborative
hotspots of the
world account
for 26% of
all international
co-inventions.
Chapter 2
Global networks of
innovation hotspots
For much of the 20th century, multinational Starting in the 1980s and 1990s, the situation changed.
companies (MNCs) from high-income countries The creation of new scientific and technological knowl
restricted their foreign-based research and edge increasingly required interaction between insti
development (R&D) operations to other rich tutions and organizations, whether public or private,
economies, notably the U.S., Western European national or multinational, irrespective of their location.
countries and, later, Japan. This was in marked Gradually, China, India, Eastern Europe and other
contrast to the trend in manufacturing activities, middle-income economies gained in importance both
which saw increasing outsourcing from richer to as targets for R&D-oriented foreign direct invest
middle-income and developing economies.1 ment (FDI) by multinational firms and as sources of
new knowledge.
The rising need for complex and specialized knowledge

and technological interaction at both national and
international level has resulted – paradoxically – in both
geographical concentration and dispersion of innova
tion creation, as highlighted in Chapter 1. On the one
hand, organizations have sought to locate innovation
activities and interactions wherever high quality and
lower costs are available. On the other, market forces,
economies of scale and the need for more face-to-
face communication and multidisciplinary interaction,
because of the complexity of the interactions, have
pulled in the direction of geographical proximity.
Global innovation networks have been a key centrifugal

force in the geographical distribution of knowledge-
creation activities. Knowledge-seeking FDI does not
target whole countries, but specific locations within
them. Most international collaborations, investments or
movements of skilled workers occur between specific
knowledge-production centers. But global innova
tion networks do not merely span frontiers, they link
specific locations within countries and reinforce the
national prominence of these locations; within national
borders, inter-regional innovation subnetworks coexist
with global ones.
In view of these considerations, it is crucial to under

stand empirically the geographical concentration and
spread of the world’s scientific and technical knowl
edge production and interactions. This requires fine-
grained mapping of innovation activities within national
borders and how these contribute to the worldwide
dispersion of knowledge exchanges. In particular, it is
important to examine whether the growth of national
31
knowledge-production centers or hubs results in an

overall increase in international collaboration and
investment or whether they simply suck-in innovative Box 2.1 Patent and scientific publication
activities to the detriment of other areas in the country geocoded data
or beyond in a zero-sum game. This may be especially
relevant for developing economies, whose national Patent data
innovation systems may become less dependent on
the R&D operations of foreign MNCs, thanks both to The patent data used in this report cover all patent
the strengthening of local firms and the implementation documents – granted or not – filed from 1970 to 2017
of specific public policies to promote local innovation, in all patent offices worldwide and available in the
either substituting or leveraging their national and European Patent Office’s (EPO) PATSTAT database
international collaborations. and WIPO’s Patent Cooperation Treaty (PCT) collec
tions. The unit of analysis is the first filing for a set
Moreover, the globalization of knowledge produces of patent documents filed in one or more countries
imbalances in the distribution of innovative activities and claiming the same invention. Each set containing
within countries. As knowledge-production centers one first and, potentially, several subsequent filings
worldwide gain in importance and intensify their is defined as a patent family. In the analysis, patent
exchanges, cities and regions that do not participate in families are split into those oriented internationally
such exchanges risk being marginalized (see Chapter 5). and those oriented only domestically. Internationally-
oriented patent families refer to applicants seeking
This chapter documents the evolution of global knowl patent protection in at least one jurisdiction other
edge-creating interactions and how the centrifugal than their country of residence. These include patent
and centripetal forces described in Chapter 1 generate families containing only patent documents filed at
global networks of extremely concentrated innovation the EPO or through the PCT. Conversely, domestic
hotspots and specialized niche clusters. It uses a patent families refer only to filings in a home country,
novel database of geocoded scientific publication – for instance, a Japan-based applicant filing only at
scientific articles and conference proceedings – and the Japan Patent Office.
patent data to track developments (see Box 2.1) and
emphasize a series of long-term trends, starting from As far as possible, the geocoding – attributing
the mid-1970s. the geographical coordinates to a given location –
relates to the inventor’s address based on the best
The chapter is organized in four sections. The first available data source within a patent family.2 Many
section examines how internationalized the produc addresses are geocoded at a very precise level – i.e.,
tion of scientific and technological knowledge has street or block – but others only at the postal code
become, with a focus on the rise in the participation or other sub-city level. To remain internationally
of middle-income countries, notably China. It also comparable, but also due to the limited coverage
provides complementary evidence on how knowledge of inventors’ addresses in some national collections,
production is becoming increasingly concentrated the clustering analysis relies only on internationally-
geographically by identifying the main innovation oriented patents.
agglomerations – hotspots and clusters – within each
country. The second section analyzes the scientific Scientific publication data
and technological interactions between countries,
providing further evidence for the globalization of inno The scientific publication data used in this report
vation. It highlights the role of international knowledge comes from records from 1998 to 2017 in the
outsourcing by companies as a driving force behind Science Citation Index Expanded (SCIE) of the Web
the development of global innovation networks. The of Science, the citation database operated by the
third section explores to what extent the two types of Clarivate Analytics company. The analysis focuses
agglomerations concur to form an innovation network on observations referring to scientific articles and
that is properly global. The final section spells out the conference proceedings, which are the bulk of
chapter’s main findings. these data.
32
2 Global networks of innovation hotspots
The report assumes that research conducted for any Two decades of accelerated spread
publication has taken place at the institutions and of knowledge production
organizations to which the authors declare their affili
Figure 2.1 Evolution of patenting (top) and
ation. Virtually all of these locations were geocoded Evolution ofshare
patenting share
publication (bottom) by top economies
at the postal code or sub-city level. In the case of by top economies
authors with more than one affiliation in the same 100
publication, all different addresses are considered.
80
2.1 The two sides of global 60

knowledge production Percent
The accelerating international 40

spread of knowledge creation
Where does knowledge production take place? Is

the geographical spread of such output different to 20
that of other economic activities? Empirical evidence

indicates that activities related to knowledge produc
tion – such as R&D expenditure, patent generation or
0
scientific publication – are typically more geographically Evolution of1980–89
1970–79 publication
1990–99 share
2000–04 2005–09 2010–14 2015–17
concentrated within countries than is the case for other byU.S.
top economies
JAPAN GERMANY OTHER WESTERN EUROPE
key economic activities, the overall population, trade REP. OF KOREA CHINA REST OF THE WORLD
100
or FDI. Despite this higher concentration, the global
tendency is for the degree of international geographic
dispersal of innovation over time.3
80
For most of the period from 1970 to 2000 only three coun
tries – namely, the U.S., Japan and Germany – account
ed for two-thirds of all patenting activity worldwide
60
(Figure 2.1). Adding the remaining Western European
Percent
economies – particularly, the U.K., France, Switzerland

and Italy – takes the figure to around 90 percent.
40
Still, the rest of the world’s share in the production
of new technologies, as reflected in the number of
patents, slowly rose over the three decades, mostly at
the expense of several Western European economies. 20
The rest of the world passed from less than 6 percent

at the beginning of the 1970s to more than 13 percent
in the early 2000s. And only a fraction of this spread
0
was due to the Republic of Korea and China. 2000–04 2005–09 2010–14 2015–17
U.S.
U.S. JAPAN GERMANY
JAPAN GERMANY OTHER
OTHER WESTERN
WESTERN EUROPE
EUROPE
In the last two decades, the trend has accelerated REP.OF
REP. OF KOREA
KOREA CHINA REST OF
CHINA REST
THEOF THE WORLD
WORLD
remarkably for both technological (patents) and scien
tific outcomes. The rest of the world accounts for Source: WIPO based on PATSTAT, PCT and Web of Science data (see
Technical Notes).
almost one-third of all patenting activity in the decade Notes: Other Western Europe excludes Germany. Patent figures based
starting in 2010. Published scientific data has spread on international patent families.
33
even more widely, with the rest of the world going from Russian Federation – held the largest shares of both
less than a quarter of all scientific publication to around innovation outcomes after North America, Western
half in roughly the same period. Europe and East Asia. However, these economies
subsequently suffered a sharp drop in their share of
During this period, the rest of the world – a hetero patents and a small one in scientific publishing.
geneous group that ranges from some high-income
countries, such as Canada or the Republic of Korea, to Each of these regions also reveals high concentration
mostly middle- and low-income economies – outpaced within a few countries, especially for patents. This is the
in its share of knowledge production not only Western case of India and the Islamic Republic of Iran in South
Europe but also the U.S. and Japan. Undeniably, China and Central Asia; Singapore in Southeast Asia; the
and the Republic of Korea represent a substantial part Russian Federation and Poland in Central and Eastern
of this international dispersion, but they do not explain Europe; Brazil and Mexico in Latin America; Israel and
all of it. Indeed, including these two Asian economies Turkey in the Middle East; Australia in Oceania; and
in the same group as Western Europe, the U.S. and Egypt and South Africa in Africa. These regional leaders
Japan still results in the rest of the world increasing its account for much of the little patenting activity happen
share of both knowledge production indicators. ing in their subcontinents. They also concentrate much
of the scientific publishing, particularly Brazil in Latin
What is behind this widening spread? First and fore America and India and the Islamic Republic of Iran in
most, the rise of Asian countries as global innova South and Central Asia.
tion players: since the 2000s, Asia as a whole has
increased its share of total patenting from 32 percent Innovations can differ in their scientific and technologi
to 48 percent and its share of total scientific publish cal value. Seminal and disruptive scientific and tech
ing from 17 percent to 36 percent. This reflects the nological outputs influence subsequent ones and, as a
rise of China and the Republic of Korea and comes result, are more cited. High-income economies spend
despite the relative decline in Japan’s share of patents more on producing such seminal innovative outputs.
and publications. Even if an imperfect indicator of economic value, cita
tions of patents and scientific publication also reflect
Furthermore, considering their low starting point, many how visible and appreciated the research is to other
economies in West, South, Central and Southeast innovators and therefore how valuable.
Asia have seen a remarkable increase in their share
of patenting (Table 2.1). This is also true for scientific Both patent and scientific publication data indicate
publication where the share increased from more than that innovation is more concentrated when it is more
5 percent to more than 10 percent in only two decades. valuable (more cited) (Figure 2.2). In particular, the U.S.
Within these economies, Turkey, Israel, India, Singapore holds a disproportionate share of top-cited patents
and the Islamic Republic of Iran stand out as the largest and scientific publication, dwarfing the shares of other
innovation producers. economies. Still, even here there is a trend toward
dispersion. In the last two decades, the U.S., Japan and
Economies in other continents have also contributed Western Europe have seen less overall concentration of
to the geographical spread of innovation in the last more valuable innovation outcomes. Again, China and
two decades, especially with respect to scientific the Republic of Korea stand out. But other economies
publication. Oceania – mostly pushed by Australia – also contributed to the spread of top-cited innovation,
has seen a small but steady increase in its share of even if the spread has not been as fast as for less cited
scientific publication, although its share of patents scientific publication and patents.
has decreased since the early 2000s. Latin American
and the Caribbean economies have experienced a In sum, China seems to largely explain the world
36 percent increase in their share of scientific publica wide spread of scientific and technological innovation
tion in the last two decades and doubled their share of activities in the last two decades, although many other
patents since the 1970s, although from a very low start countries have contributed to this trend. But many
ing point. African countries had a high relative increase lower-income countries are systematically excluded
in share of scientific publication, but their already very from international innovation.4 Curiously, the recent rapid
small patent share fell. At the outset of this period, rise of China and, to a lesser extent, that of the Republic
Central and Eastern European countries – led by the of Korea, also signifies a global reconcentration of
34
Asia’s share in innovation rises strongly

Table 2.1 Evolution of patenting and scientific publishing, by regions and selected countries
Region Patents Publications

(country) 1970–79 1980–89 1990–99 2000–04 2005–09 2010–14 2015–17 2000–04 2005–09 2010–14 2015–17
SCSE Asia 0.1% 0.1% 0.6% 1.0% 1.6% 2.1% 2.0% 3.2% 4.8% 6.7% 7.5%
India 0.0% 0.0% 0.1% 0.5% 1.0% 1.4% 1.3% 2.0% 2.6% 3.2% 3.5%
Singapore 0.0% 0.0% 0.1% 0.3% 0.4% 0.4% 0.3% 0.4% 0.5% 0.5% 0.5%
CEE 3.2% 3.8% 4.9% 1.1% 1.3% 1.4% 1.3% 5.8% 5.9% 5.8% 5.6%
Russian
0.7% 1.4% 2.7% 0.4% 0.5% 0.5% 0.4% 2.4% 1.9% 1.7% 1.8%
Federation
Poland 0.2% 0.1% 0.1% 0.1% 0.1% 0.2% 0.2% 1.1% 1.3% 1.3% 1.3%
LAC 0.3% 0.3% 0.3% 0.4% 0.5% 0.6% 0.6% 3.0% 3.5% 4.0% 4.0%
Brazil 0.1% 0.1% 0.1% 0.2% 0.2% 0.3% 0.3% 1.5% 2.0% 2.3% 2.3%
Western Asia 0.3% 0.3% 0.7% 1.1% 1.4% 1.6% 1.7% 2.3% 2.8% 3.0% 3.1%
Turkey 0.0% 0.0% 0.0% 0.1% 0.2% 0.3% 0.4% 1.0% 1.5% 1.7% 1.7%
Israel 0.2% 0.3% 0.6% 0.9% 1.2% 1.1% 1.1% 0.9% 0.8% 0.6% 0.6%
Oceania 0.8% 1.1% 1.1% 1.4% 1.3% 0.9% 0.9% 2.4% 2.4% 2.6% 2.8%
Australia 0.7% 1.0% 1.0% 1.2% 1.1% 0.8% 0.8% 2.0% 2.1% 2.3% 2.5%
Africa 0.3% 0.2% 0.2% 0.3% 0.2% 0.2% 0.2% 1.1% 1.3% 1.6% 1.8%
Egypt 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.3% 0.3% 0.4% 0.5%
South Africa 0.2% 0.2% 0.2% 0.2% 0.2% 0.1% 0.1% 0.3% 0.4% 0.4% 0.4%
Total 4.8% 5.8% 7.8% 5.3% 6.4% 6.8% 6.7% 17.8% 20.7% 23.6% 24.9%
Source: WIPO based on PATSTAT, PCT and Web of Science data (see Technical Notes).
Notes: CEE = Central and Eastern Europe; LAC = Latin America and the Caribbean; SCSE Asia = South, Central and Southeast Asia. Patent figures
based on international patent families.
innovation production shares, but in different locations. 50 states concentrate almost 40 percent of inventive
This reconcentration coincides with a similar trend production (patents) and almost 30 percent of scien
observed for R&D expenditure shares after 2008 at tific production (publication). And the U.S. is the least
the onset of the Great Recession, with both China and geographically concentrated among the largest econo
the Republic of Korea increasing their share of global mies. In Japan, three out of 47 prefectures concentrate
R&D expenditure. All in all, innovation production has 56 percent of patents and 35 percent of scientific
increased in volume and spread more globally, but publication. In China, three out of 33 provinces gather
there is still a limited set of countries that produces 60 percent of patents and almost 40 percent of scien
the bulk of it. tific publication. In Europe, concentration is higher, but
the number of regions is smaller. In Germany, three
out of 16 states concentrate two-thirds of the patents
Increasing concentration: a local affair and half the scientific publication. Similarly, three out
of 18 French regions accumulate about 60 percent of
The geographical distribution of inventive and scientific knowledge production.
activities within each country is uneven. In the context
of the increase of innovation production and its inter Regional concentration of patents within these econo
national spread, an interesting phenomenon occurs – mies has increased over the last decade. In all cases
there is no clear evidence that knowledge production but France, the top three regions (Table 2.2) accu
has spread within countries. mulate more patents in the latter 2011–2015 period,
evidencing within-country concentration, not disper
A few administrative areas in each economy often sion. Interestingly, the top three regions are not neces
accumulate the lion’s share of scientific and techno sarily the same in the two periods, but the changes
logical production (Table 2.2). In the U.S., three out of are small. For scientific publication, however, the
35
More the value, more the concentration

Figure 2.2 Evolution of top-cited patents (left) and scientific publications (right)
by top economies and regions
Evolution of top-cited patents share Evolution of top-cited scientific publications share
100 100
75 75
Percent
Percent
50 50
25 25
0 0
9
7
–9
–0
–0
–1
–1
–9
–0
–0
–1
–1
–0
–0
–1
–1
–0
–0
–1
–1
70
00
05
10
15
70
00
05
10
15
00
05
10
15
00
05
10
15
19
20
20
20
20
19
20
20
20
20
20
20
20
20
20
20
20
20
10% Top cited 1% Top cited 10% Top cited 1% Top cited
U.S. JAPAN GERMANY OTHER WESTERN EUROPE REP. OF KOREA CHINA REST OF THE WORLD
Notes: Patent figures based on international patent families.
Shares of top innovation subnational regions within countries

Table 2.2 Top three large administrative areas in patent and scientific
publication concentration by period, selected countries
Country Patents Publications

(level) 1991–95 % 2011–15 % 2001–05 % 2011–15 %
Beijing Guangdong Beijing Beijing
China
Guangdong 42.3 Beijing 60.3 Shanghai 45.5 Shanghai 39.4
(provinces)
Shanghai Jiangsu Jiangsu Jiangsu
Baden-Württemberg Bayern Bayern Nordrhein-Westfalen

Germany
Bayern 63.8 Baden-Württemberg 65.0 Nordrhein-Westfalen 49.4 Baden-Württemberg 50.0
(states)
Nordrhein-Westfalen Nordrhein-Westfalen Baden-Württemberg Bayern
Île-de-France Île-de-France Île-de-France Île-de-France

France
Auvergne-Rhône-Alpes 64.1 Auvergne-Rhône-Alpes 59.9 Auvergne-Rhône-Alpes 63.1 Auvergne-Rhône-Alpes 62.7
(regions)
Grand Est Occitanie Occitanie Occitanie
Greater London Greater London Greater London Greater London

U.K.
Hertfordshire 17.9 Cambridgeshire 23.9 Cambridgeshire 35.8 Oxfordshire 38.7
(counties)
Cambridgeshire Oxfordshire Oxfordshire Cambridgeshire
Maharashtra Karnataka Maharashtra Tamil Nadu

India
Karnataka 51.6 Maharashtra 60.1 Tamil Nadu 36.4 Maharashtra 36.1
(states)
Telangana Telangana NCT of Delhi NCT of Delhi
Tokyo Tokyo Tokyo Tokyo

Japan
Kanagawa 51.5 Kanagawa 56.3 Osaka 35.8 Osaka 35.4
(prefecture)
Osaka Osaka Ibaraki Aichi
California California California California

U.S.
New York 30.8 New York 36.5 New York 28.2 Massachusetts 28.7
(states)
Texas Texas Massachusetts New York
Source: WIPO based on PATSTAT, PCT and Web of Science data.

Notes: Patents and scientific publications were attributed to regions according to the geocoded addresses of inventors and affiliations of authors.
See Box 2.1 and the Technical Notes. Patent figures based on international patent families.
36
top three regions show little change during the two density-based algorithms from the economic geog
periods shown in the figure. China’s top three prov raphy literature.7 In a nutshell, the report uses the
inces are the only ones to evidence some noticeable DBSCAN – Density-based spatial clustering of appli-
geographical spread of scientific production. The fact cations with noise – clustering algorithm to identify
that the location and budget of academic institutions “clusters” separately from the geocoded patent and
are the result of complex and long decision-making scientific publication data. The borders of each
processes can partially explain these stable scientific scientific publication and patent agglomeration are
publication trends. determined using a concave polygon approach. The
overlapping polygons are merged keeping only the
These trends apply not only to the main innovative econ outer borders of all concerned agglomerations. The
omies described in the previous section. In most coun resulting outer areas are referred to as global inno-
tries, a few areas have become truly innovation hotspots, vation hotspots (GIHs) or, more plainly, hotspots.
way ahead of the rest of the country. This is the case in To allow for scientific and technological specializa
India, Australia and several countries in Southeast Asia, tion, the above method is repeated for 25 subsam
the Middle East, Latin America and Africa. ples of the same publication and patent data, which
refer to 12 scientific fields and 13 technological
However, there are substantial difficulties – common ones, respectively.8 Only the resulting polygons of
to all questions of economic geography – in making these 25 iterations not contained within a hotspot
comparisons across countries based on existing are kept. From these, the overlapping polygons are
subnational administrative areas. Administrative areas merged in the same way as for hotspots. The final
can differ markedly in size, population and density of outer areas are referred to as specialized niche
innovative activity, all of which complicate comparison. clusters (SNCs) or, more plainly, niche clusters.
In addition, the administrative borders may not coincide
with the limits of the innovation agglomeration or hub.5 By definition, the resulting areas: (1) are internation-
A given administrative area may even encompass ally comparable, i.e., the same scientific publica
two or more agglomerations, or the agglomeration or tion or patent (specialized) density would have
agglomerations may span several administrative areas, determined the same hotspot (cluster) anywhere
even crossing country borders. in the world; (2) can have different scientific and
technological density, i.e., hotspots and niche
A vast literature in spatial analysis documents this clusters need only scientific publication or patent
well-known issue, which goes under the name of the high concentration, but not necessarily both; (3)
modifiable areal unit problem (MAUP) and its resulting have different specialization density, i.e., niche
statistical distortions.6 The solution requires creating clusters are defined with lower density thresholds
ad hoc comparable areas that can be used in place than hotspots; (4) are distinct geographical areas,
of administrative ones. Box 2.2 provides a detailed i.e., the polygons are non-overlapping within and
explanation of the solution. across hotspots and niche clusters; and (5) have
non-predefined boundaries, i.e., hotspots and
niche clusters can have different sizes and include
more than one city, state/province or country.
Box 2.2
How to measure local agglomerations
of innovation
The report aims to provide an internationally compa Based on this methodology, there are 174 global inno
rable measure of agglomeration of scientific and vation hotspots and 313 specialized niche clusters
technological activities. It makes use of all interna worldwide, which together concentrate 85 percent
tionally-oriented patents from 1976 to 2015 and all of all patents and 81 percent of all scientific articles
scientific publications from 1998 to 2017 to identify and conference proceedings published worldwide.
the main geographical concentration of innovation. The contribution of niche clusters is relatively small.
Of course, these also include collaborations – i.e.,
The agglomerations are organically defined co-inventions and co-publication – with partners
using a cluster identification approach based on outside of these innovation-dense areas.
37
Population density does not ensure high innovation density

Figure 2.3 Patents and scientific articles in the top 35 largest cities
Tokyo
London New York City
Paris
Mexico City
Istanbul Beijing Seoul
Tianjin
Rio de Janeiro Shanghai
Osaka
Buenos Aires Moscow
Los Angeles
Sao Paulo
200,000 Tehran
Shenzhen
Publications (1998−2017)
Xi'an Guangzhou
Delhi
Cairo Mumbai
Kolkata
Bangkok Chengdu
Chongqing
Karachi
20,000
Lima Manila
Dhaka
Jakarta
Lagos
2,000
5 50 500 5,000 50,000 500,000
Patents (1976−2017)
Population (ca. 2017)

GIH SNC LESS INNOVATION DENSE
Min: 10,852,000 Max: 37,800,000
Source: WIPO based on PATSTAT, PCT and Web of Science data (see Boxes 2.1 and 2.2), and top cities from The City Mayors Foundation. Based
on the 35 largest metropolitan populations in the list of largest cities in the world retrieved from The City Mayors Foundation, www.citymayors.com/
statistics/largest-cities-population-125.html, September 2019.
Notes: Size of bubble refers to the metropolitan area population (circa 2017). Axis in logarithmic scale. Due to low scientific publication or patent
values, Kinshasa and Shijiazhuang are omitted from the chart area. Patent figures based on international patent families.
To a great extent, these innovation-dense areas coin technological fields. This is the case of the niche clus
cide with the large, urban, cosmopolitan and prosper ters found in Bangkok, Cairo, Chongqing and Kolkata,
ous areas of the world. As noted, innovation is even among others. Despite concentrating much of their
more concentrated than general economic activity national innovation output, several highly populated
and population. For instance, only 22 out of the 35 metropolitan areas – such as Jakarta, Karachi or Manila
most-populated metropolitan areas in the world are – do not generate sufficient innovation to classify as
part of a global innovation hotspot (Figure 2.3). A huge hosts of a hotspot or niche cluster.
disparity exists among those that are: Beijing, London,
Los Angeles, New York, Seoul and Tokyo concentrate On the other hand, less dense urban areas in many
a large amount of both patents and scientific publish high-income and innovative countries can host high
ing, while Buenos Aires, Delhi, Istanbul, Mexico City, innovation density, especially in some specialized
Moscow, São Paulo and Tehran, for example, are part fields. These niche clusters – such as Ithaca in the
of hotspots concentrating a fair, but appreciably lower, U.S., Stavanger in Norway or Bern in Switzerland – are
output of scientific articles and very few patents. Other highly innovative due to the strong innovation footprint
highly populated urban centers have only enough of local academic institutions, industries or, sometimes,
innovation density in some specialized scientific or a key company. In their specialized fields, these niche
38
Innovation density and urban density largely coincide

Figure 2.4 Worldwide distribution of innovation (GIHs and SNCs) and DMSP nightlight
GIH SNC NIGHTLIGHT
North America, Western Europe and East Asia host most hubs
Figure 2.5 Global Innovation hotspots and specialized niche clusters, by region
Source: WIPO based on PATSTAT, PCT and Web of Science data (see Boxes 2.1 and 2.2). Nightlight data from the U.S. National Oceanic and
Atmospheric Administration’s (NOAA) National Geophysical Data Center.
Note: DMSP = Defense Meteorological Satellite Program.
39
clusters outperform metropoles with much higher urban A few locations concentrate most
and overall innovation densities. inventive and scientific activities
Table 2.3 Concentration of patenting and
Figures 2.4 and 2.5 expand this comparison globally,
publishing among GIHs and among less
based on the world distribution of nightlight as a proxy
innovation-dense countries, 1998–2017
for urban-dense areas.9 As shown in Figure 2.4, the
nightlight is not well distributed across the world or within Top 30 hotspots (as share of all GIHs in the world)
borders. Innovation follows a similar pattern of agglom
eration, but it is even more geographically skewed. These Hotspots (%) 30 (17.2%)
agglomerations or hotspots – which are by definition

Countries (%) 16 (47.1%)
denser in scientific knowledge or patent generation –
typically coincide with the brightest – in terms of night Patents (%) 3,234,850 (69.2%)
light – areas in the world. Niche clusters also coincide
with bright locations, although their specialist nature Scientific articles (%) 10,987,971 (47.8%)
means the urban areas can be less dense.

Top 30 agglomerations in non-innovation dense countries
Europe – particularly the west – has the most homoge Agglomerations (%) 30 (5.0%)
nous territorial distribution of nightlight and, not surpris
ingly, concentrates more than one-third of all global Countries (%) 24 (14.4%)
innovation hotspots and specialized niche clusters in

Patents (%) 11,491 (64.1%)
the world. Despite this, there are still several illuminated
areas without a corresponding innovation agglomera Scientific articles (%) 484,689 (61.0%)
tion. In Europe, Germany, the U.K. and France lead in
quantity of innovation agglomerations, but even they Source: WIPO based on PATSTAT, PCT and Web of Science data (see
Boxes 2.1 and 2.2).
have several dense urban areas without any corre Notes: Only data from 1998 to 2017 reported. Top 30 are calculated
sponding scientific publication or patent density. separately for patent and publication data. Top 30 agglomerations in
non-innovation-dense countries are based on the same methodology
described for GIHs in Box 2.2. Patent figures based on international
North America hosts more than a quarter of hotspots patent families.
and niche clusters, mostly in dense urban areas along

the east and west coasts. Most of the main cities in the contrary, Africa and Latin America have mostly dense
center and south of the country also host innovation urban areas with no corresponding innovation density.
agglomerations; but many dense urban areas – particu
larly in the Midwest and southern U.S. states – do not As shown in Table 2.3, the inventive and scientific activi
have sufficient innovation to host a global innovation ties across locations are highly skewed at all levels of
hotspot or niche cluster. innovation density. The 174 hotspots represent the
most innovative-dense areas in the world; nonetheless,
Asia hosts slightly more than a quarter of the total a limited number – mostly in high-income and middle-
hotspots and niche clusters. Japan, China, the Republic income countries – consistently produce most of the
of Korea and India account for the bulk of Asia’s inno scientific and technological knowledge created within
vation agglomerations. In Japan, and to some degree global innovation hotspots.
the Republic of Korea, there is a high correspondence
between nightlight and innovation agglomeration. Only 30 hotspots in 16 different countries are respon
Despite their numerous innovation agglomerations, sible for the creation of almost 70 percent of the
China and India still have many dense urban areas with patents and around 50 percent of the scientific arti
no corresponding innovation density. cles produced.
The large continental territories of Oceania, Latin Very little inventive and scientific activity is produced
America and Africa host vast areas without dense outside the hotspots and niche clusters, and even less
urban locations. Within the former, Australia has a high is produced outside the few countries hosting these.
co-location of urban and innovation-dense areas, with Indeed, there are more than 160 countries not hosting
virtually no bright locations without a corresponding any hotspot or niche cluster. Even in these less inno
global innovation hotspot or niche cluster. On the vation-dense areas, most of the knowledge produced
40
Persistent concentration of innovation in a few hotspots

Table 2.4 Top three GIH concentration, patents and publications, selected countries
Patents Publications
Country
1991–95 % 2011–15 % 2001–05 % 2011–15 %
Beijing Shenzhen–Hong Kong Beijing Beijing
China Shanghai 36.5 Beijing 52.2 Shanghai 43.9 Shanghai 35.8
Shenzhen–Hong Kong Shanghai Nanjing Nanjing
Frankfurt Frankfurt Frankfurt Frankfurt

Germany Cologne–Dusseldorf 37.4 Stuttgart 29.4 Cologne 34.4 Cologne 34.2
Stuttgart Cologne–Dusseldorf Berlin Berlin
Paris Paris Paris Paris

France Lyon 47.1 Grenoble 42.8 Lyon 51.0 Lyon 49.4
Grenoble Lyon Grenoble Toulouse
London London London London

U.K. Manchester 30.0 Cambridge 35.0 Cambridge 39.8 Oxford 41.8
Cambridge Oxford Oxford Cambridge
Bengaluru Bengaluru Delhi Delhi

India Mumbai 41.9 Hyderabad 46.2 Mumbai 27.7 Mumbai 24.6
Delhi Delhi Bengaluru Kolkata
Tokyo Tokyo Tokyo Tokyo

Japan Osaka 80.5 Osaka 83.4 Osaka 64.3 Osaka 64.8
Nagoya Nagoya Nagoya Nagoya
New York City San Jose–San Francisco New York Boston

U.S. San Jose–San Francisco 19.4 New York City 23.4 Washington DC–Baltimore 21.2 New York 21.4
Boston Boston Boston Washington DC–Baltimore
Source: WIPO based on PATSTAT, PCT and Web of Science data (see Boxes 2.1 and 2.2).
is generated in just a few dense urban areas. Just 30 three patenting hotspots concentrate less inventive
agglomerations located in only 24 different countries activity than two decades earlier.
produce roughly 64 percent of patents and 61 percent
of scientific articles within these non-innovation-dense Overall, concentration of scientific publication has
countries (Table 2.3). Despite this concentration in a few also remained relatively stable at high rates. Within
agglomerations, the gap with the world’s top hotspots is these selected economies, only China and, to a lesser
huge. The volume of patents and scientific publication extent, India show some tendency to dispersion, but
of the top 30 agglomerations in less innovation-dense their top three hotspots still hold at least a quarter and
countries is only 0.4 percent and 4 percent, respectively, a third of all national scientific publication, respectively.
of that of the world’s leading 30 hotspots. Comparing publication and patents, it is interesting to
observe (Table 2.4) how in certain countries, scientific
But even within innovation-dense areas strong national publication is more concentrated than patents (which
differences emerge. Table 2.4 looks at the top three is not the general rule). This is the case of the U.K. and,
hotspots and niche clusters for selected countries in to a lesser extent, France. Both countries host capital
two different periods and the share of patenting and cities that are worldwide centers of scientific production
scientific publication they accumulate in their respective and which lead the way in their respective countries.
countries. First, the list of top three innovation-dense
areas barely differs in time and between patents and
scientific publication, showing the stability of the 2.2 Global networks of
concentration phenomenon. Second, in all countries collaboration and sourcing
shown, the share the top three accumulates is quite
high, ranging from around 20 percent up to more than
80 percent. Next, in the majority of countries the share Just how globalized is collaboration?
of the top three patenting hotspots either remains quite
stable or increases, showing that, within countries, The production of scientific and technological knowl
inventive activities do not spread much geographically edge is increasingly collaborative. Already in 1998,
and in some cases even reconcentrate. Germany and, teams produced the majority of scientific papers. By
to a lesser extent, France are exceptions where the top 2017, lone wolf scientists had become half as important
41
Increased collaborative innovation

Figure 2.6 Inventor (left) and scientific (right) team size, by period
Inventor
Inventorteam
teamsize,
size,bybyperiod
period Scientific
Scientificteam
teamsize,
size,by
byperiod
period
100
100 100
100
80
80 80
80
60
60 60
60
Percent
Percent
Percent
Percent
40
40 40
40
20
20 20
20
00 00
1970–99
1970–99 2000–04
2000–04 2005–09
2005–09 2010–14
2010–14 2015–17
2015–17 2000–04
2000–04 2005–09
2005–09 2010–14
2010–14 2015–17
2015–17
TEAM SIZE:
TEAM
TEAM SIZE: 1
SIZE: 11 2 32 2 4 33 5 46+
4 55 6+
6+ TEAM
TEAMSIZE:
SIZE: 11 22 33 44 55 6+
6+
Source:
Source:WIPO
WIPObased
basedon
onPATSTAT,
PATSTAT,PCT
PCTand
andWeb
WebofofScience
Sciencedata
data(see
(seeBox
Box2.1).
2.1).
Source: WIPO based on PATSTAT, PCT and Web of Science data (see Box 2.1).
as they were 20 years before. The size of the teams is needs; (ii) access a special talent pool; or (iii) simply
also increasing. In 2017, the average scientific paper lower the researcher costs.10
required almost two more researchers – on average
– than 20 years previously (see Figure 2.6). Moreover, Increasingly, collaboration in scientific production
the average team size has shifted upward across the (publication), as opposed to technological produc
board, making teams of six or more scientists the most tion, involves teams from organizations in at least two
common in the production of scientific knowledge. different countries (Figure 2.7). In only two decades,
the share of international scientific collaboration prac
Teams collaborating to achieve technological innova tically increased by half, growing from 17 percent to
tions (patents) are smaller but follow a similar increasing 25 percent of scientific articles published. International
trend, with the average team number doubling since the co-inventorship is a much less frequent phenomenon.
early 1970s. By the mid-2010s, two thirds of inventions Despite the lower shares, however, collaborative inter
were collaborative efforts. All team sizes of inventors national patent production showed an impressive
are increasing at the expense of single-inventor patents. growth trend until the second half of the 2000s, more
than doubling from less than 5 percent to almost
Teams are also increasingly international. As seen in 11 percent. Since 2010, the share has fallen slightly.11
Chapter 1, the forces pushing academia and companies
to cross borders seeking partners for innovation are The fact that international teams account for a higher
manifold. The scientific community has a long tradition percentage of published scientific articles than of
of engaging in international collaboration. MNCs also patents indicates once again that science production
seek efficiency gains from the international division of is more internationalized than technology production.
their R&D and through international collaboration. For Figure 2.8 breaks down data for international inventive
instance, they may collaborate with R&D teams in other and scientific teams by country, for the top innovative
countries to: (i) adapt technologies to different market countries worldwide. With the exception of Japan and,
42
Collaboration for innovation is to a lesser extent, the Republic of Korea, most top-filing
increasingly international countries show a large international co-inventorship
Worldwide share of international share. The U.S. and Western European countries
co-inventorship
Figure and co-publishing
2.7 International co-inventorship and
show mostly a rising trend. Smaller economies with
co-publishing, percent
30 internationally linked and dense urban and innovation
areas – such as Switzerland – are very prone to engag
ing in international collaborations. India also shows a
high rate of international co-inventorship. In East Asia’s
20 top economies things are different. Before the 2000s,
the share of international co-inventorship in China
Percent
was extraordinarily large, but the volume was small.

Thereafter, when the volume of Chinese patenting
10
picked up, the share of international co-inventorship
dropped dramatically, becoming comparable to the
very low shares of Japan and the Republic of Korea.
Trends for international co-publication reveal a very

0
1980 1990 2000 2010 different picture. All main scientific publishing countries
INTERNATIONAL
INTERNATIONAL CO-INVENTORSHIP
CO-INVENTORSHIP
have larger shares of international co-publication than
INTERNATIONAL
INTERNATIONAL SCIENTIFIC
SCIENTIFIC CO-PUBLISHING
CO-PUBLISHING
co-inventorship, with the exception of India. Moreover,
Notes: Int. co-inventorship = share of patents with more than one inventor located in at least two countries;
these shares increased steadily over the period.
Source: WIPO= based on PATSTAT,
articles withPCT
more and Web of Science
located in atdata
int. co-publications
(see Box 2.1).
share of scientific than one affiliation least two countries.
However, the figures show East Asian countries being
Notes: Int. co-inventorship = share of patents with more than one less internationally open than the U.S. and Western
inventor located in at least two countries; int. co-publications = share
of scientific articles with more than one affiliation located in at least two Europe when it comes to scientific publishing too.
countries. Patent figures based on international patent families.
Large economies are highly internationalized

FigureInternational
2.8 International co-inventorship
Internationalco-inventorship
co-inventorshipshare,
share,by
(left)
bycountry
and co-publication
country
(right),
International
by countryshare,by
Internationalco-publication
co-publicationshare, bycountry
country
7575
4545
6060
3030
Percent
Percent
Percent
Percent
4545
1515
3030
00 1515
1970–99
1970–99 2000–04
2000–04 2005–09
2005–09 2010–14
2010–14 2015–17
2015–17 2000–04
2000–04 2005–09
2005–09 2010–14
2010–14 2015–17
2015–17
U.S. U.S.
U.S.
JAPAN JAPAN
JAPAN SWITZERLAND
SWITZERLAND
SWITZERLAND GERMANY
GERMANY
GERMANY U.S.
U.S. JAPAN
JAPAN SWITZERLAND
SWITZERLAND GERMANY
GERMANY
CHINA CHINA
CHINA
INDIA INDIA
INDIA
REP. OF REP.
REP.OF
KOREA OFKOREA
KOREA CHINA
CHINA INDIA
INDIA REP.
REP.OF
OFKOREA
KOREA
Source:
Source:WIPOWIPObased
basedononPATSTAT,
PATSTAT,PCT PCTandandWeb
WebofofScience
Sciencedatadata(see
(seeBoxBox2.1).
2.1).
Source:
Notes:
Notes: Int. WIPO based on
Int.co-inventorship
co-inventorship PATSTAT,
= =share PCT with
shareofofpatents
patents and Web
withmore ofthan
morethan Science
one data located
oneinventor
inventor (see Box 2.1).
located ininatatleast
leasttwo
twocountries;
countries;
int.
int.
co-publication
co-publication
Notes: = =share
shareofofscientific
Int. co-inventorship scientific
= sharearticles
articles
with
withmore
of patents more
withthan
thanone
more one affiliation
than affiliationlocated
locatedin
one inventor inatatleast
located least
intwo
two
at countries.
countries.
least two countries; int. co-publication = share of scientific
articles with more than one affiliation located in at least two countries. Patent figures based on international patent families.
43
Concentration and spread of international collaborations

Figure 2.9 International co-inventorship (left) and international
co-publication (right) by country pairs, 1998–2002 and 2011–2015
International co-inventorship by country pairs, 1998–2002
International co-inventorship by country pairs, 2011–2015
2,000 5,000
Notes: Int. co-inventorship = share of patents with more than one inventor located in at least two countries; int. co-publication = share of scientific
articles with more than one affiliation located in at least two countries. Only top 10 percent international links of each period reported. Bubbles report
the share of links only for selected countries and regions. Patent figures based on international patent families.
44
International co-publication by country pairs, 1998–2002
International co-publication by country pairs, 2011–2015
10,000 40,000
45
A select club of innovation outsourcing recipients

Figure 2.10 Companies’ patenting with inventors in a different country (%), selected regions
1970–1989 1990–1999
Company's country Company's country
Western Europe
Western Europe
Rep. of Korea
Rep. of Korea
Western Asia
Western Asia
SCSE Asia
SCSE Asia
Germany
Germany
Japan
Japan
China
China
Africa
Africa
India
India
CEE
CEE
U.S.
LAC
U.S.
LAC
U.S. U.S.
Japan Japan
Germany Germany
Western Europe Western Europe
Inventor's country
CEE Inventor's country CEE

Rep. of Korea Rep. of Korea
China China
India India
Western Asia Western Asia
SCSE Asia SCSE Asia
LAC LAC
Africa Africa
<.5% .5–1% 1–5% 5–7.5% <.5% .5–1% 1–5% 5–7.5%

7.5–10% 10–15% >15% 7.5–10% 10–15% >15%
2000–2009 2010–2017
Company's country Company's country
Western Europe
Western Europe
Rep. of Korea
Rep. of Korea
Western Asia
Western Asia
SCSE Asia
SCSE Asia
Germany
Germany
Japan
Japan
China
China
Africa
Africa
India
India
CEE
CEE
U.S.
LAC
U.S.
LAC
U.S. U.S.
Japan Japan
Germany Germany
Western Europe Western Europe
Inventor's country
Inventor's country
CEE CEE
Rep. of Korea Rep. of Korea
China China
India India
Western Asia Western Asia
SCSE Asia SCSE Asia
LAC LAC
Africa Africa
<.5% .5–1% 1–5% 5–7.5% <.5% .5–1% 1–5% 5–7.5%

7.5–10% 10–15% >15% 7.5–10% 10–15% >15%
Source: WIPO based on PATSTAT and PCT data (see Technical Notes).
Notes: CEE = Central and Eastern Europe; LAC = Latin America and the Caribbean; SCSE Asia = Southern (excluding India), Central and Southeast
Asia; These regions are closely based on the geographic regions from the U.N. Statistics Division’s (UNSD) methodology (unstats.un.org, accessed
March 2019). The only differences are that CEE includes all countries in the UNSD’s Northern and Southern Europe categories not included in
Western Europe and that SCSE includes Mongolia. Other Western Europe excludes Germany. Western Europe includes the 15 economies that were
members of the EU prior to May 1, 2004, along with Andorra, Iceland, Liechtenstein, Malta, Monaco, Norway, San Marino and Switzerland. Patent
figures based on international patent families.
46
Innovation within hotspots is more likely to be international

Figure 2.11 Percentage of international patent (left) and publication
teams (right), inside vs. outside GIHs and SNCs
2020
4040
1010
2020
00 00
1970–99
1970–99 2000–04
2000–04 2005–09
2005–09 2010–14
2010–14 2015–17
2015–17 2000–04
2000–04 2005–09
2005–09 2010–14
2010–14 2015–17
2015–17
INSIDE INSIDE
INSIDE
INSIDE (TOP INSIDE
INSIDE
10% (TOP
(TOP10%
CITED) 10%CITED)
CITED)
OUTSIDE INSIDE
INSIDE
OUTSIDE (TOP 10% CITED) INSIDE
INSIDE(TOP
(TOP10%
10%CITED)
CITED)
OUTSIDE
OUTSIDE OUTSIDE
OUTSIDE(TOP
(TOP10%
10%CITED)
CITED) OUTSIDE
OUTSIDE OUTSIDE
OUTSIDE(TOP
(TOP10%
10%CITED)
CITED)
International collaboration is also concentrated among a these central economies made up only 2 percent
few main countries, although concentration is decreas of all international co-inventions in 1998–2002 and
ing as new stakeholders enter the network (Figure 2.9). 7 percent in 2011–2015. The subnetwork for scientific co-
Scientific co-publishing only between the U.S., Western publication is slightly greater, starting from 5 percent in
Europe and Japan accounted for 54 percent of all inter 1998–2002 and reaching 13 percent of all international
national co-authorships in 1998–2002 and 42 percent in ties in 2011–2015. Some larger economies outside the
2011–2015. Co-inventorship among these three regions big three – such as China, India, Singapore and, to
was 69 percent of overall international co-inventorship a lesser extent, Argentina, Australia, Brazil, Mexico
in 1998–2002 and 49 percent in 2011–2015. and South Africa – have increased their participation
in the subnetwork, although mostly for scientific co-
The three are also involved in most of the collaboration publication. But their connections still mostly involve
undertaken with other economies (Figure 2.9). While one of the big three – particularly the U.S. and Europe –
collaboration within Europe is increasingly important, rather another non-core location.
the U.S. is the main partner for most European coun
tries. Canada and the U.S. – certainly due to geographic Overall, the collaboration trends suggest that the
and cultural proximity – represent one of the strongest globalization of inventive activities mostly concern the
linkages in international collaboration networks at all U.S. and Western Europe along with China and India.
times. Most of the remaining Canadian ties are with
Western Europe, with few connections elsewhere.
New entrants to these networks – such as China, India, MNCs seek innovation further afield
Australia or Brazil – also mostly link with these three
economies, typically with the U.S. and a few Western From the late 1990s, as discussed in Chapter 1, MNCs
European countries, such as the U.K. and Germany. increasingly started to outsource R&D activities to
middle-income, developing economies, such as
Collaboration between countries and economies China, India and countries in Eastern Europe.12 While
outside the U.S.–Western Europe–Japan triangle is initially they were adapting their technologies to local
much sparser. International co-inventions not involving market needs, they slowly moved toward cutting-edge
47
Dispersion of scientific publication, reconcentration of patenting

Figure 2.12 GIHs’ and SNCs’ share of co-inventorship (left) and co-publication (right) interactions,
GIHs’
by and SNCs’
partner share of co-inventorship
location GIHs’ and SNCs’ share of co-publication
interactions, by partner location interactions, by partner location
100 100
80 80
60 60
Percent
Percent
40 40
20 20
0 0
1970–99 2000–04 2005–09 2010–14 2015–17 2000–04 2005–09 2010–14 2015–17
INTERNATIONAL NATIONAL INTERNATIONAL NATIONAL

INTERNATIONAL
LOCAL NATIONAL LOCAL
NO COLLABORATION NO COLLABORATIONLOCAL NO COLLABORATION
R&D – comparable to that undertaken in high-income Western European economies – such as Germany,
economies – and developing new products for the France and the U.K. – come the closest, while compa
worldwide market.13 The dynamism of certain middle- nies from the main East-Asian countries – i.e. Japan,
income countries was a great attractor of R&D-related the Republic of Korea and China – are far less inter
FDI, especially in India and China. nationalized.
Outward R&D from the U.S. has increased more than There is a clear pattern of companies from all around
fivefold in the last 25 years, with most of this innovation- the world increasing and widening their patenting with
related investment going to Germany, the U.K., Japan, foreign inventors. However, as noted, most international
Canada or France.14 The trend for U.S. companies patent sourcing still happens between companies
patenting with foreign inventors has followed a very and inventors from high-income economies, particu
similar pattern (Figure 2.10). In the 1970s and 1980s, larly from the U.S., Japan and Western Europe. Within
only 9 percent of patents filed by U.S. companies had these, Japanese companies are the least foreign-
foreign inventors; by the 2010s, this share had risen oriented, while U.S. companies rely considerably on
to 38 percent. The outsourcing of technology from Japanese inventors.
Canada, Japan and Western Europe kept growing until
the early 2000s, before flattening. Since then, most of In the last two decades, China and the Republic of
the increase in innovation outsourcing by U.S. compa Korea have made a case to be added to this select
nies has taken place elsewhere, mainly in China, India group. They certainly have the volume of companies’
and, to a lesser extent, Israel. So, a large part of the patenting and of inventors participating in patenting by
U.S. knowledge-diversification strategy has involved foreign companies. Companies from the Republic of
expansion to non-high-income countries. Korea rely more intensively on Japanese and U.S. inven
tors than the latter do on Korean companies. Chinese
Internationalization of R&D has not been limited to companies used to rely intensively on Japanese inven
U.S. companies, although no other large economy has tors during the 1990s, but since the 2000s they have
been so open to such collaboration (Figure 2.10). Large shifted to an increasingly nationally-oriented profile.
48
Differences in country trends, particularly for patents

Figure 2.13 GIHs’ and SNCs’ share of co-inventorship (top) and
GIHs’ and SNCs’ share of co-inventorship interactions
co-publication (bottom) selected
by partner location, interactions, by partner location, selected countries
countries
100
75
Percent
50
25
0
1970–99
2000–04
2005–09
2010–14
2015–17
1970–99
2000–04
2005–09
2010–14
2015–17
1970–99
2000–04
2005–09
2010–14
2015–17
1970–99
2000–04
2005–09
2010–14
2015–17
1970–99
2000–04
2005–09
2010–14
2015–17
1970–99
2000–04
2005–09
2010–14
2015–17
1970–99
2000–04
2005–09
2010–14
2015–17
China Germany India Japan Rep. of Korea Switzerland U.S.
GIHs’ and SNCs’ shareNATIONAL
INTERNATIONAL of co-publication
LOCALinteractions,
NO COLLABORATION
by partner location, selected countries
100
75
Percent
50
25
0
2000–04
2005–09
2010–14
2015–17
2000–04
2005–09
2010–14
2015–17
2000–04
2005–09
2010–14
2015–17
2000–04
2005–09
2010–14
2015–17
2000–04
2005–09
2010–14
2015–17
2000–04
2005–09
2010–14
2015–17
2000–04
2005–09
2010–14
2015–17
China Germany India Japan Rep. of Korea Switzerland U.S.
INTERNATIONAL NATIONAL LOCAL NO COLLABORATION

INTERNATIONAL NATIONAL LOCAL NO COLLABORATION
49
Different MNCs have different connectivity strategies

Figure 2.14 Global co-inventor network of selected companies
São José dos Campos Nürnberg
San Jose–San Francisco Munich
Los Angeles Köln–Dusseldorf
Minneapolis Berlin
New York City Frankfurt
Seoul Stuttgart
Campinas Braunschweig
Chicago Regensburg
Detroit–Ann Arbor Karlsruhe Siemens

Embraer
Embraer Siemens
Florianópolis Orlando
Montréal Bamberg
New Haven New York City
Rio de Janeiro Aachen

2010s 2010s
Herning 2010s
São Paulo 2010s
2000s 2000s
2000s
Washington–Baltimore 2000s Grimsby
0 10 20 50 60 0 10 20 30
Percent Percent
San Jose–San Francisco Bengaluru
New York City Hyderabad
Chicago Delhi
Zürich Mumbai
London Chennai
Boston Pune
Los Angeles Thiruvananthapuram
Seattle Kolkata
San Diego Lucknow

Google
Google Infosys
Infosys
Sydney Chandigarh
Beijing Mysore
Washington–Baltimore Tank
Tel Aviv Milan

2010s 2010s
Bengaluru 2010s Dallas 2010s
2000s
2000s
2000s
2000s
Seoul Reykjavík
0 5 10 30 50 0 6 12 18 50 60
Percent Percent
Shenzhen–Hong Kong Tokyo
Ottawa Osaka
Beijing Malmö
Munich San Diego
San Jose–San Francisco Nagoya
Stockholm Beijing
Shanghai San Jose–San Francisco
New York City Fukuoka
San Diego London

Huawei
Huawei Sony
Sony
Dallas New York City
Chicago Kitakata
Boston Eindhoven
Chengdu Helsingborg
2010s 2010s
Wuhan 2010s Los Angeles 2010s
2000s
2000s 2000s
2000s
Los Angeles Sendai
0 2 4 80 90 0 3 6 9 70 80
Percent Percent
Notes: Top 15 GIHs by inventor locations of patents having the company as applicant. Patent figures based on international patent families.
50
Chinese companies are currently only slightly more The same gap applies to co-inventions inside and
open to international inventors than are Japanese ones. outside hotpots and niche clusters. In the second half
of the 2010s, 11 percent of inventions from hotspots
Nevertheless, despite the increase seen in recent and niche clusters had international partners – and
decades in outsourcing by MNCs involving middle- almost 16 percent in the case of top-cited patents –
income developing countries, companies from the while only 6 percent of patents originating outside
latter are still more likely to draw on the innovation of of these had an international co-inventor. However,
high-income economies than the other way around. there is no evidence of the gap increasing. In fact,
Companies from India, Asia, Central Eastern Europe, international co-invention inside and outside agglom
Latin America and Africa rely intensively on the ingenu erations shows a similar stagnating and to some
ity of inventors in the U.S., Western Europe, China and, extent declining trend, starting in the second half of
to a lesser extent, Japan and the Republic of Korea the 2000s and probably linked to a wider slowdown
to create for them patentable technologies. It is worth in globalization (see below).
recalling, however, that companies in these economies
have low patenting volumes in comparison to those in Figure 2.12 shows several noteworthy patterns. As
the U.S., Western Europe, Japan, China and Republic discussed in Section 2.2, the percentage of scientific
of Korea. Last and not least, there is much less direct and inventive output in these innovation-dense agglom
patenting activity between companies and inventors erations that does not involve any local, national or
from non-high-income countries. international collaboration has decreased. Inventions
with a single inventor went from one-third in the 1970s
There is some degree of regional collaboration. However, and 1980s to less than a quarter by 2017. Scientific
this follows the same pattern described above. Mexican publication by a sole author went from more than
companies source more intensively from inventors 40 percent in the early 2000s to less than 25 percent
in the U.S. and Canada than the other way around. in the second half of the 2010s. The more the hotspots
The same applies to Germany, France and the U.K. and niche clusters collaborate, the denser the network
in Europe, particularly with Central Eastern Europe. of knowledge they create.
Companies from all Asia reach out more intensively to
inventors in Japan, the Republic of Korea, China and, In other respects, the picture differs depending on
to some degree, India than vice versa. Less markedly, whether it is inventive or scientific output. For patents,
inventors in Brazil and South Africa appear as regional the share of local-only teams is larger than that of
sources for Latin American and African companies. national and international ones, while this is not the
However, companies in Asian, Latin American or African case for scientific publications. Nevertheless, for
non-high-income economies interact mostly with inven scientific publication, international co-publication
tors outside their respective continents, typically in the has continuously grown faster than national and
U.S. and Western Europe. local collaborations. The same trend is observed for
patents from the early 1980s until the second half of
the 2000s.15
2.3 Local innovation and global
networks of innovative hubs Since around 2005, however, there has been a fresh
rise in the share of local-only patents. This change
coincides with a slowdown in the pace of globaliza
The globalization of agglomerations tion and internationalization generally, as reflected
in slower growth of trade, FDI flows and financial
Not only do hotspots and niche clusters concentrate integration. It also coincides with a decrease in the
more scientific publication and patent output, they also share of patents generated by teams that are national
collaborate more internationally (Figure 2.11). The differ but not just local. The explanation for the latter could
ence is even greater for highly cited patents and scientific be that part of the slowdown in the globalization of
articles. During the last two decades, international scien knowledge creation and innovation has to do with
tific collaboration went from 19 percent to 29 percent of the rise of local hotspots rather than with the devel
all scientific articles produced inside innovation-dense opment of new national innovation systems. As will
areas and the most-cited within this international collabo be shown, this pattern is stronger in some specific
ration went from 28 percent to 43 percent. Asian countries.
51
Global network of innovative agglomerations: a small world?

Figure 2.15 Top 10 percent co-invention ties among GIHs and SNCs, 2011–2015
52
Source: WIPO based on PATSTAT,

PCT and Web of Science data
(see Boxes 2.1 and 2.2).
Notes: Only the world’s
10 percent largest links reported.
Green lines connect GIHs/SNCs
from the same country and purple
lines connect those from different
countries. Bubbles represent
the top 10 hotspots in terms
of connectivity volume. Patent
figures based on international
patent families.
53
Agglomerations in a few economies are central to the global innovation network

Figure 2.16 Patent co-invention network, 2001–2005 and 2011–2015
U.S. GERMANY JAPAN REP. OF KOREA U.K. CANADA FRANCE ITALY CHINA OTHER
Notes: Only the world’s 10% largest links reported. Bubble size reflects patent volume. Bubbles positioned according to their network centrality.
Patent figures based on international patent families.
The situation also varies considerably across countries, countries, the share of international co-inventions has
as shown in Figure 2.13, which extends the analysis of grown or only slightly stagnated in recent years.
Figure 2.12 by showing the breakdown for hotspots and
niche clusters in a selection of top innovating countries. MNCs can have very different needs and strategies
The country-specific trends for scientific publication for where to source for talent, and these can change
largely follow what was observed in Figure 2.12, with over time (Figure 2.14). For example, in the 2010s,
virtually all countries showing similar patterns and San Jose–San Francisco accounted for 53 percent
increases in collaboration. There are some differences, of Google’s patents. Similarly, Nuremberg – the most
however. In the U.S., Japan, Germany and Switzerland, important source of patents for Siemens – accounted
the increasing share of international co-publication for 32 percent during the same period. As expected,
is the principal cause of a reduction there in non- Tokyo and Shenzhen–Hong Kong are the most impor
collaborative scientific research. China, India and, to tant sources of invention for Sony and Huawei, concen
some extent, the Republic of Korea have seen less trating 71 and 81 percent, respectively. Interestingly,
vibrant growth of international scientific collaboration. comparing figures for the 2010s to the 2000s, Google
In these countries, the decline in the share of non- and Siemens have concentrated more inventive activi
collaborative scientific publication largely reflects an ties within their top hubs, whereas the reverse holds
increase in national and local collaboration. for Sony and Huawei.
In line with the trends for MNCs in the previous section, MNCs from middle-income countries – such as
the trends for patent co-inventorship vary substan Brazil or India – also seek out talent in different ways.
tially across countries. Some countries – like India or Technology services company Infosys has a wide
Switzerland – can be extraordinarily open to interna spread but predominantly Indian network. Brazilian
tional co-invention, with the Republic of Korea, Japan plane-maker Embraer remains very concentrated in
and, more recently, China at the other extreme. There São José dos Campos, which is also the company’s
has been a noticeable drop in the share of international headquarters. But in the 2010s, Embraer replaced
teams in patent production in some countries, particu its second main national hub, São Paulo, with more
larly in China, due in the latter case to a sharp growth international connections, including San Jose–San
in local-only co-invention. However, for the majority of Francisco, Los Angeles or Seoul, among others.
54
Size plays a role in network centrality, but it is not everything

Figure 2.17 SNC network and the global innovation subnetworks
of Los Angeles and Daejon, 2011–2015
U.S. GERMANY JAPAN REP. OF KOREA U.K. CANADA FRANCE ITALY CHINA OTHER
Notes: Bubble size reflects patent volume. Bubbles positioned according to their network centrality. Grayed bubbles do not belong to
the subnetwork.
55
Global network of hotspots international gatekeepers. With obvious proximity

and niche clusters advantages, Canada has a national network well inte
grated to the U.S. one. The role of Canadian hotspots
Innovation-dense agglomerations worldwide form a in the overall North American network contrasts with
network – within and outside their own countries – that the absence of equivalent Mexican co-invention ties.
concentrates most inventive and scientific activities, to
the possible detriment of non-agglomerated actors.16 Notwithstanding the dominance of Bengaluru, India has
In particular, these innovative agglomerations form a a fairly active national innovation network, with sever
thick web of national and international ties between al hubs directly connecting internationally. Similarly,
hotspots and niche clusters in the U.S., Europe and despite its remote location and vast territory, Australia
Asia. Only 10 hotspots account for 26 percent of all manages to have several hotspots that connect inter
international co-inventions between hotspots of the nationally and a fairly interconnected national network.
world (Figure 2.15). These are San Jose–San Francisco, One region less connected than the rest is Latin
New York, Frankfurt, Tokyo, Boston, Shanghai, London, America, where the large majority of connections of
Beijing, Bengaluru and Paris. its few hotspots and niche clusters are with leading
economies outside the region. There are no national
Figure 2.15 also depicts the top 10 percent co-invention or regional networks in Latin America comparable to
links between all the world’s global innovation hotspots those depicted for other regions and countries.
and specialized niche clusters. Even in the U.S., niche
clusters and smaller hotspots often have only national The above discussion demonstrates that it is not only
connections. But despite their extensive geographical geography that shapes the global innovation networks.
spread, global innovation hotspots and specialized niche From a network analysis perspective, an innovation
clusters in the U.S. form a far denser national innova agglomeration is more “central” within a global network
tion network than is the case in the rest of the world. the more international connections it concentrates.
Nevertheless, in the U.S., the larger hotspots concen Figure 2.16 depicts such centrality by grouping the
trate the greater portion of both national and international hotspots and niche clusters with most connections
connections with other hotspots and niche clusters. in the center and scattering the less connected ones.
In Europe, a similar pattern can be detected. A few As noted, U.S. agglomerations are among the more
large hotspots in each country act as gatekeepers connected nodes, hence more central in the networks
that connect the national innovation system to global in both of the periods (Figure 2.16). In the center of the
innovation networks. Clear examples can be found in picture are other global innovation hotspots which are
France, with Paris connecting other French cities with arguably highly connected, such as Tokyo, London,
the rest of the world, and in the U.K., with London Shanghai, Beijing, Seoul or Paris. But they are much
being a central actor. Germany shows some hier less central than the U.S. hotspots. The network also
archical structure too, though access points to the evolved over time, getting more nodes and more
global innovation networks are more numerous and connections and denser at its center.
the national innovation network is very dense. Japan
and the Republic of Korea also stand out, with very Size plays only a limited role. Smaller clusters are
dense national innovation networks, although their connected to big and highly connected ones in the
international connections are less widespread and same country, reflecting the hierarchical pattern
mostly directed to the U.S. and the main Western discussed earlier. This is clearly the case for agglom
European hotspots. erations in the U.K., Japan and the Republic of Korea.
On the other hand, several hotspots that are larger
Hotspots and niche clusters in the remaining regions or similar in size to the top U.S. agglomerations – for
of the world have substantially fewer connections example, Tokyo – fail to occupy the same kind of central
than those in the U.S., Western Europe, Japan and position in the global network. This reflects the lower
the Republic of Korea, although among those that do, international connectivity of Japanese hotspots.
China, India, Canada and Australia stand out. China
has a dense national innovation network where a Figure 2.17 depicts subnetworks of the same 2011–2015
hierarchical structure is also evident, with Shanghai, co-invention network presented in Figure 2.16. It shows
Beijing and Shenzhen–Hong Kong acting as the top the subnetwork of all niche clusters by graying the
56
network connections of all hotspots. As is apparent, Part of this limited geographical spread of knowledge
these specialized innovation-dense areas cannot activities is due to the emergence of global innovation
compete with the hotspots in volume of connections. networks, which first link more traditional innovation
The few connections among niche clusters are virtually countries and then bring in middle-income econo
always within the same country. mies. However, networks among core countries domi
nate and innovation networks involving only non-core
The figure also depicts the specific subnetworks of two economies are of marginal importance for patents. For
hotspots – Los Angeles and Daejeon in the Republic scientific publication, some middle-income economies,
of Korea – which are of equivalent size in terms of the and even subnetworks among these, are beginning to
number of patents they generate. The Los Angeles play a larger role.
hotspot is highly connected – both nationally and inter
nationally – making it a relatively centered node in the Overall, knowledge production and interactions are
global network. Daejeon is not as central, because it is becoming increasingly global in terms of their reach,
mostly connected only to other Korean agglomerations. through the spread of knowledge-creating hubs and
Daejeon’s international connections are largely limited the formation of international teams. There has been
to Shanghai, San Francisco and New York. some stagnation of co-inventorship networks, reflect
ing a more general slowdown on globalization, but
Geography alone, therefore, does not determine the no let-up in international teamwork for the publish
importance, or the “centrality,” of a top innovation ing of scientific articles. However, as discussed in
agglomeration within a network. Many other factors Chapter 1, truly global innovation networks cannot
also have to be considered. be confined to networks based mostly in a few high-
income countries. Several regions of the world still
have much work to do to integrate themselves into
2.4 Conclusions international networks and, eventually, become part
of global innovation networks. Certainly, international
This chapter exploited a rich data set of patent appli collaboration with top innovation hotspots is a way
cations and scientific publications in order to answer forward. It has worked to some extent for East Asian
several questions arising from two current phenomena economies, notably China.
connected to the way knowledge is produced and
shared worldwide – its geographical spread interna Another important message relates to the geographical
tionally and – at the same time – its concentration in a distribution of knowledge production within countries
few geographical hotspots. (both within well-established knowledge producers
and emerging ones). Despite the increasing world
The production of patents and scientific articles wide spread of knowledge production, there is no
has not stayed within the traditional knowledge- equivalent spreading within countries; there is even
producing economies (Europe, Japan and the U.S.). increased concentration in some. This may have, of
This is a notable development as knowledge-related course, important consequences for the distribution
phenomena, such as patenting, scientific production, of economic benefits within countries, which will need
R&D investment and so on, have always been more to be addressed properly (see Chapter 5).
concentrated than other aspects of globalization, like
trade or FDI. These agglomerations – identified as hotspots and
niche clusters – do not only keep concentrating a
Yet, a few Western economies, plus Japan and the larger share of the production of innovative ideas.
Republic of Korea, account for almost 80 percent of They also increasingly concentrate connections with
internationally-oriented patent activity and around other hotspots, both within their own countries and
57 percent of all scientific publication, which is a across borders through a global innovation network of
lot. In fact, it seems that by far most of the spread in relatively few hotspots. This is bad news for areas of a
knowledge production is due to a handful of develop country that not only produce less innovation, but also
ing, middle-income economies, notably China. At the lack the necessary connectivity to the outside world.
same time, large areas of the world, notably in Africa Lack of connectivity can leave countries or areas locked
and Latin America, are left out of the full process of into non-innovative development paths.
knowledge globalization.
57
Notes
1 This chapter draws on Miguelez 5 Alcácer and Zhao (2016). known technical distortions in
et al. (2019). 6 See a review in Miguelez et relation with overglow, gas flares,
2 The data work relies on the al. (2019). the aurora and zero lights.
research efforts and generosity 7 See Ester et al. (1996). 10 UNCTAD (2005) and Cantwell
conducted by many others. In 8 See Technical Notes for and Janne (1999).
particular, it relies on geocoded further information. 11 For a discussion on this
patent data from Yin and 9 Economists have found that co-inventorship slowdown see
Motohashi (2018), Ikeuchi et nightlight data is a relatively Miguelez et al. (2019).
al. (2017), Li et al. (2014), de good proxy for population and 12 Branstetter et al. (2015).
Rassenfosse et al. (2019), establishment density (see 13 He et al. (2017) and
Morrison et al. (2017) and Mellander et al., 2015), but UNCTAD (2005).
PatentsView (www.patentsview. there are also limitations. There 14 Branstetter et al. (2018).
org, accessed March 2019). is a known weaker link with 15 See Miguelez et al. (2019) for
3 See Miguelez et al. (2019). other economic indicators – for full series.
4 Amendolagine et al. (2019). example, wages – and some 16 See Chaminade et al. (2016).
58
References
Alcácer, J. and M. Zhao (2016). de Rassenfosse, G., J. Kozak and Miguelez, E., J. Raffo, C. Chacua,
Zooming in: a practical manual for F. Seliger (2019). Geocoding of M. Coda-Zabetta, D. Yin, F. Lissoni
identifying geographic clusters. worldwide patent data. papers.ssrn. and G. Tarasconi (2019). Tied In: The
Strategic Management Journal, 37(1), com/sol3/papers.cfm?abstract_ Global Network of Local Innovation.
10–21. doi.org/10.1002/smj.2451 id=3425764 WIPO Working Paper No. 58,
November. Geneva: WIPO.
Amendolagine, V., C. Chaminade, Ester, M., H.-P Kriegel, J., Sander
J. Guimón and R. Rabellotti (2019). and X. Xu (1996). A density-based Morrison, G., M. Riccaboni and F.
Cross-Border Knowledge Flows algorithm for discovering clusters in Pammolli (2017). Disambiguation of
Through R&D FDI: Implications for large spatial databases with noise. patent inventors and assignees using
Low- and Middle-Income Countries. Proceedings of the 2nd International high-resolution geolocation data.
Papers in Innovation Studies No. Conference on Knowledge Discovery Scientific Data, 4. doi.org/10.1038/
2019/09. Lund: CIRCLE, Lund and Data Mining (KDD-96), Portland, sdata.2017.64
University. Oregon, August 2–4, Menlo Park, CA:
AAAI Press, 226–231. UNCTAD (2005). World
Branstetter, L., B. Glennon and investment report 2005:
J.B. Jensen (2018). Knowledge He, S., G. Fallon, Z. Khan, Y.K. Lew, transnational corporations and
Transfer Abroad: The Role of K.H. Kim and P. Wei (2017). Towards the internationalization of R&D –
US Inventors within Global R&D a new wave in internationalization overview. Transnational Corporations,
Networks. Working Paper No. 24453. of innovation? The rise of China‘s 14(3), 101–140.
Cambridge, MA: National Bureau of innovative MNEs, strategic coupling,
Economic Research. and global economic organization. Yin, D. and K. Motohashi (2018).
Canadian Journal of Administrative Inventor Name Disambiguation with
Branstetter, L., G. Li and F. Veloso Sciences, 34(4), 343–355. doi. Gradient Boosting Decision Tree and
(2015). The rise of international org//10.1002/CJAS.1444 Inventor Mobility in China (1985–2016),
co-invention. In Jaffe, A.B. and B.F. RIETI Discussion Paper Series, 18-E-
Jones (eds), The Changing Frontier: Ikeuchi, K., K. Motohashi, R. Tamura 018. www.rieti.go.jp/en/publications/
Rethinking Science and Innovation and N. Tsukada (2017). Measuring summary/18030018.html
Policy. Chicago: University of Chicago Science Intensity of Industry
Press, 135–168. using Linked Dataset of Science,
Technology and Industry. RIETI
Cantwell, J. and O. Janne (1999). Discussion Paper Series, 17-E-
Technological globalisation and 056. www.rieti.go.jp/en/publications/
innovative centres: the role of summary/17030073.html
corporate technological leadership
and locational hierarchy. Research Li, G.-C., R. Lai, A. D’Amour, D.M.
Policy, 28, (Issues 2–3), 119–144. doi. Doolin, Y. Sun, V.I. Torvik and L.
org/10.1016/S0048-7333(98)00118-8 Fleming (2014). Disambiguation and
co-authorship networks of the U.S.
Chaminade, C., C. De Fuentes, G. patent inventor database (1975–2010).
Harirchi and M. Plechero (2016). The Research Policy, 43, 941–955.
geography and structure of global
innovation networks: global scope Mellander C., J. Lobo, K. Stolarick
and regional embeddedness. In: and Z. Matheson (2015). Night-time
Shearmur, R., C. Carrincazeaux light data: a good proxy measure for
and D. Doloreux (eds), Handbook economic activity? PLoS ONE 10(10):
on the Geographies of Innovation. e0139779. doi.org/10.1371/journal.
Cheltenham: Edward Elgar, 370–381. pone.0139779
59
Collaboration between car
manufacturers and tech
companies is beginning
to shift the geography of
innovation in the sector.
Chapter 3
Auto and tech companies – the

drive for autonomous vehicles
In 2004, the United States (U.S.) Defense The automobile industry has been envisioning self-
Department staged a novel off-road race in the driving2 or autonomous vehicles at least since General
Mojave Desert. The novelty lay in it being open Motors presented its “Futurama” concept at the 1939
only to driverless or self-drive cars. First prize World’s Fair. Even in those early days, GM was not
for winning the “Grand Challenge” over the the only one dreaming of a self-driving future, and
240km course was $1 million. Nobody lifted the several attempts toward realization of AVs were made
prize, because nobody finished the race.1 in subsequent years. But it is since the mid-2000s that
huge advances in robotics and, particularly, artificial
But a year later, the Department’s Defense intelligence (AI)3 have begun to turn a long-held aspira
Advanced Research Projects Agency (DARPA) tion into something closer to reality.
staged the competition again and doubled the
prize. It attracted dozens of entrants and this The AV industry is still in its infancy and fully autono
time a number completed the course. The desert mous vehicles (Level 5) are years from reaching the
race was won by “Stanley,” an autonomous market. Nevertheless, robotics and AI are already
vehicle (AV) entered by Stanford University, with reshaping the car industry – so much so that new
vehicles from Carnegie Mellon University (CMU) technologies are posing a significant existential threat
taking second and third places. to the incumbent automakers. AI, data analytics and
a slew of connected devices and components are
reformulating the industry’s business model toward
services and the so-called “platform economy.”
Traditional automakers fear being supplanted and

reduced to bit-players in their core competency – the
making and marketing of cars. To tackle these chal
lenges a menu of options is available to them – from
investing in internal knowledge development, recruiting
human capital and strategic alliances, to acquisitions of
new entrants, or a combination of these.4 It is not clear
which single or combination of the above strategies will
yield the most successful results. What is clear though
is that neither the incumbents nor the new entrants, on
their own, currently have all the required competencies
for producing AVs. They either need to join forces or else
develop internally the respective skills they now lack.
Against this background, this chapter seeks to analyze

current innovation clusters in the automotive industry
and understand how AV is affecting the geographical
spread and concentration of innovation (see Chapter 1).
Understanding the relationship between the new
entrants and the incumbents can offer pointers to the
evolution of current innovation clusters. How firms
react to AV technology will determine which firms will
be the market leaders and which regions will be the
AV technological hubs.
61
In the following sections the chapter looks into the Core aspects of computer-backed driving
contemporary evolution of AV technology and its key
Figure 3.1 Three basic functional components
players. It also briefly discusses two other related tech
of any computer-automated system
nologies: mobility and connectivity. Next, it explores
the impact of AV technology on the automotive industry
from two perspectives. First, whether AV technology Feedback
is changing the nature of innovative collaborations
between and within incumbents and entrants. Second, Monitoring Agency Action
whether it is changing the geography of innovation. It
concludes with a discussion on potential positive and - Sensors - Software - Powertrain
negative impacts. - Driver input - Processors - Brake
- Connected- - Digital maps - Steering
vehicle data
Source: Center for Automotive Research (CAR).

3.1 Definitions
1990s and early 2000s that were the building blocks

Basic components of a driving of modern AV technology.
automation system
There are three basic functional components of any 3.2 Technological evolution
computer-automated system: monitoring, agency, of the automotive industry
and action – as depicted in Figure 3.1. Monitoring
can be understood as sensing and paying attention, Industry evolution literature5 divides the life cycle of
while agency consists of decision-making, and action any given industry into five stages: the introductory
involves implementing decisions. Furthermore, auto embryonic stage, growth, shakeout, maturing and
mated systems can also include various feedback loops, decline. The early stages are ripe with high uncertainty
possibly including machine learning. and numerous entries and exits. Later on, the emer
gence of a dominant design will leave only a handful of
firms standing. Names like Sprite, Unito, Wolfe, Angus,
Levels of driving automation Empire do not exactly ring a bell and that is because
these early car companies were some of the thousands
The established Society of Automotive Engineers (SEA) that exited the industry more than a century ago when
industry standard for terms relating to automated the first automobiles started mesmerizing the world.
vehicles is Taxonomy and Definitions for Terms Related
to Driving Automation Systems for On-Road Motor Until a few years ago, the automotive sector was
Vehicles (SAE J3016). It was initially published in 2014 considered a mature industry with well-established
and substantially revised in 2018. players and for which the key technological questions
had been answered in the 1930s.6 The initial innovations
The SAE standard introduced and defined six levels were fundamental as they defined the basic structure
of driving automation (Figure 3.2), including level 0 for of the automobile. These included the development
systems that perform no sustained dynamic driving of water-cooled engines placed in the front of the
tasks. Levels 1 and 2 are termed driver assistance car, shaft-driven transmissions, streamlined bodies
and partial driving automation, respectively. The lower and pressed steel frames.7 The remaining product
levels of automation require the driver to at least actively and process innovation in the years after the Second
supervise the driving automation system. Driving World War, and particularly after the 1970s, was attrib
automation systems that assume the entire dynamic uted to rising oil prices, cost pressure arising from
driving task are classified as SAE levels 3, 4 and 5 intensifying international competition and changes in
and are collectively described as automated driving consumer demands.
systems (ADS). While the main focus of this chapter is
on level 3+ technologies, in the empirical analysis we At the turn of the millennium this picture changed;
do not exclude the historical innovations of the 1980s, the increasing processing power of computers in
62
3 Auto and tech companies – the drive for autonomous vehicles
From manual to fully automated main ingredients for the realization of AVs are both the
“V” and the “A.” An AV unit is basically chassis and
Figure 3.2 Six levels of driving automation
engine, plus an intelligence that brings full autonomy to
Active the physical aspect. The incumbent automakers’ core
0: No driving automation Safety
Systems competency9 lies with the “V.” Creating all the software
Driving Automation Systems
(e.g., artificial intelligence) and hardware elements (e.g.,

1: Driver assistance
sensors and cameras) required for autonomy – the “A”
2: Partial driving automation – is within the core competencies of the tech companies.
3: Conditional driving automation The incumbent automakers’ core competencies are

Automated
Driving
mass manufacturing, mechanical engineering and
4: High driving automation
Systems jumping through the thousands of regulatory hoops
(ADS)
5: Full driving automation
that lead to the final car being on the road. They are
the result of decades of accumulated tacit knowledge
– knowledge that is not easily replicable – and know-
Source: Center for Automotive Research (CAR) based on SAE 2016.
how. Mastering these competencies is not immediate
and straightforward.
New entrants’ technological competencies are in

conjunction with the widespread adoption of the hardware and software, especially the deep-learning
Internet and, consequently, smartphones, opened and real-time control algorithms needed for vehicle
several avenues for innovation. Many established old- autonomy. They are beyond the spectrum of expertise
line industries – like newspapers, the music business, of most automakers and their suppliers, which have
TV and retail – woke up to the waves of technological little prior knowledge of them.
disruption that advances in software and the hardware
side of computer technology had triggered. These Core competencies of the automakers are more or less
affected not only their core competencies, but also their familiar to most people, but not so the technological
complementary assets – those needed to commer waves that are transforming the industry. The following
cialize and market products – and their distribution sections will briefly discuss three technological waves
channels. Many of these industries were rattled and that are somewhat related. A fourth wave, electric
reshuffled by the digital era. The automotive industry vehicles, although equally affecting the industry, is not
– although with some lag – has not been untouched by within the focus and scope of this chapter.
the waves. For instance, in 2018, the global electric
vehicle fleet exceeded 5.1 million,8 achieving almost
2.1 percent of market share. This number is expected Autonomous vehicles: scientists
to increase to around 30 percent by 2030. behind their contemporary rise
Industry life-cycle literature discusses how industries, The genesis of a set of AV-related startups and tech
as they reach maturity, are subject to new technological firms stems from the Massachusetts Institute of
shocks which can be the seeds for the beginning of a Technology (MIT). MIT has been a global leader in
new cycle. Whether the new cycle is actually realized or robotics technology for decades and has contributed
not depends on the existence of various technological to an agglomeration of firms specializing in AV-related
and non-technological competencies. The participants robotics technology in the Cambridge and Boston area.
in the new cycle may be from within the same industry MIT graduates have also produced several robotics-
or from previously non-competing industries whose related spin-offs, including a few specifically interested
competencies meet the technological requirements in deploying autonomous vehicles.
for entering the new cycle.
In 2007, DARPA held a follow-up competition to its
Competencies required for the development of AVs “Grand Challenge,” this time providing a 60-mile course
have allowed players from the tech industry to enter through a simulated urban traffic environment, includ
the automotive sector, with the ultimate goal of creating ing interaction with other vehicles and compliance with
fully autonomous vehicles that require no driver. The traffic laws. CMU and Stanford again led the pack,
63
Grand challenge scientists and their spin-offs

Figure 3.3 Many leading players in today’s AV industry started in the DARPA grand challenges
Jesse Levinson Zoox co-founder
Kevin Peterson Astrobatic Marble co-founder

Stanford University
Sebastian Thrun Udacity co-founder
Michael Montemerlo Kitty Hawk CEO
Google Inc.
Dimitri Dolgov Pronto co-founder
Waymo LLC
Anthony Levandowski Otto Uber formerly

Carnegie Mellon University
Bryan Salesky Argo AI co-founder
Dave Ferguson Nuro co-founder
Chris Urmson Aurora co-founder
Source: Stanford and Carnegie Mellon University
with CMU’s “Boss” taking first place. In all, six teams

completed the 2007 course – demonstrating the rapid
development of self-driving technology within several Box 3.1 The AV patent mapping strategy and
universities. Silicon Valley tech giants, notably Google, its limitations10
later recruited many Stanford and CMU participants
of the DARPA challenges. (Waymo originated as a The AV industry is a combination of various tech
self-driving project of Google before it became a stand- nologies applied to a specific use – automating the
alone subsidiary.) Most of the scientists involved subse operation of ground-based vehicles. Thus, search
quently have founded their own spin-offs, including strategies to identify AV-related technologies and
such tech startups as Aurora, Udacity, Nuro and Argo scholarship are inherently imprecise and require
AI, all of which are at the forefront of the AV industry creativity and several iterations. Defining clear-cut
(see Figure 3.3). boundaries is very difficult.
The DARPA challenges have been a milestone in the Against these limitations, this chapter makes use
history of modern AV technology. Although there’s of technological codes of the Cooperative Patent
no evidence of their causal effect, we observe an Classification (CPC), an international system for
increasing trend in innovative activity in AV technology classifying patent documents. A list of CPC classes
(measured by patents, see Box 3.1) in the mid-2000s that corresponds to the technologies used in AV
that coincides with the DARPA initiatives, with a major was compiled. The list was divided into two groups.
innovative spike after 2010. Despite this upward trend, First, the smaller number of niche classes where it
AV technology is still very niche and comprised less is relatively safe to say the entirety were relevant
than 0.1 percent of total patent filings globally even at to AV. Second, the classes that were broader and
the height of that spike in 2016 (see Figure 3.4). had patents that may not be relevant to AV. For this
second group, a list of keywords was added to the
search. These keywords were some permutation
64
AV technology has taken off since the mid-2000s

Figure 3.4AV
AVshare
shareof
ofall
allpatent
patentfirst
firstfilings
filingsand
and key
key milestones
milestones over
over time
time
0.100
0.075
Baidu Autonomous Driving Unit
Lyft, Grab and DIDI founded
Percent
Waymo and Uber founded

0.050
2nd DARPA Challenge
Eureka PROMETHEUS Project Navelab 2

DARPA-CMU Navlab 1
0.025
Shakey the robot (Stanford)
0.000
1970 1975 1980 1985 1990 1995 2000 2005 2010 2015
of autonomous vehicle, car, taxi, truck, etc. These Lyft (founded in 2012) in the U.S. came to fruition.
keywords were used to identify the patents that Soon, others with similar business models started
belonged to the selected CPCs and had one of popping up all around the globe: Ola Cabs in India
these keywords mentioned either in their patent (founded in 2010), Grab (founded in 2012) in Singapore
abstract or title. and DiDi Chuxing (founded in 2012) in China. These
companies provided services like ride-hailing and/
The same list of keywords was used to search for or car-sharing. Many of them have expanded their
scientific publications that had mentioned some businesses to other services, including deliveries,
permutation of the keywords in their abstracts logistics and bike-sharing.
or titles. From these selected sets of papers a
new list of keywords was compiled, for example, Uber’s former CEO, Travis Kalanick, described the
predictive cruise control. As publications only have development of “robotaxis” (self-driving taxis) as “exis
broad categories, with no level of granularity similar tential” to the company. If the future of automobiles is
to CPCs, the subject-level category was used to driverless, mobility companies have a vested interest
eliminate those false positive articles that belonged in AV technology for multiple reasons. First, removing
to areas that are intuitively far from AV technology the driver from the equation will reduce their costs.
– microbiology, zoology, etc.
Second, their business model has the potential to
change the economics of the automotive industry.
The MaaS business model can lead to a reduction
of private car ownership and a shift to a more fleet-
oriented system, where the revenue model would be
Mobility as a service based on mileage instead of the number of cars sold.
AV technology can enable a system where people
Parallel to these efforts, Mobility-as-a-Service (MaaS), buy access to transportation as opposed to owning
which integrates various transport services into a vehicles. A rough calculation based on the number
single service available on demand, became a popular of cars on the road and their average annual mileage,
concept. Companies like Uber (founded in 2009) and compared to what mobility companies charge per mile,
65
shows that if all existing cars were to convert to AVs, The rise of AI, robotics and mobility
automakers could make a profit and charge far less services is the main driver of the
than mobility companies. technological shift in the mid-2000s
Figure 3.5 Sectoral breakdown of AV-related
Third, mobility companies are sitting on abundant data
Sectoral by
patents breakdown of AV-related patents
frequency
and information about customer behavior and prefer by frequency
ences, which would give them a significant advantage in 3,500
a sales environment that is increasingly about custom
ized and bespoke experience. 3,000
2,500
Number of patents
Connected vehicles 2,000
1,500
Another branch of technology that has intertwined with
autonomous driving is “connected vehicle technology.” 1,000
A vehicle can be connected without being autono
mous, therefore the two terms are not interchangeable 500
and should not be confused. The connected vehicle

0
technologies allow vehicles to communicate with each 1998 2003 2008 2013 2018
other and the world around them. They aim to increase Auto Tech Other
AUTO TECH OTHER
efficiency and road safety for both drivers and pedes
trians. Popular use cases for connected vehicles are Source: WIPO based on PATSTAT and PCT data (see Technical Notes).
sharing braking data, real-time high-definition maps,
road hazards, closure updates, fleet tracking and
infotainment. All of these require minimum latency applicants, led by names such as Ford (357 patents),
(delay in implementation of commands) and maximum Toyota (320) and Bosch (277) have generated around
precision in the transmission of data. That is why 5G half of the total patents. Non-automakers also feature
cellular network technology is becoming the future in the list of top patent applicants. Google and its
of autonomous and “connected” vehicles.11 Several AV subsidiary Waymo lie in eighth position, with 156
tech companies, notably Huawei, Intel and Ericsson, patents, ahead of automakers like Nissan, BMW and
are exploring this field. Hyundai. They are followed by other companies like
Uber and Delphi, which each have 62 AV patents and
are ranked joint 31st.
3.3 Technological shift
The sectoral breakdown of AV patenting over time 3.4 Competition and

supports the idea that the rise of AI, robotics and mobil cooperation in AV
ity services is the main driver of the technological shift.
In the years immediately after 2005 almost half of the Thus far it is established that the auto sector is in the
patents seem to be from the tech sector.12 However, the early phases of a period of technological disruption,
traditional auto sector later regained dominance (see with several new entrants, both from the auto and the
Figures 3.5 and 3.6). Not surprisingly, the majority of the tech side, joining this bandwagon. Standardization and
patent applicants are companies, roughly 20 percent regulatory issues are not yet being deeply discussed
are individuals and only 10 percent are universities or and there is still no consensus on basic definitions and
other public entities. terminologies. AV technology is an extremely costly
endeavor not only in terms of capital but also time.
A quick look at the list of the top applicants13 in the Therefore, players in this industry have high incentives
1990s shows manufacturing and auto companies. Later to collaborate with each other to share the risks and
lists tell a different story. Google, Qualcomm, Mobileye, costs. But who collaborates with whom? And why?
Uber, Baidu are not among the usual suspects of the Theoretically speaking, three types of collaboration
auto industry, but from the mid-2010s they appear can form: incumbent automakers with each other, tech
in the top 100 AV patent applicants. These top 100 firms with each other, or automakers with tech firms.
66
In the years immediately after enormous costs of designing and building computer-
2005, the tech sector comprised powered vehicles has already prompted Honda to pool
almost half of the patents in AV its efforts with General Motors, while Volkswagen is
pursuing talks with Ford about an alliance on autono
Figure 3.6 Sectoral breakdown of AV-related
mous cars.
Sectoral by
patents breakdown
share of AV-related patents
by share
100
Collaboration among tech companies
80
Tech firms also would need to collaborate with each
other to share the technology’s large risks and costs.
60
Most tech firms, especially the smaller startups, occupy
Percent
niches, focusing on hardware, software, mobility services,

40 connectivity, communications and many more (see
Figure 3.7 below). With the exception of Waymo –
20
which develops all its hardware and software stack16
in-house – no single tech company has the necessary
expertise in all these areas. So, collaboration among
0
1998 2003 2008 2013 tech companies is not uncommon. Taiwan-based VIA
2018
Auto Tech Other

Technologies Inc. announced in 2018 that it is partner
AUTO TECH OTHER
ing with AI vision startup Lucid to deliver AI-based
Source: WIPO based on PATSTAT and PCT data (see Technical Notes). depth sensing in dual- and multi-camera devices
for use in security, retail, robotics and autonomous
vehicles.17 This is just one of a long list of examples of
Collaboration among auto companies collaboration between tech companies.
In the face of the AV technological shock, auto compa Some tech companies have also decided to give open
nies have an incentive to join forces to share the costs access – free of cost or other access barriers – to their
and risks but also defend their market position, which closely guarded data and technologies. For instance,
is being threatened by outsiders. The common threat Waymo has decide to sell one of its three LIDAR
they are facing is “commoditization” of their core sensors – called Laser Bear Honeycomb, which uses a
competency; that is, becoming simply a supplier of laser to measure distances – to third parties interested
a commodity good, which in this case is a car. The in using the technology for purposes other than self-
tech companies would be the ones generating the driving cars. Some believe the LIDAR sensor devel
value added and therefore reaping the largest benefits. opment curve is similar to Moore’s Law in computer
Global automakers Daimler and BMW announced chips – every 18 months, resolution will double and the
they would partner in a new long-term partnership to price drop by half18 – so granting open access offers
co-develop automated driving technologies. the chance to scale up with reduced costs.
The joint effort will involve 1,200 technicians from both Waymo is making some of the high-resolution sensor
companies. The technicians will be based at BMW’s data gathered by its fleet of autonomous vehicles avail
autonomous driving campus in Unterschleissheim, near able to researchers for free. It is not the first company
Munich, its Mercedes subsidiary’s technology center to release an open dataset. In March 2019, global
in Sindelfingen, near Stuttgart, and Daimler’s testing technology company Aptiv was one of the first large
and technology center in Immendingen in southern AV operators to publicly release a set of its sensor data.
Germany. The two companies aim to launch their next- Uber and Cruise, the autonomous division of General
generation, self-driving passenger cars by 2024.14 Audi, Motors, have also released their AV visualization tools
another German automaker, has announced that it is to the public.19
to join forces with them.15
These decisions are in line with the “open innovation”20
While some may be surprised to see long-time foes strategies that firms adopt as a response to highly
becoming friends, it is not rare in AV development. The complex innovative ideas.
67
Mapping the involvement of AV companies

Figure 3.7 Examples of companies working in various AV technologies
Automated driving systems Automated vehicle service applications
Perception/ Cognition/ Advanced Mobile

Actuation Freight and
sensors core intelligence driver-assistance Ridehailing industrial
delivery
systems (ADAS) robots
Google Google Tier 1s
Velodyne Cruise Automation Tier 1s Uber, Lyft Daimler Embark
Nvidia Google Volvo Trucks Starsky
Cruise Peleton Robotics
Wireless Localization and Intelligent Fleet

communication navigation infrastructure management
Telecoms DeepMap G4 Apps eSentire

Qualcomm TomTom Mojio ISARA
Nokia
Vehicle design Vehicle design Simulation Validation Certification
OEMs OEMs Google IAV Intertek

Startups Tier 1s Nvidia Veoneer Emtek
Startups Baidu Zenuity
Source: Center for Automotive Research (CAR).
Collaboration between tech While it seems logical for auto companies to collaborate
and auto companies with tech companies, the reverse is not so straight
forward. Some might even argue that tech giants do
AV technology is not rendering the upstream core not need auto companies and that they can, and will
knowledge of automakers completely obsolete. In eventually, directly enter the auto sector.24 Their argu
fact – at least for now – AV is a type of technologi ment focuses on the costs. Since IT giants like Alphabet,
cal discontinuity that needs the incumbent’s core Amazon and Apple in the U.S. and Alibaba, Baidu and
competency to achieve its goal. Research shows21 that Tencent in China have deep pockets they can easily
– historically – incumbents can survive the discontinuity afford the costs of designing and manufacturing a car.
if they cooperate with the entrants challenging their Others do not agree.25 Excelling at complex mass manu
core knowledge. In presence of strong “appropri facturing, organizing quality value chains, dealing with
ability regimes,” the new entrants have the incentive complex regulatory issues is neither trivial nor negligible.
to license out their technologies. The literature 22 U.S. energy and automotive company Tesla’s financial
defines strong appropriability regimes as environ losses and struggles to keep up with delivery schedules
mental factors – legal protection (e.g., patents) or of its Model 3 electric sedan car attest to this issue. The
the needed knowledge is difficult to pass on (tacit) or ecosystem in which automakers operate and lobby is
codified – that allow the tech company to recuperate their stronghold. Even if the tech companies had the tech
its investment. nological capacity to produce cars, they would still have
difficulties challenging the current socio-technical regime
AV technology shows characteristics of strong appro unless they collaborate with the incumbent automakers.
priability. This allows the new entrants to cooperate
with incumbents while securing their benefits without Therefore, tech companies also have an incentive to
fear of imitation.23 By partnering with tech companies, collaborate and see where their strengths complement
automakers gain a better understanding of the key those of the automakers. This division of labor, at least
technologies that are transforming the industry and at this stage of the industry, allows each side to focus
accelerate the learning process that can keep them on what they do best and is the shortest and safest
competitive in a rapidly changing environment. route to AV success.
68
The types of collaboration outlined are not mutually traditional automobile and mainly related to level 1 or 2
exclusive and they coexist. The high uncertainty makes of driving automation (See Figure 3.2).
firms simultaneously bet on multiple combinations of
the three options – “build,” “borrow” and “buy.”26 In the later years, we observe some developing coun
tries that are not traditional automaking countries
By default, much of the above collaboration may not be also engaging in this technology. The most notice
captured by patent or scientific publication data. The able change is the emergence of China and India. As
main reason is that most are formal partnerships and discussed earlier, the changing nature of technology
alliances, joint ventures, investments or acquisitions. can be one explanation of this expansion. The new sets
Out of more than 100 formal collaborations identified,27 of technologies – AI and robotics – allow for “leapfrog
in terms of frequency, the largest share belongs to ging” of countries/regions with no longstanding ties to
auto–tech, followed by tech–tech and auto–auto. Finally, the auto-manufacturing sector.30 Despite this, the top
a small portion of the collaboration is between tech countries involved are still the U.S., Japan, Germany,
companies and national or regional government enti the Republic of Korea and Sweden, with the U.S. and
ties. For instance, Detroit-based Quadrobot and the China latterly being the most active.
Chinese Postal Service are partnering to produce
autonomous delivery vans. When looking at scientific publication we observe that
more developing countries in the Middle East, Latin
America and Africa – that are not captured in the patent
3.5 Role of geography ing data – are highly active in generating basic research
in AV technology and scientific articles. Iran would be an example of a
country highly active in scientific publication but with
almost no patenting presence in this field. Scientific
Spread over time publication data complements patents in giving a better
picture of the innovation landscape in AV technology.
Until a few years ago, no one would have associat
ed places like Boston, San Francisco and Pittsburg,
Singapore or Jerusalem with the automotive industry. 3.6 AV innovation, countries
The more familiar names were Detroit, Toyota City and cities
in Japan and Stuttgart in Germany. But advances in
robotics and AI as general-purpose technologies,28 with
multi-faceted applications in various fields, have created North America
avenues for new entrants. Naturally, these entrants reside
in the main tech hubs, such as the U.S. Silicon Valley and Boston, Massachusetts
others around the world. However, places like Singapore
or Jerusalem, with no history in the automotive sector Boston is not a traditional automotive industry
but with booming and vibrant tech and startup scenes, cluster. However, the Massachusetts Institute of
have become highly active in AV technology. Technology (MIT) has been a global leader in robot
ics technology for decades and has contributed to
A historical look at innovative activity in AV shows its an agglomeration of firms specializing in AV-related
geographical evolution and global spread. Figure 3.8 robotics technology. One major company that has
displays the regions involved in patenting29 and publish taken advantage of the Boston robotics cluster for AV
ing scientific articles concerning AV-related technolo development is the Toyota Research Institute (TRI),
gies, before and after 2005. Not surprisingly, in the earli which located one of its three offices in Cambridge
er period, regions that traditionally led the auto market (the other two offices are in Michigan and California).
also show high patenting activity. But even then, there TRI sponsors MIT’s Computer Science and Artificial
was significant patenting activity from Silicon Valley and Intelligence Laboratory (CSAIL), where researchers
Singapore. The focus in the earlier period was still on study various aspects of AI and machine learning
areas like advanced driver assistance systems (ADAS) applied to vehicle automation.31
and automated highway systems (AHS), technologies
that are not directly related to AI/robotics approaches. MIT has produced several robotics-related spinoffs,
These patents were closer to the operations of the including a few specifically interested in deploying
69
East Asia has become very active in AV technology in recent years

Figure 3.8 Geographical distribution of AV-related patents (this page) and
publication (next page) in selected regions, pre- (left) and post-2005 (right)
North America
Europe and the Middle East
East Asia
PATENTS SCIENTIFIC PUBLICATIONS
70
North America
Europe and the Middle East
East Asia
71
autonomous vehicles. One of these, nuTonomy, was Ontario, Canada

purchased in 2017 by Aptiv – a global tier-132 automotive
supplier historically tied to Detroit and General Motors.33 The interest in automated vehicle systems has brought
Aptiv maintains a technology center in Boston, along additional attention to centers of software development
with centers in Pittsburgh and California.34 nuTonomy and artificial intelligence. One research cluster that has
is running trials in Boston and Singapore, where the benefited is the Canadian province of Ontario, including
state Economic Development Board has taken a stake Toronto, Waterloo and Ottawa.
in the company.35 Another MIT spinoff, Optimus Ride,
has partnered with multiple Silicon Valley and automo Ontario is an established automotive industry cluster,
tive firms to deploy low-speed, self-driving shuttles in owing mainly to its proximity to Detroit. Ontario is
defined geo-fenced routes.36 also strong in the computer software industry. The
University of Waterloo, for example, has outstanding
math and computer programs. The Waterloo Centre
Detroit, Michigan for Automotive Research (WatCAR) has several distinct
groups researching advanced vehicle and mobil
Detroit is the historical center of the North American ity technology.41 The University of Toronto also has
automotive industry. General Motors and Ford main programs focused on vehicle automation, connectivity
tain headquarters and multiple research centers in and cyber-security.42
the Detroit metropolitan area, as do several interna
tional automakers (Fiat Chrysler Automobiles (FCA),
Hyundai/Kia and Toyota) and dozens of large automo Pittsburgh, Pennsylvania
tive suppliers. Nearly all automakers involved in the
North American market have some presence in the Carnegie Mellon University (CMU) in Pittsburg has
Detroit area. been a center of autonomous driving technology for
decades. CMU researchers tested the very first proto
Michigan is not a historical locus for robotics. However, type on-road hands-free driving automation system in
it is among the top areas worldwide for research, 1986, the Navlab 1 project,43 which was followed by
development, design and manufacturing of advanced Navlab 2 in 1990. CMU teams also were among the
automotive systems. While Detroit-based firms have most successful in the DARPA grand challenges that
opened regional offices in robotics hubs, such as helped usher in the current era of AV research.44 To
Pittsburgh, Boston and Silicon Valley, AV-focused some extent, the CMU robotics program has been
startups have opened offices near Detroit to lever a victim of its success, as dozens of established
age the local expertise in engineering and validating researchers have been hired away by AV startups.
robust automotive-grade systems. The automotive The most notable example of this is Uber, which first
focus of the technology is also leading to increased began a strategic partnership with CMU and opened
investment in software development facilities in the a nearby research center. However, eventually, it hired
Detroit area, including significant investments by Ford, over 50 CMU researchers away from the university.45
GM and Toyota.
CMU has also spawned some AV startups, such as
Waymo – perhaps the most advanced autonomous Argo.AI, of which Ford has purchased an ownership
vehicle developer in the industry – plans to renovate stake and announced it will deploy a robotaxi service
a historic Detroit facility to fit vehicles with its propri in 2021.46 Boston-based nuTonomy, now owned by
etary automotive technology.37 Waymo is partnering tier-1 auto supplier Aptiv, has facilities in Pittsburgh
with Magna International – a tier-1 automotive supplier and is actively expanding. Many other CMU robotics
based in Aurora, Canada, with multiple facilities in the alumni are now scattered throughout the AV diaspora,
Detroit area.38 Previous to the partnership with Magna, including some of the most significant names in the
Waymo contracted with another major Detroit-area industry, as shown earlier. CMU’s Robotics Institute
engineering firm – Roush.39 Roush, meanwhile, has carries on, though with less emphasis on driving
expanded its involvement in automated-vehicle engi automation than in previous decades.47 Meanwhile,
neering, opening a new research center focused on AV Pittsburgh has become one of the world’s most popular
software and systems integration.40 cities for on-road testing and development of prototype
autonomous vehicles.48
72
Silicon Valley, California out an embedded control system, AliOS, and a smart
assistant called Tmall Genie. Tencent, another tech
It now feels as though Silicon Valley (the area surround giant, has its own system called “AI in Car.”
ing San Francisco, California), has always been the
center of the AV industry. However, it began with Moreover, the Government’s New Generation Artificial
Google (now Waymo) taking an interest in on-road Intelligence Development Plan,51 announced in 2017,
autonomous vehicles following the DARPA grand shows China’s determination to become a global leader
challenges. Google started to hire grand challenge in artificial intelligence, including autonomous driving
participants in 2009, including CMU team leader Chris technology. China also heavily invests in infrastructure
Urmson, who became chief technology officer (CTO) and building roads and streets that are compatible
of the project. Other researchers were available locally. with connected and autonomous vehicles. Roads in
Stanford University had an established robotics and Beijing’s E-Town52 are among 44 roads (total 123 km)
automated-driving research program on the same marked out for testing AVs. Besides Beijing, extensive
level as CMU and otherwise unparalleled in the world. testing is also taking place in 15 other cities around
China, including Shanghai, Shenzhen and Guangzhou
Google announced its self-driving project in 2010 with in Guangdong province, Hangzhou in Zhejiang province,
a compelling video of a blind man taking a self-driving Wuhan in Hubei province and Chongqing.53
car to a Taco Bell restaurant. The video did not convey
the extent of preparation required for the demonstration, Baidu, Pony.ai and WeRide are the leaders in self‐
but it did show a level of driving automation capability driving technology in China.54 However, even Baidu
that surprised the automotive industry and sparked a is not yet considered to be among the top 10 globally.
movement of industry stalwarts and startups to catch In California, Baidu’s test vehicles required human
up and get in on the self-driving action. intervention every 41 miles driven, compared with
every 5,596 miles driven by Waymo’s vehicles.55 The
Google’s self-driving car project, combined with the Baidu Apollo Automated Vehicle Platform has, however,
pre-existing pool of AI and software engineers, cata attracted more than 100 global partners. Apollo has
lyzed the development of Silicon Valley as a global an AV simulation system, vehicle test data and high-
leader in AV development. definition maps.56 Both CarLife and DuerOS are incor
porated in Apollo. Further, Baidu has pledged to deploy
It would be difficult to tabulate the number of firms self‐driving taxis in geo-fenced areas of Changsha,
pursuing automated driving in the Valley. But as of Hunan Province, in 2019.57 Baidu is attracting the major
early 2019, 62 entities have received a permit from ity of investment and attention in the Chinese market
the California Department of Motor Vehicles to test for autonomous driving. However, the city govern
prototype ADS on public roads in the state.49 ment of Beijing, which has begun requiring reporting
for testing of automated vehicles on city streets, has
received reports from seven other companies in addi
China tion to Baidu.58 Many Chinese companies also maintain
research facilities in Silicon Valley, including Baidu.59
The three waves of technological disruption discussed Chinese companies, including Baidu, NIO, Tencent,
earlier (AV, MaaS and connected cars), have created a Alibaba, FAW, SAIC, ChangAn, BAIC, Great Wall, GAC,
window of opportunity for Chinese auto firms as they Dongfeng, Geely, BYD and Lifan have started testing
have no legacy disadvantage vis-à-vis foreign multi their vehicles in China. Waymo has also opened a
nationals.50 Even in China, however, local tech giants subsidiary in Shanghai, although the filing says the
have the upper hand over the auto sector. Chinese subsidiary will focus on logistics consulting, supply
tech giants, such as search engine company Baidu, chain and AV parts and product design and not AVs.60
e-commerce concern Alibaba and ride-hailing firms
Didi Chuxing, Dida and Ucar are more or less on par
with their foreign counterparts. With regard to connec Japan
tivity, Baidu’s CarLife – a system that allows mobile
phones to control the infotainment in a car – has been Japan’s AV work was slow to start due to particularly
up and running since 2015. Baidu’s voice assistant restrictive laws on self-driving. However, with the 2020
technology is called DuerOS. Alibaba has also rolled Olympics approaching, there has been an explosion
73
in the AV industry in order to show off the country’s eye toward automation.62 The U.K. has also published
cutting-edge technology. Japan has introduced legis an advisory document to guide testing automated-
lation to ease restrictions on self-driving cars and it driving technologies on public roads.63
intends to use Toyota autonomous self-driving shuttles
at the 2020 Olympics to transfer competitors around The U.K. has also benefited from EU‐funded research
the athletes’ village. Evidently, Toyota has become a programs, such as the EU GATEway project. That
large part of Japan’s foray into AV tech, but Toyota and project funded U.K.‐based Oxbotica – an Oxford
other Japanese companies are doing much more than University spin‐off – to deploy a low‐speed autono
just the Olympics push. mous shuttle on a mixed‐use pathway.64 Another U.K.
project supported the development of an automated
Within Japan, Toyota has partnered with Japanese podcar,65 which operates on purpose-built guideways,
tech investment firm Softbank, most known for its to be developed by the U.K.‐based RDM group, an
USD 100 billion Vision Fund to buy stakes in fast-grow automotive supplier. That project resulted in RDM
ing technology companies, to create a joint venture – spinning‐off an independent company, Aurrigo, which
MONET – that will focus on development of driverless now has facilities in the U.S., Canada and Australia.66
technology and MaaS solutions in Tokyo. Additionally,
MONET has investment from Japanese automakers Cambridge is a global center of innovation for AI –
Honda and Hino. Waymo has partnered with Renault dating back to 1936 when Alan Turing invented the
and Nissan, a Franco-Japanese alliance, to bring its “universal computing machine” at Kings College.67
AV mobility services to France and Japan. Also, ZMP, Cambridge is also the home of ARM – a global leader
a Japanese AV company, and Hinomaru Kotsu, a in high‐performance processors – which has taken an
Japanese taxi company, have paired to develop an interest in automated driving.68
autonomous taxi, which they are hoping will be ready
for the 2020 Olympics. However, the Japanese AV Other U.K. universities are heavily invested in devel
space isn’t just populated by private companies. Both oping an AV cluster. The universities of Warwick,
the University of Tokyo and Keio University have smart/ Birmingham and others contribute to the AI talent
advanced mobility projects developing AV technology. pipeline supporting clusters. Oxford University boasts
an exceptionally strong robotics program and, as previ
Internationally, the main player is Toyota, again. It ously mentioned, gave birth to Oxbotica.
has partnered and invested internationally for AV
advancements with a myriad of companies, includ
ing Uber, May Mobility, Hui, Grab, Getaround, Nvidia France
and AT&T. Furthermore, on the international front,
Softbank has invested USD 2.25 billion in GM’s Cruise France’s automobile industry is doing its part to
Automation, a robotaxi firm, and was involved in provid remain engaged in the development of next‐genera
ing USD 1 billion in funding for Uber to use on its AV divi tion automated vehicles. Renault has pledged “eyes‐
sion. Chinese company SenseTime is one of the world’s off/hands‐off” functionality in production vehicles
highest-valued AI startups, and they have opened a as soon as 2021.69 The Groupe PSA (whose brands
self-driving facility in Joso, just outside of Tokyo. include Peugeot, Citroen and DS) is pursuing its
Autonomous Vehicle for All (AVA) program.70 PSA
is testing AV technology on roads in Europe and
United Kingdom (U.K.) China.71 The global automotive tier‐1 supplier Valeo
is also investing heavily in driving automation.72 Valeo
The U.K. is an established hub of automotive and is building a research center for AI in Paris and has
engineering talent. The Government has taken a keen secured multiple research partnerships. Such efforts
interest in autonomous vehicles and has worked to have been buoyed by a national effort to make France
leverage existing capabilities to remain a significant an AI leader.73
contributor to an emerging AV industry. For example,
the U.K. Autodrive project funded trials of prototype Europe has seen dozens of low‐speed autonomous
automated vehicles made by several manufacturers.61 shuttle trials deployed by several firms. France is a
The UK CITE Consortium is an industry‐led group center of R&D for autonomous shuttles. One of the larg
focused on connected vehicle technology but with an est and best‐known companies, Navya, was founded
74
in France in 2014. It has pilot deployments of shuttles level 3 automation in 202085 and BMW is targeting
around the globe and has a facility in Michigan. Well deployment of consumer‐available autonomy in 2021.86
over 100 shuttles have been produced.74 Keolis, a
France‐based private operator of public transit systems, German tier‐1 suppliers are also very active in this
operates many of these deployments.75 space. Tire company Continental has long manufac
tured components of vehicle automation and has even
Another of the world’s largest autonomous shuttle constructed its own autonomous shuttle.87 Conti also
companies is EasyMile. It was founded in Toulouse in aspires to provide an ADS platform as a supplier to
2014 following the EU‐funded CityMobile2 Project. Over automakers.88 German firm ZF has been partnering
100 EasyMile shuttles have been produced and used in and maneuvering for years to become integrated into
test deployments around the globe.76 TransDev, another a global supply chain for AVs and is also developing a
France‐based private operator of public transit systems, prototype vehicle.89 Bosch is another major tier‐1 with
manages many of these deployments.77 TransDev has ambitions to provide AV technology and is working
also partnered with U.S.‐based Torc Robotics to test with Daimler among others to deploy the technology in
autonomous shuttles in France.78 future consumer vehicles.90 The maturity of this cluster
has supported dozens of autonomy and mobility-
Many French companies are making efforts to expand related startups.91
in the North American market.79 While many of the
French AV‐focused companies remain small, many are
actively partnering with other companies and institu Israel
tions, which demonstrates their global ambitions.
The agglomeration of technology companies in this
small country is remarkable. By one count, as of
Germany mid‐2018, nearly 1,000 Israeli startups were using or
developing AI technology, and well over a dozen new
Germany may be second only to the U.S. as a well firms were being established every month.92
spring of innovation and development of AV technology.
The EU‐sponsored PROMETHEUS research program in Global AI and software firms have maintained facili
the 1980s paralleled DARPA‐sponsored research and ties in Israel for some time to take advantage of this
established German institutions, such as Universität der ecosystem, and the auto industry has followed. For
Bundeswehr München (UBM), as sources of AI and AV example, General Motors was once notable for having
expertise.80 The very first consumer-available level 2 no significant presence in Silicon Valley (this is no longer
driving automation system, which provides such things the case after GM acquired Cruise Automation), but it
as steering, braking and acceleration support to the established a research center for automated vehicle
driver, was introduced by Mercedes Benz and was a technology in Israel in 2008 and expanded it in 2016.93
legacy of the PROMETHEUS program.81 Several other automakers have expanded or opened
research centers in Israel since 2016.94
The German auto industry, including Daimler, BMW
and Volkswagen, has created numerous partnerships Perhaps the best‐known Israeli firm contributing to the
inside and outside of Germany in efforts to bring about global AV ecosystem is Mobileye – a supplier of vision
a new era of autonomous shared mobility. These systems for multiple automakers. Mobileye started in
activities include not only minor research partner 1999. It had its initial public offering (IPO) in 2014 and
ships and investment tie‐ins, but also large consor was acquired by Intel in 2017 for USD 15 billion. As an
tia.82 The German automakers have been among the established supplier, Mobileye claims that its technol
most aggressive in communicating goals for public ogy has been embedded in over 27 million vehicles
deployment of automated driving. VW’s Audi brand across 25 different brands.95 Mobileye is now the face
announced that the 2018 Audi A8 would have an of Intel’s foray into the automotive supply chain and
option for the world’s first consumer‐available level 3 is pursuing fully autonomous driving in earnest. Intel
ADS, called “Traffic Jam Pilot,”83 which would allow for has announced a partnership with Israel’s Champion
highly automated driving. But it later cited regulatory Motors and Volkswagen to deploy driverless taxis in
barriers as delaying its appearance.84 Mercedes has Israel with commercialization scheduled for 2022.96
announced its flagship S‐Class sedan will include Beyond its supplier role, Intel/Mobileye’s activity in
75
strategic R&D partnering has become increasingly have 16.1 and 9.7 percent of AV patents but only 1 and
extensive and global.97 4.8 percent, respectively, of general patents.99
In order to test whether tech companies have moved

3.7 Is AV technology changing physically closer to automakers, the same exercise
the geography of innovation in was repeated100. The selected companies were
the automotive industry? Google, Waymo, Delphi, Mobileye, DeepMap, Magna
Electronics, Qualcomm, Uber and Apple. No systematic
Innovation has a geographical dimension.98 Research trend toward auto clusters was observed. As with auto
has shown that industries tend to co-locate in the makers, the lion’s share of both total and AV patenting
vicinity of each other (see Chapters 1 and 2). The two happens in the same top cluster.
types of players in the auto industry, the incumbents
and the new entrants, have their own geographical The geography of Uber’s AV patents is interesting.
clusters. The new entrants belong to the tech clusters While 39.6 percent of its patents are in San Francisco,
of the world (e.g., Silicon Valley), whereas the incumbent Silicon Valley is not its top cluster when it comes to
automakers are well established in their manufacturing AV. Around 48.5 percent of Uber’s AV patents are in
clusters (e.g., Detroit). The key question is whether the Pittsburgh, where it has been hiring and collaborating
emergence of AV has made the automakers and tech with CMU researchers. Uber has also been testing AVs
companies seek greater geographical proximity. If the in Pittsburg since late 2018.
answer is yes, in which direction? The automakers are
appearing in the tech clusters or vice versa. These results indicate that, while there is some shifting
geography at the margin, auto and tech companies’
While it is too early to give a definitive answer to the innovation is still largely home based. However, the
above questions, evidence based on patent data evidence available, although interesting, should be
can shed some light. This section looks at the top treated with caution. The numbers, particularly for AV
global auto industry companies’ patents, selected patents, are very limited and the weight of this limited
from three geographical areas: the U.S. (Ford and set of patents may distort the overall picture. Moreover,
GM), Germany (Daimler, BMW, Audi, Volkswagen and patent data is made public with at least 18 months’
Bosch) and Japan (Toyota, Honda and Nissan). These delay after being first filed. And the actual innovation
companies’ total patent portfolio was examined, and may have been developed months, if not years, before
a subset of patents related to AV technology identified the patent request was made. Finally, applicants’ name
and flagged. Based on this data the share of each disambiguation issues may have impacted the results
company’s total patenting for different clusters is for some companies.
calculated together with that of AV patents. For instance,
72.6 percent of Daimler’s total patents are in Stuttgart,
with 76.9 percent of its AV patents also being there. 3.8 Potential positive and
negative impacts of AVs
The major chunk of automakers’ AV patents is still
generated in the same main clusters where most Despite the high anticipation that surrounds them,
of their patenting happens. Nevertheless, there are fully autonomous vehicles are, if not decades, defi
also important variations. More than 82 percent of nitely years away.101 Multiple intertwined technological
Japanese automakers’ total and AV patents belong advances are creating new rules for an industry that
to their primary, Japan-based clusters, a far higher had not changed its way of doing business for almost
percentage than that of the two U.S. companies, as a century. Key players from the tech and traditional
can be seen from Table 3.1 below. automobile sectors – although with different incentives
– are pooling resources to realize the goal of self-driving
A quick look at the list below of second-line clusters cars. However, the obstacles are not simply technical.
reveals some interesting differences. A number of Every technological shock at the early stages faces
clusters, such as San Jose, Berlin, Los Angeles and some level of socio-technical inertia in the sense that
Osaka, have strong AV specialization (in the sense that new technology requires organizational changes that
their AV share is large relative to their total patent share). also affect the interaction of people and technology.
For Volkswagen, for example, San Jose and Berlin each Oftentimes, change is not easily welcomed.
76
While there is some shifting geography at the margin, auto and

tech companies’ innovation is still largely home-based
Table 3.1 Comparison of the total share of patents with the AV patents of selected automakers in
different clusters
Cluster name Total share (%) AV share (%) Cluster name Total share (%) AV share (%)
Audi GM
Ingolstadt 60.1 60 Detroit–Ann Arbor 45.3 54.7
Munich 10.7 18.8 Waterford 5.1 11.3
Frankfurt 3.9 6.2 Los Angeles 4.5 8.5
San Jose–San Francisco 0.4 6.2 Frankfurt 16.6 7.5
BMW Honda
Munich 72.5 84.1 Tokyo 90.8 82.3
Nürnberg 1.3 6.1 Los Angeles 0.2 3.7
Würzburg 0.4 3.7 Osaka 2.6 2.4
San Jose–San Francisco 0.4 3.7 Nagoya 3.1 1.8
Bosch Nissan
Stuttgart 69.1 77.6 Tokyo 97.0 87.7
Munich 2.6 5.0 Osaka 1.5 8.6
San Jose–San Francisco 1.0 4.6 San Jose–San Francisco 0.0 3.1
Braunschweig 0.5 4.1 Nagoya 1.2 2.5
Daimler Toyota
Stuttgart 72.6 76.9 Nagoya 95.4 93.7
Ulm 5.8 7.4 Tokyo 5.4 5.2
Frankfurt 5.1 4.1 Osaka 2.3 3.0
Aachen 0.7 4.1 Shizuoka 0.2 1.1
Ford Volkswagen
Detroit-Ann Arbor 65.0 71.5 Wolfsburg 47.9 46.8
Cologne–Dusseldorf 8.8 6.6 Braunschweig 37.1 40.3
San Jose–San Francisco 1.4 3.6 San Jose–San Francisco 1.0 16.1
Aachen 4.8 2.9 Berlin 4.8 9.7
Note: The sum of the percentages may be more than 100 percent, due to the fact that a single patent can be assigned to more than one cluster so
there is double counting.
The current ecosystem of the automotive industry – Advocates of AV technology see it solving several
its market power and its social and political position, grave urban problems. For example, it could reduce
for example – has been in place for decades and is traffic jams and air pollution and improve road safety.
very strong. This ecosystem is not so likely to change Increased precision in the movement of vehicles and
easily unless the key players in the industry change the elimination of human error can reduce traffic fatali
(i.e. existing automakers exit the market or the market ties. Connected “smart” vehicles can safely travel much
is totally taken over by the tech companies), there is a closer together – a technique known as “platooning.”
drastic transformation of policy and regulatory issues or This, together with automated highway systems, should
customer demand and preferences shift considerably. increase road capacity and lead to other efficiency
At the same time, public opinion is still split over AV. gains, such as lower fuel consumption and better
77
energy efficiency, which will also have a positive impact system” applications could potentially be used for
on the environment. purposes not related to driving. The ability of hack
ers to crack the system, and alter information or the
Hours would no longer be wasted “behind the wheel” identity of another vehicle is one of the many serious
and those who would once have been driving could security worries. Legal and regulatory systems already
instead dedicate time to relaxing, working or even sleep have trouble keeping up with the fast pace of change
ing. Children, senior citizens and disabled people would in the automotive industry. It is still not clear, in the
have more independence and mobility. Land currently case of an accident, who would be legally liable – the
devoted to parking lots could be put to other uses. company that runs the software system, the hardware
or the mobility platform.
Not everyone is so positive about self-drive cars,
however. In 2018, the death of a pedestrian in Arizona Moreover, countries and regions are at different levels
in an accident involving a test vehicle operating in self- of infrastructure readiness for AVs. Uneven degrees
driving mode was a huge setback. Some companies of preparedness may exacerbate inequality between
temporarily halted road testing. Whatever the state of richer and poorer areas within countries and between
play technologically, the general public may not yet be regions. All these changes will ripple through other
ready for AVs to go mainstream. Some critics question industries – from insurance to repair, trucking to taxi
whether AVs would really help solve urban issues such driving. AV technology has an impact that goes beyond
as traffic jams and pollution. The new technology could the boundaries of a single industry.
simply increase the number of vehicles on the road, and
therefore congestion. And with cars being self-driving, Until the auto and tech world can address all these
commuters might be prepared to “drive” further to work technical, ethical, security and legal issues, the AV
rather than take a train, which is less polluting. future will continue to be a dream.
Privacy and cyber-security are also major concerns.

Data about drivers collected through autonomous,
connected vehicles and other “intelligent transport
78
Notes
1 This section draws on Dziczek et 18 See Randall (2019). Research Projects Agency (n.d.).
al. (2019). 19 See Hawkins (2019). 45 See Lowensohn (2015).
2 In this chapter, terms like 20 See Chesbrough (2003) 46 See Vasilash (2018).
autonomous vehicle, self- 21 See Arora and 47 See Carnegie Mellon
driving, driverless, etc. are used Gambardella (1990). University (n.d.).
interchangeably and are meant 22 See Teece (1986). 48 Wiggers (2019).
to refer to the same phenomenon. 23 See Gans and Stern (2003) and 49 See California Department of
3 See WIPO Technology Trends Cozzolino and Rothaermel (2018). Motor Vehicles (n.d.).
2019 - Artificial Intelligence 24 See Perkins and Murmann (2018). 50 See Teece (2019).
4 See Tripsas (1997) on internal 25 See MacDuffie (2018), Jiang and 51 See full translation: FLIA (2017).
knowledge development Lu (2018), Teece (2018). 52 Economist (2019).
strategies, Zucker and Darby 26 See Capron and Mitchell (2012). 53 See Feifei (2019).
(1997) on recruiting human 27 The majority of the data was 54 See Silver (2018).
capital, Rothaermel (2001) on collected from the latest media 55 See Teece (2019) and Jing (2018).
strategic alliances, Higgins and and company announcements. 56 Visit apollo.auto.
Rodriguez (2006) on acquisitions However, at times this info may be 57 See Xinhua (2019).
of new entrants, and Rothaermel misleading as other motivations 58 See Liao (2019).
and Hess (2007) for combinations like market signaling and gaining 59 Visit research.baidu.com.
of these strategies. venture capitalist attention might 60 See Korosec (2018).
5 See Klepper (1997), Audrestsch be behind the announcements. 61 Visit www.ukautodrive.com/
and Feldman (1996), Abernathy 28 See Bresnahan and the-uk-autodrive-project.
and Utterback (1978), Jovanovic Tratjenberg (1995). 62 See Fleet News (2018).
and MacDonald (1994). 29 The patent and scientific 63 See U.K. Department for
6 See Abernathy and Clark (1985) publication data used in this Transport (2015).
and Klepper (1997). section are a sub-sample of 64 See Dennis and Brugeman (2019).
7 See Klepper (1997). those explained in Chapter 2. 65 Personal Rapid Transit (PRT),
8 See IEA (2019). For more information about also referred to as podcars, is a
9 See Prahalad and Hamel (1997). detailed search strategy and public transport mode featuring
10 See Zehtabchi (2019) for more data collection please check the small automated vehicles
detailed information about AV respective working papers. operating on a network of
patent and scientific publication 30 See Lee and Lim (2001). specially built guideways.
search strategy. 31 See Toyota Research Institute- 66 See Dennis and Brugeman (2019).
11 See Intel (n.d.). CSAIL (n.d.). 67 See Taylor (n.d.).
12 Tech includes: electronics, ICTs, 32 See Stone (2018). 68 See ARM (n.d.).
semiconductors and audio- 33 See Abuelsamid (2017). 69 See Poulanges (2017).
visuals. Auto includes: instruments, 34 See nuTonomy (2017). 70 See PSA Groupe (n.d.).
material, machines, engines 35 See Singapore Economic 71 See PSA Groupe (n.d.).
and transport, civil engineering. Development Board (2016). 72 See Valeo (2015).
Others include: biopharma, 36 See Engel (2017). 73 See Ministère de l'Enseignement
chemicals and environment and 37 See Bigelow (2019a). supérieur, de la Recherche et de
consumer goods. 38 See Bigelow (2019b). l'Innovation (2019).
13 See Zehtabchi (2019). 39 See Nicas (2017). 74 See Dennis and Brugeman (2019).
14 See Hummel (2019). 40 See Snavely (2017). 75 Visit www.keolis.com/en.
15 See Reuters (2019). 41 See University of Waterloo 76 See Dennis and Brugeman (2019).
16 A technology stack is the list of (n.d.) and McKenzie and 77 Visit www.transdev.com/en.
all the tools and technologies McPhee (2017). 78 See McQuilkin (2019).
used to build and run a 42 See University of Toronto (2019). 79 See UBI Mobility‐Connected
single product. 43 See Carnegie Mellon (1986). Cars France (2018).
17 See VIA Technologies (2018). 44 See U.S. Defense Advanced 80 See Dickmanns (2002).
79
81 See Gregor et al. (2002), Daimler 88 See Continental AG (2018). 94 See Leichman (2017).
(2016), Oagana (2016). 89 See Behrmann and 95 See Scheer (2018).
82 See Taylor and Rauwald (2018). 96 See Intel (2018).
Wissenbach (2019). 90 See Daimler (n.d.). 97 See Reichert (2019).
83 See Audi (2017). 91 See Initiative for Applied Artificial 98 See Saxenian (1996) and (2007).
84 See Ulrich (2019). Intelligence (n.d.). 99 See Zehtabchi (2019).
85 See Hetzner (2018). 92 See Singer (2018). 100 See Zehtabchi (2019).
86 See DeMattia (2018). 93 See South Africa Israel Chamber 101 See Ghemawat (1991).
87 See Continental AG (n.d.). of Commerce (2016).
80
References
Abernathy, W.J. and K.B. Clark (1985). Bigelow, P. (2019b). Waymo to build Cozzolino, A. and F.T. Rothaermel
Innovation: mapping the winds of self-driving cars in Detroit, invest (2018). Discontinuities, competition,
creative destruction. Research Policy, $13.6 million in factory. Crain’s Detroit and cooperation: Cooperative
14(1), 3–22. Business, April 23. dynamics between incumbents and
entrants. Strategic Management
Abernathy, W.J. and J.M. Utterback Bresnahan, T.F. and M. Trajtenberg Journal, 39(12), 3053–3085.
(1978). Patterns of industrial (1995). General purpose technologies
innovation. Technology Review, ‘Engines of growth’? Journal of Daimler. (2016). The PROMETHEUS
80(7), 40–47. Econometrics, 65(1), 83–108. project launched in 1986: Pioneering
autonomous driving. Press release,
Abuelsamid, S. (2017). Delphi acquires California Department of Motor September 20.
nuTonomy for $450M, advancing Vehicles. (n.d.). Testing of autonomous
push for automated Driving. Forbes, vehicles with a driver. www.dmv. Daimler. (n.d.). Reinventing safety:
October 24. www.forbes.com/sites/ ca.gov/portal/dmv/detail/vr/ a joint approach to automated
samabuelsamid/2017/10/24/delphi- autonomous/testing driving systems. www.daimler.com/
acquires-automated-driving-startup- innovation/case/autonomous/
nutonomy-for-450m Capron, L. and W. Mitchell (2012). reinventing‐safety‐2.html
Build, Borrow, or Buy: Solving the
ARM. (n.d.). Automotive Autonomous Growth Dilemma. Cambridge, MA: DeMattia, N. (2018). Klaus Fröhlich
Drive. www.arm.com/solutions/ Harvard Business Press. talks about BMW iNEXT. BMW
automotive/autonomous‐car Blog, September 24. www.bmwblog.
Carnegie Mellon. (1986). NavLab 1. com/2018/09/24/klaus-frohlich-to-
Arora, A. and A. Gambardella (1990). Robotics Institute History of Self- talks-about-bmw-inext
Complementarity and external Driving Cars. YouTube video, www.
linkages: the strategies of the large youtube.com/watch?v=ntIczNQKfjQ Dennis, E.P. and V.S. Brugeman
firms in biotechnology. The Journal of (2019). Automated and Connected
Industrial Economics, 38(4), 361–379. Carnegie Mellon University. (n.d.). No Vehicle Deployment Efforts: A Primer
Hands Across America. www.cs.cmu. for Transportation Planners. Ann
Audi. (2017). Audi piloted edu/~tjochem/nhaa/nhaa_home_ Arbor, MI: Center for Automotive
driving. media.audiusa.com/models/ page.html Research; Lansing, MI: Michigan
piloteddriving Department of Transportation.
Chesbrough, H.W. (2003). Open
Audretsch, D.B. and M.P. Feldman Innovation: The New Imperative Dickmanns, E. (2002). The
(1996). Innovative clusters and the for Creating and Profiting from development of machine vision for
industry life cycle. Review of Industrial Technology. Cambridge, MA: Harvard road vehicles in the last decade. In
Organization, 11(2), 253–273. Business Press. Proceedings Intelligent Vehicle
Symposium 2002, Versailles, June
Behrmann, E. and C. Rauwald (2018). Continental AG (2018). Continental 17–21. Piscataway, NJ: IEEE, 268–281.
ZF plans $14 billion autonomous expands automated driving
vehicle push, concept van. tests on the autobahn. Press Dziczek, K, E.P. Dennis, Q. Hong, Y.
Automotive News, September 19. Release, April 26. www.continental. Chen, V. Sathe-Brugeman and E.
com/en/press/press-releases/ Marples (2019). Automated Driving
Bigelow, P. (2019a). Waymo firms up cruisingchauffeur-128928 Technology Report. Unpublished
plans for autonomous car assembly background report for the World
plant in Detroit. Automotive News, Continental AG. (n.d.). Driverless Intellectual Property Organization.
April 23. mobility. www.continental-automotive.
com/en-gl/Landing-Pages/CAD/
Automated-Driving/Hidden-Pages/
Driverless-Mobility
81
Economist (2019). Chinese firms Gregor, R., M. Lutzeler, M. Pellkofer, Intel. (2018). Volkswagen, Mobileye
are taking a different route to K-H. Siedersberger and E. Dickmanns and Champion Motors to invest in
driverless cars. The Economist, (2002). MS-Vision: a perceptual Israel and deploy first autonomous EV
October 12. www.economist.com/ system for autonomous vehicles. ride-hailing service. Intel Newsroom,
business/2019/10/12/chinese-firms- IEEE Transactions on Intelligent October 29. newsroom.intel.com/
are-taking-a-different-route-to- Transportation Systems, 3(1), 48–59. news-releases/volkswagen-mobileye-
driverless-cars champion-motors-invest-israel-
Hawkins, A.J. (2019). Waymo is deploy-first-autonomous-ev-ride-
Engel, J. (2017). Optimus Ride drives making some of its self-driving car hailing-service/#gs.bt6x8i
off with $18M for autonomous vehicle data available for free to researchers.
tech. xconomy, November 2. xconomy. The Verge, August 21. www.theverge. International Energy Agency (IEA).
com/boston/2017/11/02/optimus‐ride‐ com/2019/8/21/20822755/waymo- (2019). Global EV Outlook 2019.
drives‐offwith‐18m‐for‐autonomous‐ self-driving-car-data-set-free- Paris: IEA. www.iea.org/publications/
vehicle‐tech research reports/globalevoutlook2019
Feifei, F. (2019). Autonomous Hetzner, C. (2018). Mercedes plans Jiang, H. and F. Lu (2018). To be
vehicles gaining more ground. advanced self-driving tech for next friends, not competitors: a story
China Daily, January 15. www. S class. Automotive News, October 11. different from Tesla driving the
chinadaily.com.cn/a/201901/15/ Chinese automobile industry.
WS5c3d2bb0a3106c65c34e46e2. Higgins, M.J. and D. Rodriguez (2006). Management and Organization
html The outsourcing of R&D through Review, 14(3), 491–499.
acquisitions in the pharmaceutical
Fleet News (2018). UK CITE enters industry. Journal of Financial Jing, M. (2018). Baidu’s self-driving
second phase of Coventry autonomy Economics, 80(2), 351–383. cars require more human intervention
testing with Jaguar Land Rover. Fleet than Alphabet’s Waymo. South China
News, July 2. www.fleetnews.co.uk/ Hummel, T. (2019). BMW, Daimler Morning Post, May 7. www.scmp.
news/environment/2018/07/02/ seal self-driving tech partnership. com/tech/enterprises/article/2144863/
uk-cite-enters-second-phase-of- Automotive News Europe, July 4. baidus-self-driving-carsrequire-more-
coventry-autonomy-testing-with- europe.autonews.com/automakers/ human-intervention-alphabets
jaguar-land-rover bmw-daimler-seal-self-driving-tech-
partnership Jovanovic, B. and G.M. MacDonald
Foundation for Law and International (1994). The life cycle of a competitive
Affairs (FLIA) (2017). China’s New Initiative for Applied Artificial industry. Journal of Political Economy,
Generation of Artificial Intelligence Intelligence. (n.d.). German Startup 102(2), 322–347.
Development Plan. flia.org/notice- Landscape of Autonomous
state-council-issuing-new-generation- Driving. appliedai.de/insights/ Klepper, S. (1997). Industry life cycles.
artificial-intelligence-development- autonomous-driving Industrial and corporate change,
plan 6(1), 145–182.
Intel. (n.d.). 5G Is key to fully realizing
Gans, J.S. and S. Stern (2003). The connected and autonomous Korosec, K. (2018). Waymo
product market and the market for vehicles. www.intel.com/content/ opens subsidiary in China.
“ideas”: commercialization strategies www/us/en/communications/5g- TechCrunch, August 24. techcrunch.
for technology entrepreneurs. connected-vehicle.html com/2018/08/24/waymo-opens-
Research Policy, 32(2), 333–350. subsidiary-in-china
Ghemawat, P. (1991). Market Lee, K. and C. Lim (2001).

incumbency and technological inertia. Technological regimes, catching-up
Marketing Science, 10(2), 161–171. and leapfrogging: findings from the
Korean industries. Research Policy,
30(3), 459–483.
82
Leichman, A.K. (2017). Porsche Ministère de l'Enseignement supérieur Prahalad, C.K. and G. Hamel
invests 8-figure sum in Israeli auto de la Recherche et de l'Innovation (1997). The core competence of
innovation. Israel21c, June 4. www. (2019). Lancement de 4 Instituts the corporation. In Strategische
israel21c.org/porsche-invests-8- Interdisciplinaires d’Intelligence Unternehmungsplanung/Strategische
figure-sum-in-israeli-auto-innovation Artificielle (3IA) et ouverture de deux Unternehmungsführung. Heidelberg:
appels à projets complémentaires. Physica, 969–987.
Liao, R. (2019). Search giant Baidu Press release, April 24. www.
has driven the most autonomous enseignementsup-recherche.gouv.fr/ PSA Groupe. (n.d.). On the road to the
miles in Beijing. TechCrunch, cid141320/lancement-de-4-instituts- autonomous car! www.groupe‐psa.
April 2. techcrunch.com/2019/04/02/ interdisciplinaires-d-ia-3ia-et- com/en/story/en‐route‐vers‐la‐voiture‐
baidu-self-driving-2018 ouverture-de-deux-appels-a-projets- autonome
complementaires.html
Lowensohn, J. (2015). Uber just Randall, T. (2019). Waymo starts
announced its own self-driving car Nicas, J. (2017). Google parent retires selling sensors to lower cost of self-
project. The Verge, February 2. www. ‘Firefly’ self-driving prototype. Wall driving cars. Bloomberg Hyperdrive,
theverge.com/2015/2/2/7966527/ Street Journal, June 13. March 6. www.bloomberg.com/news/
uber-just-announced-its-own-self- articles/2019-03-06/waymo-starts-
driving-car-project nuTonomy. (2017). Aptiv opens Boston selling-sensors-to-lower-cost-of-self-
Technology Center. Press release, driving-cars
MacDuffie, J.P. (2018). Response to December 12. www.aptiv.com/media/
Perkins and Murmann: Pay attention article/2017/12/12/aptiv-opens- Reichert, C. (2019). CES 2019:
to what is and isn’t unique about boston-technology-center Mobileye inks autonomous vehicle
Tesla. Management and Organization deals across China. ZD Net, January
Review, 14(3), 481–489. Oagana, A. (2016, January 25). A 8. www.zdnet.com/article/ces-2019-
short history of Mercedes‐Benz mobileye-inks-autonomous-vehicle-
McKenzie, R. and J. McPhee (2017). autonomous driving technology deals-across-china
Research and educational programs AutoEvolution. Retrieved from
for connected and autonomous www.autoevolution.com/ Reuters (2019). Audi to join
vehicles at the University of Waterloo. news/a‐short‐history‐ofmercedes‐ Mercedes, BMW development
Mechanical Engineering, 139(12), benz‐autonomous‐driving‐ alliance: paper. www.reuters.com/
S21–S23. www.memagazineselect. technology‐68148.html article/us-volkswagen-audi-bmw-
asmedigitalcollection.asme.org/article. daimler/audi-to-join-mercedes-
aspx?articleid=2676826 Perkins, G. and J.P. Murmann (2018). bmw-development-alliance-paper-
What does the success of Tesla mean idUSKCN1VC0YT
McQuilkin, K. (2019). Blacksburg’s for the future dynamics in the global
Torc Robotics is launching automobile sector? Management and Rothaermel, F. T. (2001). Incumbent's
autonomous shuttle buses in France. Organization Review, 14(3), 471–480. advantage through exploiting
RichmondInno, January 8. www. complementary assets via interfirm
americaninno.com/richmond/tech- Poulanges, M. (2017). Renault cooperation. Strategic management
news-richmond/blacksburgs-torc- presents eyes-off/hands-off journal, 22(6‐7), 687--699.
robotics-is-launching-autonomous- technology for the autonomous
shuttle-buses-in-france vehicle of the future. Groupe Renault, Rothaermel, F.T. and A.M. Hess
June 28. group.renault.com/en/news/ (2007). Building dynamic capabilities:
blog-renault/renault-presents-eyes- innovation driven by individual-,
offhands-off-technology-for-the- firm-, and network-level effects.
autonomous-vehicle-of-the-future Organization Science, 18(6), 898–921.
Saxenian, A. (1996). Regional

Advantage. Cambridge, MA: Harvard
University Press.
83
Saxenian, A. (2007). The New Taylor, M. (n.d.). AI in Cambridge: U.K. Department for Transport.
Argonauts: Regional Advantage in a the machine learning capital of the (2015). The Pathway to Driverless
Global Economy. Cambridge, MA: U.K.? Luminous PR. luminouspr.com/ Cars: Code of Practice for
Harvard University Press. cambridge‐the‐ai‐capital‐of‐the‐uk Testing. assets.publishing.service.
gov.uk/government/uploads/system/
Scheer, S. (2018). Mobileye gets Taylor, E. and I. Wissenbach (2019). uploads/attachment_data/file/446316/
self-driving tech deal for 8 million As Google races ahead, German pathway-driverless-cars.pdf
cars. Automotive News Europe, carmakers look to go faster on
May 17. europe.autonews.com/ autonomous driving. Reuters, Ulrich, L. (2019). 2019 Audi A8 review:
article/20180517/ANE/180519817/ January 23. www.reuters.com/article/ Tech-packed flagship delivers almost
mobileye-gets-self-driving-tech-deal- us-germany-autos/as-google-races- everything, except level 3 autonomy.
for-8-million-cars ahead-german-carmakers-look-to- The Drive, January 31. www.thedrive.
go-faster-on-autonomous-driving- com/new-cars/26252/2019-audi-a8-
Silver, D. (2018). Baidu brings the idUSKCN1PH18C review-tech-packed-flagship-sedan-
Waymo model to China. Forbes, delivers-almost-everything-except-
November 1. www.forbes.com/ Teece, D.J. (1986). Profiting from level-3-autonomy
sites/davidsilver/2018/11/01/ technological innovation: implications
baidu-brings-the-waymo-model-to- for integration, collaboration, licensing University of Toronto. (2019). News:
china/#d0b52193d961 and public policy. Research Policy, Self-driving cars. www.utoronto.ca/
15(6), 285–305. news/tags/self-driving-cars
Singapore Economic Development
Board. (2016). World’s first Teece, D.J. (2018). Tesla and the University of Waterloo. (n.d.). Centre
driverless taxi system comes to reshaping of the auto industry. for Automotive Research. uwaterloo.
Singapore. www.edb.gov.sg/en/news- Management and Organization ca/centre‐automotive‐research/
and-events/insights/innovation/world- Review, 14(3), 501–512. research‐expertise/connected‐
s-first-driverless-taxi-system-comes- andautonomous
to-singapore.html Teece, D.J. (2019). China and the
reshaping of the auto industry: a U.S. Defense Advanced Research
Singer, D. (2018). Israel’s artificial dynamic capabilities perspective. Projects Agency. (n.d.). The grand
intelligence start-ups. StartupHub. Management and Organization challenge. www.darpa.mil/about-
ai. www.startuphub.ai/israels- Review, 15(1), 177–199. us/timeline/-grand-challenge-for-
artificial-intelligence-startups-2018 autonomous-vehicles
Toyota Research Institute‐CSAIL.
Snavely, B. (2017). Roush expands in (n.d.). Joint Research Center. toyota. Valeo (2015). The Autonomous
Troy, will hire 150 engineers of self- csail.mit.edu Car Takes to the Road. Retrieved
driving tech. Detroit Free Press. May 9. from www.valeo.com/en/the‐
Tripsas, M. (1997). Unraveling the autonomous‐takes‐to‐the‐road
South Africa Israel Chamber of process of creative destruction:
Commerce. (2016). Autonomous complementary assets and incumbent Vasilash, G. (2018). Argo AI and
cars herald new era for Israeli high- survival in the typesetter industry. getting Ford to Level 4 autonomy.
tech. saicc.co.za/general‐motors‐to‐ Strategic Management Journal, Autoblog, September 8. www.
triple‐size‐of‐rd‐israelicentre 18(S1), 119–142. autoblog.com/2018/09/08/argo‐ai‐
ford‐level‐4‐autonomy‐self‐driving‐
Stone, A. (2018). What the UBI Mobility-Connected Cars France car
three tiers of automotive (2018). French delegation: Connected
marketing mean today. Forbes, autonomous vehicles.
June 28. www.forbes.com/sites/ 1419891vq14j2fapah1bpghjzyq.
forbesagencycouncil/2018/06/28/ wpengine.netdna‐cdn.com/
what-the-three-tiers-of- wpcontent/uploads/2018/05/French‐
automotive-marketing-mean- delegation_Ubimobility‐2018‐Detroit‐
today/#c449dec26510 Final.pdf
84
VIA Technologies (2018). VIA partners Wiggers, K. (2019) 5 companies Zehtabchi, M. (2019). Measuring
with Lucid to develop industry- are testing 55 self-driving Innovation in the Autonomous
leading VIA Edge AI 3D developer kit cars in Pittsburgh. Venture Vehicle Technology. WIPO Economic
powered by Qualcomm APQ8096SG Beat, April 26. venturebeat. Research Working Paper No.
embedded processor. www.viatech. com/2019/04/26/5‐companies‐ 60. Geneva: World Intellectual
com/en/2018/11/via-partners-with- are‐testing‐55‐self‐drivingcars‐in‐ Property Organization.
lucid-to-develop-industry-leading-via- pittsburgh
edge-ai-3d-developer-kit Zucker, L.G. and M.R. Darby (1997).
Xinhua (2019) Baidu’s self-driving Present at the biotechnological
WIPO (2019. WIPO Technology taxis to run in Changsha in late revolution: transformation of
Trends 2019. Artificial Intelligence. 2019. Xinhuanet.com, April 5. www. technological identity for a large
Geneva: WIPO. xinhuanet.com/english/2019- incumbent pharmaceutical firm.
04/05/c_137952253.htm Research Policy, 26(4–5), 429–446.
85
The impact of plant biotech
innovation reaches far beyond
the lab. Innovation produced
in a metropolitan hotspot can
benefit 75 times its land mass.
Chapter 4
Plant biotechnology – connecting

urban innovation and rural application
Humans began making genetic improvements Conventional breeding of new crop varieties and traits
to plants thousands of years before anyone involves the sexual reproduction of two compatible
even knew what a gene was. Starting circa crop varieties to produce a mutated offspring with the
10,000 BCE, they initially began by selecting and desired biological traits.2 This method often requires
domesticating crops from the natural biological many crosses to get to the right combination of genes
diversity of plants. These crops differed from to produce the desired crop. It also needs the crops
their wild predecessors through the propagation to be sexually compatible.
of carefully chosen specific plant materials,
which were cultivated for human consumption Today, new crop varieties can be achieved by biotech
and use.1 nology. This modern technique relies on an understand
ing of a plant’s genetic make-up and uses different
The techniques used to select and propagate methods of genetic engineering to make changes to its
crop varieties with desirable traits – known deoxyribonucleic acid (DNA), molecules of nucleotides
as cultivars – can generally be divided into which carry the genetic instructions for the develop
three categories: the traditional, which began ment, functioning, growth and reproduction of all
with domestication, the conventional and the known organisms.
modern. All three methods are in use today to
varying degrees. Biotechnology refers to “any technological applica
tion that uses biological systems, living organisms,
or derivatives thereof, to make or modify products or
processes for specific use.”3 It can also involve the
implementation of advanced molecular and cellular
technologies and techniques. In both the broader and
narrower sense of its application, agricultural biotech
nology relies on the discoveries and research tools of
a relatively new science field.
It is changing the agriculture industry.4 Advances

have produced crops that are resistant to certain
diseases, that result in higher yields than before, that
can grow in extreme soil conditions, such as in arid
and salty environments, and that are even infused
with nutrients.5
Biotechnology innovation has the potential to increase

agricultural productivity and quality, ultimately rais
ing incomes for farmers across the world. It can also
address environmental concerns about the use of
chemical pesticides. Klümper and Qaim (2014) show
that genetically modified technology had increased
farmer profits worldwide by 68 percent, crop yields
by 22 percent and reduced use of chemical pesticides
by 37 percent. Brookes (2018) estimates that each
extra U.S. dollar spent on transgenic soy seed – seed
containing genes from another organism – relative
87
to conventional seed raises a farmer’s income by How biotechnology found its

USD 3.88. The gains reflect higher yields and lower way into agriculture
costs from using fewer pest and weed controls.
Moreover, the technology is seen as a potential solu The origin of agricultural biotechnology can be traced
tion to global issues of hunger and poverty. back to 1866 when an Augustinian friar, Gregor Mendel,
postulated the fundamental laws of genetic inheritance,
This chapter uses the case of agricultural biotechnol based on his work on pea plants. He laid the ground
ogy, specifically plant biotechnology, to illustrate the work for scientific breeding and genetic engineering.
workings of a global innovation network.6 It relies on
information contained in patent documents and scien Subsequent breakthroughs and discoveries in the
tific publications to identify the actors and locations of 1920s and 1930s on methods of chromosome and gene
innovation. It further exploits these two complementary mutation, followed by the discovery of the double helix
measures of innovative activities to demonstrate how structure of DNA in 1953, at Cambridge and London
different innovation clusters link to one another. in the United Kingdom (U.K.), led to an explosion in
research in genetics – the study of genes, genetic
The first section of this chapter describes the evolution variation and heredity in organisms.
of the plant biotechnology industry and identifies the
factors that drive innovation. The second looks at how However, it was the development of recombinant
the industry’s innovation landscape has evolved and DNA (rDNA) technologies – the splicing together of
how more countries are involved in innovation than strands of DNA from one organism to another – in bacte
ever before. The penultimate section examines the ria in 1974 by researchers at Stanford University and the
links between innovation centers in different parts of University of California, San Francisco, in the United
the world. The chapter concludes with how the plant States of America (U.S.), that cleared the way for genetic
biotechnology’s global innovation landscape as well engineering to take place in plants and other organisms.
as the network may change due to new developments
in the industry. Table 4.1 lists a few breakthrough discoveries as well as
innovations that form the basis of biotechnology meth
ods and their application in plant biotechnology today.
4.1 The rising importance
of plant biotechnology The commercial application of biotechnology tools and
techniques first found its way into the field of medicine
Plant biotechnology generally involves three areas in the mid-1970s; agricultural use began a few years
of the farming industry: (i) plant breeding and later.7 This was primarily because molecular biology
seeds, (ii) soil health and fertility and (iii) pest control was mainly developed in medical schools and universi
and pesticides. ties, which were not much concerned with agriculture.8
The application of biotechnology in plant breeding and However, as the use of biotechnology in medicine and
seeds refers to the development of new varieties and for human health became more prominent, scientists
traits through hybridization, outcrossing (interbreeding), began to apply biotechnology to veterinary science
mutation, tissue culture, grafting and cloning of plants, for animal health and then to plant breeding. Animals
genetic engineering and editing of the genome, which came first, because of their relative genetic proximity
is the whole of the hereditary information encoded in to humans.9
a plant’s DNA, to name a few. Most innovation is in
this area. By the mid-1980s, the crop biotechnology industry
had begun to grow. Several landmark legal decisions
For soil health and fertility, biotechnology involves using in the U.S. regarding whether living organisms may be
biofertilizers – the culturing and use of microbes for soil patented led to the granting of patents on genetically
amendment and plant growth. Lastly, biotechnology for engineered plants.10 Toward the end of the decade, field
pest control and pesticides deals with biocontrol strate trials of transgenic plants were underway in Australia,
gies, biopesticides, breeding and genetic engineering Canada, the U.S. and some European countries.
of pest resistance traits in crops, as well as mutation Mexico, a developing economy, also began conduct
and genetic engineering for herbicide tolerance. ing field trials of transgenic crops around the same
88
4 Plant biotechnology – connecting urban innovation and rural application
A brief history of key scientific biotech advances

Table 4.1 Selected discoveries or scientific breakthroughs in crop biotechnology
Year Discovery/scientific breakthrough Affiliation
Stanley Cohen and Herbert Boyer developed a technique –

Stanford University and University of California,
1974 rDNA – that would splice together strands of DNA from more
San Francisco, California, U.S.
than one organism, paving the way for genetic engineering
DNA sequencing methods were independently

Harvard University, Cambridge, Massachusetts,
1977 devised by Walter Gilbert with graduate student
U.S., and Cambridge University, U.K.
Allan Maxam, and Frederick Sanger
George Willems and Robert Schilperoort

1981 genetically engineered first plant (tobacco) using University of Leiden, Leiden, Netherlands
the bacterium Agrobacterium (see Box 4.1)
Complete sequencing of Arabidopsis thaliana

Consortium of universities as well as public research
2000 (a small plant) genome, published in 2000 as
institutions in the U.S., Japan and Europe
part of the Arabidopsis Genome Initiative
University of California, Berkeley, California, the U.S;

and University of Vienna, Austria; Massachusetts
2012 A new genome editing technique, CRISPR-Cas9, is developed
Institute of Technology and Harvard, Cambridge,
Massachusetts, the U.S.; Vilnius University, Lithuania
Source: Graff and Hamdan-Livramento (2019).
time.11 Meanwhile, China became the first country to What shapes innovation in the field?
commercially farm a pest-resistant, transgenic tobacco
plant in 1988. However, Chinese farmers stopped Policies, rules and regulations shape and affect innova
farming the plants in the mid-1990s as many tobacco- tion in the plant biotechnology industry. They include
producing companies expressed concerns about using the availability of IP rights as a mechanism for ensuring
these genetically improved organisms (GIOs) in their a return on investment in innovation, and regulations on
products. In particular, consumers were apprehensive health and safety and on protecting the environment,
about consuming GIO cigarettes. These landmark among others.
events coincided with a rise in the number of special
ized agricultural biotechnology startups, particularly in
the U.S., during the 1980s and 1990s. Appropriate returns on investments
Figure 4.1 plots the rising global demand for patents Most jurisdictions do not allow for the patentability of
for plant biotechnology inventions from 1970 to 2016. things that exist in nature, including biological organ
The red line shows the total number of patent applica isms. However, the lines have become blurred with new
tions in the field filed at different intellectual property technological advances in biotechnology.13
(IP) offices worldwide.12 By contrast, the dark red line
represents the number of initial filings for new plant Concerns with patentability on agricultural biotechnol
biotechnology patents, also known as first filings. The ogy innovation are similar to those expressed about
difference between the total number of patent filings patenting in the biotechnology field generally.14 Granting
and the first filings is referred to as the subsequent exclusive rights on research tools may dampen follow-
filings. The subsequent filings represent the number of on innovation.15 In crop biotechnology, patents could
applications filed in multiple countries, or jurisdictions, make it difficult for poorer economies to benefit from
for the same inventions. The widening gap between research that could alleviate poverty and address
the two lines from the 1980s onward shows how inven world hunger problems. In addition, critics have argued
tors increasingly requested patent protection for their that most of the patents granted are too broad and
inventions from competitors in more than one market. are likely to infringe on other proprietary technology,
More importantly, it illustrates the growing commercial resulting in the relatively high amount of litigation seen
importance of plant biotechnology inventions globally. in the industry.
89
The
Totalrise of genetic
patent filings ofengineering coincided
plant biotechnology, with an increase in subsequent patent filings
1970–2016
Figure 4.1 Total patent filings of plant biotechnology, 1970–2016
30,000
CRISPR-Cas9 developed
Arabidopsis genome sequencing complete

Number of patents
20,000
France approves first GMO (tobacco)
US DoA approves commercial sale of transgenic tomato

China grows the first transgenic crop (tobacco) for commercial use
First US field trials of GMOs (tobacco and tomatoes)

10,000
USPTO extends protection to GE plants
Diamond v. Chakrabarty decision

rDNA technique developed
0
1970 1975 1980 1985 1990 1995 2000 2005 2010 2015
FIRSTFILINGS
FIRST FILINGS PATENT
PATENT FILINGS
FILINGS
In the U.S., two changes related to IP policy in the biotechnological result, or on complementary assets
1980s played pivotal roles in shaping the agricultural – infrastructure, capabilities or other inventions – that
biotechnology industry there. In particular, they led lead to the final crop biotechnology product.19
to the increasing reliance on IP as a way to ensure
appropriate returns on investing in innovation.16
Protecting consumers and
The first was the passing of the Bayh–Dole Act in 1980. safeguarding the environment
The Bayh–Dole Act allows for the patenting of research
from universities, even if it is funded by taxpayers. The increasing potential commercial importance of
The second was the extension of patent protection plant biotechnology led government regulators and the
to genetically modified organisms (GMOs) through a public to question when and how to ensure that these
landmark case – Diamond v. Chakrabarty – decided by purposely transformed, or transgenic, crops would not
the U.S. Supreme Court of Justice, also in 1980.17 By harm human health or the environment.
1985, the U.S. Patent and Trademark Office (USPTO)
had extended patent protection to genetically engi There are several layers of regulations on the use of
neered plants. Europe and the rest of the world soon plant biotechnology, at both the national and inter
followed suit. national levels.20 They help ensure that GIOs meet
biosafety, food safety and consumer protection stan
At the same time, the launch of the World Trade dards. For example, at the international level, the United
Organization (WTO) in 1995 included internationally Nations’ Codex Alimentarius sets the guidelines for
binding rules for the protection of IP rights in signatory food safety standards, the Cartagena Protocol on
countries. This opened the way for many multinational Biosafety, an international agreement, provides guide
companies (MNCs) to file for patent protection on their lines for biosafety regulations, and another international
plant biotechnology inventions. But some developing pact, the Aarhaus Convention, gives the general public
economies, like Brazil, restrict the patenting of certain a right to access information about policy decisions
plant biotechnology products, particularly those that affecting the environment.21
relate to seeds or new plant varieties. Instead, the
private sector in Brazil relies on sui generis rights At the national level, there are generally at least
to protect their innovations.18 Some file for patents three regulatory processes before a new transgenic
on the development process itself, rather than the plant can be commercially farmed. They include:
90
Box 4.1 Key differences between and conventional methods are often unpredict
breeding techniques able. Breeders choose the parents with the desired
traits to cross but the progeny may not carry the
There are two ways to introduce desired traits into genotype with the desired traits or display it,
plants and they differ according to plant type. Dicots, the phenotype.
or broad-leaved crops, such as cotton, soybean and
tomato, rely on the transformation brought about by Modern breeding techniques, such as genetic
a bacterium known as Agrobacterium tumefaciens. engineering, allow for targeted transfer of desir
In nature, this bacterium infects plants, inserting able crop traits – the transgene – and the breeding
some of its own DNA directly into the DNA of the of new transgenic plants in an efficient and fast
plant. By modifying the bacterium to exclude its manner. These transgenic crops are also known
unwanted traits and include the gene of interest, a as GMOs. The modern techniques simplify the
crop may be transformed through bacterial infection. breeding process by bypassing the need for the
The cells containing the new gene subsequently sexual compatibility of the plants with the desired
can be identified and grown using plant cell-culture traits and allow for the selection of desirable traits
technology into a whole plant that now contains the from any living organism. The desired traits can
new transgene incorporated into its DNA. come from the same species or a cross-species;
they can even come from a modification of the
Monocots, or grass species, such as maize, wheat expression of the plant’s own genes. Targeting of the
and rice, are transformed by physically shooting desired gene, tracking it and inserting it into a crop’s
small tungsten balls coated with an external DNA DNA ensures a clean breed of the crop and excludes
into the plant’s genome. Some of the DNA comes the potential for unwanted, ancillary traits, which are
off and is incorporated into the DNA of the recipient often a by-product of traditional and conventional
plant. Those cells can also be identified and grown breeding. Moreover, there is a faster turnaround in
into a whole plant that contains the foreign DNA. the development of new crop varieties in modern
breeding techniques than its predecessors.
The differences between traditional and conven
tional breeding and their modern counterpart boil
Source: FAO (2004). and Persley and Siedow (1999).
down to the control over the breeding process. The
outcomes of plants bred through the traditional
(i) approvals to conduct field tests, (ii) approvals to directives on GMOs. 24 During the moratorium, the
farm for commercial purposes and (iii) approval to sell Commission differentiated between plants whose
and market to consumers. In the U.S., the agencies genes had been edited with conventional breeding
include the U.S. Department of Agriculture (USDA) methods and those that had been genetically edited
and the U.S. Environmental Protection Agency (EPA) using biotechnology tools (see Box 4.1). The measures
for field trial approval and the U.S. Food and Drug established specific requirements for conducting field
Administration (FDA) for commercial approval. tests and planting of transgenic crops, their import and
use and the labeling of GMO products.
Europe was responsible for some important break
throughs in plant biotechnology. In the early 1990s, Several commercial explanations have been put
Belgium, France and the U.K. were among the top five forward for this change in the continent’s attitude
countries in the industry, which together accounted for toward transgenic crops, besides the strong political
nearly 95 percent of transgenic crops released – the campaigns against GMOs mounted by environmental
other two were Canada and the U.S. However, by the and consumer groups. Graff and Zilberman (2007)
turn of the century, European sentiment toward trans argue that Europe’s strong agrochemical businesses
genic crops had changed significantly.22 enjoyed a comparative advantage in chemistry and
wanted to prevent their competitors from entering the
Between 1998 and 2004, the European Commission market. Sheldon (2004) proposed that EU farmers
(Commission), the European Union’s executive arm, saw the measures limiting the approval of genetically
and five EU member states imposed a de facto mora engineered plants as an opportunity to prevent agri
torium on the approval of GMOs.23 From 2003, the cultural commodities from the rest of the world from
Commission then put in place several regulations and entering the market.
91
Crop biotechnology patent filings from the U.S. have grown faster than from Europe
Figure 4.2 Share of plant biotechnology filing over total patent filings by origin (%), 1970–2016
Share of plant biotechnology filing over total patent filings by origin, 1970–2016
4.0
CRISPR-Cas9 developed
3.0
EU issues regulations on GMO
Arabidopsis genome sequencing complete

Percent
2.0 EU defacto moratorium begins
France approves first GMO (tobacco)

US DoA approves commercial sale of transgenic tomato
1.0 First US field trials of GMOs (tobacco and tomatoes)
USPTO extends protection to GE plants
Diamond v. Chakrabarty decision
rDNA technique developed
0.0
1970 1975 1980 1985 1990 1995 2000 2005 2010 2015
U.S. U.S.
EUROPE (STILL PRODUCING) EUROPE
REST (STILL PRODUCING)
OF EUROPE REST OF EUROPE
Regardless, the EU regulations have arguably had a agricultural uses toward medicinal and biofuel indus
cooling effect on research and development of agri trial applications.26 One of the major European MNCs
cultural biotechnology on the continent. in the industry, BASF, a German chemical firm, halted
the development and marketing of its transgenic crops
Figure 4.2 shows the share of patent filings by the U.S. for the EU in 2012.27
(in dark red) and several European countries (in red and
light red) of the total plant biotechnology filings world
wide.25 Until the late 1990s, the share of plant biotech Who drives innovation?
nology patents filed in the U.S. and the EU rose more
or less in parallel. However, from 1997 onward there is Innovation in the plant biotechnology industry is driven
a widening gap between patent filing growth rates in by investments by both the public and private sectors.
the U.S. and the EU. It is difficult to say conclusively if
this is due to Europe’s de facto moratorium. But since
1998, EU countries have been filing crop biotechnology Strong public sector push in
patents at a relatively similar – if not slower – rate than agricultural research
the total patent filing rate.
The public sector plays a pivotal role in plant biotech
Industry has reported that the EU stance on transgenic nology research by funding and providing important
crops has affected companies’ business strategies. infrastructure necessary for research. In Europe as
A study conducted by the USDA Foreign Agricultural well as the U.S., key policy documents and published
Service showed that many European companies have reports underscore the importance of investing in
shifted their research and development (R&D) outside genetic engineering research.28
of Europe, relocating to places such as the U.S. While
public institutions and universities in Europe continue Scientists and researchers in public research institu
to conduct basic research into plant genetics, the tions have made important discoveries that paved the
likelihood of these outputs reaching the EU market is way to genetic engineering. The importance of basic
small. In addition, the report noted that many European science to innovation in the plant biotechnology field
biotechnology firms have shifted their focus away from continues today. For example, fundamental advances
92
in molecular biology and genetics have created more Collaboration between public institutions is important,
efficient ways to identify and target specific genotypes especially when trying to commercialize GIOs in less
in organisms. Moreover, the development of CRISPR- developed economies. Most of the transgenic crops
Cas9, a gene editing technology, has significantly cut planted in these economies during the late 1990s were
costs in the field of genetic engineering.29 locally adapted germplasms of their North American
counterparts.33 In poorer countries, IARCs can act as
Agricultural research centers and universities specializ nodes in global networks of innovation by connecting
ing in agricultural science play a pivotal role in adapting agricultural scientists and breeders across many NARS,
research and diffusing crop biotechnology innovations. including agricultural research universities in the world.
These research centers are mandated to conduct and
develop work that would improve agronomics and Plant biotechnology innovation has the potential to
advance genetic improvements in crops and agricul address food scarcity and food security issues.34 Thus,
tural innovation in general. Moreover, support for their advances find strong support from the public sector,
research work includes funding from governments, including intergovernmental organizations and not-
foundations and various intergovernmental and not-for- for-profit institutions and agencies, to promote their
profit organizations and agencies. The strong mandate diffusion to the rest of the world.35
and financial support help ensure the continuity and
importance of these institutions’ work. Governments fund most of the agricultural research
in many emerging economies. In a few cases, such
In the U.S., the 1862 Morrill Act established Land-Grant as China, India and Brazil, public sector R&D spend
colleges by allocating 30,000 acres (nearly 121.5 km2) ing in agriculture has increased rapidly. From 1990 to
of federal land across the country to build colleges 2013, China’s public sector agricultural R&D spend
and universities that would teach and promote the ing grew nearly tenfold, from USD 1 billion to more
development of agriculture, among other things. The than USD 9 billion.36 At the same time, India’s spend
second Morrill Act, passed in 1890, ensured that these ing tripled, from less than USD 1 billion to almost
colleges had regular federal funding. USD 3 billion and Brazil’s almost doubled, from less
than USD 2 billion to almost USD 3 billion. By contrast,
The success of Land-Grant colleges laid the foundation public sector spending in agriculture in the U.S. grew
for establishing similar research centers in emerg only moderately from about USD 4 billion in 1990 and
ing economies.30 The International Maize and Wheat declined from 2003 onward.
Improvement Center (CIMMYT) in Mexico City, Mexico,
and the International Rice Research Institute (IRRI) But many emerging economies, especially those with
in Los Baños, the Philippines, were the first two limited capacities to innovate in plant biotechnol
such establishments. These two national agriculture ogy and/or those that lack the financial resources to
research systems (NARS) would later become part of conduct research in the field tend to rely on the work
the Consultative Group for International Agricultural of NARS and/or IARCs.37
Research (CGIAR), an umbrella organization of 15
independent, non-profit research centers focused
on innovation in agriculture. The CGIAR has shaped Incentives backed by market consolidation
the historical evolution of innovation in agricultural
biotechnology, particularly in crop genetic development. Small, university-based startups initially dominated the
plant biotechnology market in the early years. However,
Agriculture requires innovation to be adapted to differ from the 1990s onward, MNCs bought many of them
ent regional agro-ecological conditions, which include out. One study estimated that nearly 90 percent of
the combinations of soil, landform and climatic char all research and development agreements on agri
acteristics.31 This means that transgenic crops should cultural biotechnology were between startups and
be bred with local cultivars and tested in local fields. large MNCs.38
In many developing countries these cultivars and
germplasms – living genetic resources held for animal At the same time, firms in the seed, chemical and
or plant breeding or preservation – are kept by public fertilizer industries, both in richer and poorer coun
institutions, such as the NARS or international agricul tries, have experienced significant market concentra
ture research centers (IARCs).32 tion.39 There are many reasons. The high fixed costs
93
of commercializing transgenic plants require large Industry has seen significant concentration
financial resources, which many startup companies
Table 4.2 Selected alliances in the industry,
may not have. Second, the high fixed costs also neces
1996–2016
sitate increasing reliance on IP rights to ensure a return
on investment. Accumulated proprietary technologies
in plant biotechnology can be a barrier to innovation, Bayer [Germany] purchases Monsanto [U.S.] (2016)
as they are in the semiconductor industry. Firms that
collaborate are less likely to infringe on one another’s
Monsanto [U.S.] (merged with Bayer (Bought Aventis
IP. For example, Monsanto, BASF, Dow, Bayer, DuPont Pharmacia March, 2000; spun CropScience, 2001) [Germany]
off entirely August, 2002)
and Syngenta cross-licensed one another’s IP rights
on transgenic crops.40 Biotechnology Agricultural chemicals
• Agracetus [U.S.] (1995) • Hoechst [chemical,
By 2001, 30 separate firms in the seeds and agrochemi • Calgene [U.S.] (1996) Germany] merged with
• Ecogen [U.S.] (2003) Schering [pharmaceutical,
cal industries had been reduced to six – Monsanto, • Joint venture with Millennium Germany] to create Hoechst
DuPont, Swiss-based Syngenta, Bayer, Dow and Pharmaceuticals [U.S.] (1998) Schering AgrEvo (1994)
BASF. The four biggest account for almost 60 percent • Paradigm Genetics [U.S.] [Dusseldorf, Germany]
(2000), name changed • Hoechst (Agrevo) and Rhône-
of the agricultural biotechnology market. The major to Icoria (2004) Poulenc [pharmaceutical,
agrochemical-seed groups are: Bayer CropScience France] merged (and their
Seeds
and BASF in Germany, Corteva Agriscience in the • DeKalb [U.S.] (1996)
agrochemicals division
became) to Aventis
U.S. and ChemChina, which purchased Syngenta in • Asgrow [U.S.] (1997)
CropScience (1999);
2017, in China. • Holden's Foundation
• Bayer buys Aventis
Seeds [U.S.] (1997)
CropSciences in August 2002
• Cargill International
This consolidation of crop biotechnology innovation in Seeds, Plant Breeding Biotechnology
the hands of a few has not necessarily translated into International [U.S.] (1998) • Plant Genetics Systems
• Delta & Pine Land [U.S.] (PGS) (acquired by AgrEvo
a reduction of innovative activities in this field.41 (alliance, 1994; bought 2007) in 1996; became part of
• Sensako [South Africa] Monsanto in 2002) [Belgium]
Table 4.2 lists selected alliances, including mergers and (2002); Carnia [South • PlanTech [Japan] (1999)
Africa] (2002); later merged • Lion Biosciences (11.3%, 1999)
acquisitions, of the multinational seed and agrochemi under DaKalb brand • Limagrain (purchased
cal companies. It shows how the industry has become • Seminis [U.S.] (2005) the Canadian seeds
more concentrated since the 1990s. • Emergent Genetics activity; 2001) [France]
[U.S.] (2005)
Seeds
• Acquired De Ruiter
• Nunhems [Netherlands],
[Netherlands] (2008); and
Vanderhave [Netherlands],
Need for public–private collaboration Peotec Seeds S.r.l. [Italy]
Plant Genetic Systems
(2008) via Seminis
[Belgium], Pioneer
Vegetable Genetics,
Zilberman et al. (1997) conducted a survey of plant
Sunseeds (1997) [U.S.]
biotechnology firms in the U.S. and found many cases • Nunza (vegetables), Proagro
of collaboration between the public and private sectors. (India) and two Brazilian
seed companies (1999)
In particular, they reported that in most patterns of
• Fibermax (joint venture
plant biotechnology innovation the universities made with Cotton Seed Inc.
the important discovery and the private sector then of Australia, 2000)
developed and commercialized the innovation. This

pattern of collaboration between the private and public Source: Updated, based on Pray and Naseem (2003).
sectors continues.
Large chemical and seed MNCs commercialized and

cultivated all of the major transgenic crops bred through
genetic modifications in the early years.42 The only
exception was Bacillus thuringiensis (Bt) cotton, which
was developed by a Chinese public research institu
tion, the Biotechnology Research Center of the China
Academy of Agricultural Sciences (CAAS) in Shenzhen.
However, CAAS entered into a joint venture with U.S.
94
ChemChina [China] purchases Corteva Agriscience [U.S.] spinoff created in (2019),

BASF [Germany]
Syngenta [Switzerland] (2017) result of DuPont and Dow merger (2015)
Syngenta [Switzerland] Dow Chemical [U.S.]; DuPont [U.S.]

Dow AgroSciences [US]
Agricultural chemicals Agricultural chemicals Biotechnology Agricultural chemicals

• Ciba-Geigy and Sandoz • Dow purchases Eli Lilly's [U.S.] • Alliances with Human Genome • Bought corn herbicide
merged to form Novartis 40% share of Dow Elanco Sciences [U.S.] (1996) business from Sandoz
[Switzerland] (1996) for USD 900 million (1997) • Alliance with Curagen [Switzerland] (1996)
• Novartis [Switzerland] buys • Rohm and Haas Agricultural [U.S.] (1997) • American Cyanamid [U.S.],
Merck's pesticide business Chemicals [U.S.] (2001) • Purchased Verdia from crop protection subsidiary
for USD 910 million (1997) Maxygen [U.S.] for USD from American Home Products
Biotechnology
• Merger of Novartis agriculture 65 million (2004) for USD 3.8 billion (2000)
• Mycogen (1996) [ U.S.]
division [Switzerland] and
• Ribozyme Pharmaceuticals Seeds Biotechnology
AstraZeneca's Ag. Chemicals
Inc. [U.S.] (1999) • Pioneer [U.S.] (1997, 20%) • Joint venture with Institute of
[U.K.] to form Syngenta
• Contract with Proteome • Hybrinova [France] (1999) Plant Genetics and Crop Plant
[Switzerland] (1999)
Systems Limited Research [Germany] to create
Biotechnology [Australia] (1999) SunGene [Germany] (1998)
• Zeneca Ag. [U.K.] bought • Joint venture with Max Planck
Seeds
Modern International N.V. Institute [Germany] and
• Mycogen buys Agrigenetics
[Netherlands] (1997) Metanomics [Germany] (1997)
[U.S.] (1992)
• Alliance with Japan Tobacco
• United AgriSeeds Seeds
[Japan] on rice (1999)
[U.S.] becomes part of • Bought 40% of Svalöf
• Alliance with Diversa
Mycogen (1996) Weibull [Sweden] (1999)
[U.S.] (2003)
• Joint venture of Mycogen [U.S.]
• Zeneca [pharmaceutical,
with Boswell [U.S.] on cotton
U.K.] buys PSA Genetics (via
seed to form Phytogen (1998)
Garst subsidiary, 1999)
• Joint venture with Danisco
Seeds Seeds [Denmark] (1999)
• Merger between Northrup- • Agreement with Illinois
King and Ciba Seeds brings Foundation Seeds [U.S.] (1999)
together S&G Seeds, Hilleshog • Cargill Hybrid Seeds
and Rogers Seed Co. under [U.S.] (2000)
one umbrella (1997)
• ICI (Imperial Chemical
Industries, pharmaceuticals
and agrochemicals) [U.K.]
splits into Zeneca (including
ICI seeds) and ICI PLC (1993)
• Garst [U.S.] reborn as a
Zeneca company (1996)
• Zeneca [U.K.] via Garst [U.S.]
buys Agripro Seeds [U.S.]
(1998), Gutwein Seeds (2000)
95
firms, Monsanto and Delta and Pineland, and China’s The increasing need for collaboration between the
Heibei Provincial Seed to bring Bt cotton to market. private and public sectors implies some changes to
The transgenic crop was made available to Chinese the use of IP. Research institutions in many emerging
farmers in 1997.43 economies used to shy away from relying on the IP
system and focused instead on ensuring that knowl
The need for access to complementary assets in edge could be easily shared. This view has changed.
innovating in agricultural biotechnology necessitates Collaboration between the two sectors – either to help
collaboration between innovators. Commercialization with commercialization (for the research institutions) or
of research work from universities, or public research as sources of germplasms and cultivars (for the private
institutions, both in developed and developing econo sector) – necessitates a hybrid approach to IP use.
mies, may require further assistance from the private
sector. This was the case for the Chinese Bt cotton Evidence collected from patent documents shows
and continues to be the case for many joint research that the number of collaborations between the private
projects between university research labs and and public sectors is on the rise. On average, only
private companies. 18 percent of plant biotechnology patents are inventions
with co-applications.48 However, this underestimates
In many developing economies, there are a few instanc collaboration activities. Not all collaborations lead to
es of collaboration between the NARS and large MNCs patented inventions and the number does not accu
to develop transgenic crops adapted to the region.44 rately capture collaborations between subsidiaries of
These public institutions may need access to propri large MNCs in different locations, because, typically,
etary biotechnological research tools to conduct their only the headquarters appears as the applicant on
research work, and thus would need the collaboration many MNC patent applications. Moreover, some of
of the IP owners. One such example is through licensing the public–private collaborations take place during
in proprietary technologies held by private companies.45 the commercialization stage, such as during field trials,
Another is through purchasing the technology at an and these are not generally captured by either patent
agreed cost. The firm may be paid by funds raised by or scientific publication data.
donor countries. Collaboration between the IARC and
the private firm may be made available royalty-free to Figure 4.3 plots the number of co-applications involv
developing economies or on reasonable royalty terms.46 ing the private and public sectors. The trend shows an
For example, the International Potato Center (IPC) in increase in the share of patents filed with at least one
Peru made arrangements with Plant Genetics Systems, public-sector applicant since 1999.
a Belgian firm later acquired by Bayer CropScience,
to use Bt genes to test its line of genetically engi
neered potatoes.47 4.2 The innovation landscape
of plant biotechnology
Private firms may collaborate with NARS or IARCs
on research in return for exclusive commercial rights The global innovation landscape of plant biotechnology
on any resulting technology in developed markets. is spread relatively widely across the globe. Figure 4.4
Developing economies would be entitled to the result plots the innovation landscape using two proxies for
ing technology at a preferential rate. There could also innovative activities – patents and scientific publication
be a hybrid approach to IP, with the private company in the form of articles and conference proceedings (see
applying for patents in developed markets only. Chapters 1 and 2) – for two time periods, 1998–2007
(top) and 2008–2017 (bottom).
Collaborations have also been initiated by the private
sector. For example, large life sciences firms may need It shows the evolution of innovative regions in the
access to different pools of germplasms administered industry and illustrates how patenting and publication
by various IARCs and NARS for further innovation. tend to mirror one another, at least for the top plant
CGIAR has a collection of germplasm which it has biotechnology clusters. The four top countries for
committed to keep in the public domain. Access to innovation activities in plant biotechnology are China,
this pool of germplasm could help in cultivating various Germany, Japan and the U.S., with Switzerland join
different versions of transgenic crops for use in many ing them in the top five for patenting and France for
parts of the world. scientific publication.
96
Innovation by the private sector is the main driver of agricultural biotechnology

innovation, but public–private and public–public collaborations are on the rise
Patented
Figure 4.3collaborations byco-application
Trend in patent type, Patented
type by number (left)collaborations by type,
and share (right), 1980–2016
1980–2016 1980–2016, shares
100
500
500
80
400
400
Number of patents
300
300 60
200
200 Percent 40
100
100 20
00 0
1980
1980 1985
1985 1990
1990 1995
1995 2000
2000 2005 2010 2015 1980 1985 1990 1995 2000 2005 2010 2015
PRIVATE
PRIVATEAND
PRIVATE & PRIVATE
AND PRIVATE
PRIVATEPRIVATE
PRIVATEAND
& PUBLIC
PRIVATE PUBLIC PUBLIC
PUBLICAND
AND PUBLIC &PUBLIC
PUBLIC
PUBLIC PRIVATE AND PRIVATE PRIVATE AND PUBLIC PUBLIC AND PUBLIC
AND PUBLIC
In addition, Figure 4.4 also shows how some regions innovative activities in the U.S. are carried out by the
lean more toward patenting, while others toward scien private sector, which tends to rely on patents, while in
tific publication. The U.S., Europe, Japan and China China, universities and public institutions are the main
show more patenting, while developing countries sources of such activities.
generally have more regions featuring scientific publi
cation work.49 Figure 4.5 maps the international and national clusters
of crop biotechnology across the world. These clusters
The difference in innovative output as captured by are determined by the relative significant co-location of
Figure 4.4 for patents and scientific publication can inventors and authors of plant biotechnology outputs
be quite significant for plant biotechnology. There are across countries (international clusters) and within
two reasons for this. countries (national clusters).
First, patenting of plant biotechnology inventions is Box 4.2 highlights the method used to identify these
subject to different criteria across jurisdictions. Hence, clusters. It emphasizes how only international plant
using patenting as the sole indicator of crop biotech biotechnology clusters are comparable across coun
nology innovation may miss important research work tries, while national clusters are only comparable across
carried out by scientists in countries where patenting regions in one country.
possibilities are limited.
The international clusters of plant biotechnology show
Second, while both patented inventions and scientific how geographically diverse is its innovation land
publication are used to measure innovative activities, scape. The clusters include the three main corridors
there are important differences. For example, inven of innovation, namely the U.S., Europe and East Asian
tions disclosed under patenting requirements may be countries, Japan and the Republic of Korea. They
closer to the commercialization stage than research further include India, Israel, China and Singapore in
published in scientific publications, which may be Asia, Australia in Oceania, and Argentina and Mexico
more “upstream” and science related.50 Moreover, most in Latin America and the Caribbean. But at the same
97
The distribution of agricultural biotech innovation has been relatively wide since the 2000s
Figure 4.4 Distribution of plant biotechnology innovation centers by patent filings (left)
and publication (right), 1998–2007 (top) and 2008–2017 (bottom)
PATENTS SCIENTIFIC PUBLICATIONS
Note: Size of bubbles corresponds to the relative volume of patent and scientific publications, respectively.
98
99
Agricultural biotechnology clusters are spread across the globe

Figure 4.5 Global distribution of plant biotechnology innovation clusters, 1970–2017
INTERNATIONAL PLANT BIOTECH CLUSTERS NATIONAL PLANT BIOTECH CLUSTERS
time, these international plant biotechnology clusters areas. However, field trials are conducted in rural areas,
mirror their related biotechnology field clusters. which may require some innovative activities to adapt
the genetically engineered crop to local agro-ecological
Not all countries with significant innovative activities conditions – to the local combinations of soil, landform
have international clusters. Brazil, for example, is a and climatic characteristics.52
leading developing economy with important innova
tive activities in the field, but with no international Figure 4.6 maps the international and national crop
cluster. The main reason for this is that its innovative biotechnology clusters against the crop areas (shaded
activities in plant biotechnology are spread across green) of the world for four regions: North America,
seven different regions that individually do not reach Europe, Asia and Latin America. Most of the interna
volume thresholds for the production of patents tional clusters tend to be in urban areas. In the U.S., for
and scientific articles (see Box 4.2). The Brazilian example, they are in such places as San Jose, Boston
Agricultural Research Corporation (EMBRAPA), its and New York City.
NARS, mandates that its research activities should
be scattered across its different research campuses However, there are some clusters that are adjacent
and not just concentrated in its main office in Brasilia. to crop areas. The location of these clusters is not an
accident. Most of these clusters are located in major
There are two notable insights from the global mapping agriculture-related universities, for example the U.S.
of plant biotechnology’s international and national Land-Grant colleges already mentioned. One notable
clusters. First, there is an urban–rural divide between example is Des Moines, Iowa, which is both a farming
the location of innovation centers and the farming the area and an international plant biotechnology cluster.
innovation is aimed at.51 Most innovation in the plant Des Moines is home to Iowa State University, a Land-
biotechnology industry is conceptualized, researched Grant university, and Pioneer Hi-Bred, one of the first
and developed in urban areas rather than in farming startups specializing in agriculture biotechnology.
100
Box 4.2 Identifying the international and For national clusters, all patent families, along with
national clusters of crop biotechnology scientific publication, are used. The patent families
include singletons, which are patents filed in the
Identifying international and national clusters of agri same IP office as the applicant’s residence and
cultural biotechnology involves at least three steps. nowhere else.
Step 1: Identify plant biotechnology patents and Classification as an international cluster is based
scientific publication on a global threshold combining foreign-oriented
Patents: use a combination of two international tech patents and scientific publication. National clusters
nological classification schemes, the International are based only on a country-specific threshold.
Patent Classification (IPC) and the Cooperative Patent
Classification (CPC) codes, and keywords to arrive Therefore, international clusters differ from national
at crop-specific agricultural biotechnology (see clusters in two main ways. First, international clusters
Technical Notes for the complete list of codes and are calculated by only considering foreign-oriented
keywords used in the search strategy). The cate patent families. National clusters, by contrast, are
gories for crop patents include: (i) crops’ genetic based on all patents filed by the residents of a coun
improvement; (ii) pest control in crops; (iii) soil fertil try, including both singletons and foreign-oriented
ity; and (iv) climate change. patents. Second, the threshold criterion determined
at the international level is based on the average
Scientific publication: use a combination of well- volume of patents and scientific articles attributed
known top journals in agricultural biotechnology, to one region, across the globe. The threshold at the
combined with relevant plant biotechnology-specific national level is measured by the average volume of
keywords (see Technical Notes for details). patents and scientific publication attributed to one
region in a country.
Step 2: Geocoding the addresses of inventors
and authors Only the international clusters are comparable
The addresses of authors of patented inventions across countries.
and scientific articles related to plant biotechnol
ogy are geocoded and mapped. Inventors’ resident Note: See Box 2.1 and Box 2.2 of Chapter 2. See also the glossary
list in the annex of WIPO (2018).
addresses, as listed in the patented documents, are
used, while for scientific publication, the authors’
addresses are usually not disclosed. Instead, the
location of the authors’ affiliation is employed.
Step 3: Differentiating between national and

international clusters For many developing economies, the relevant inter
Once the location of both patented inventions and national and national clusters of plant biotechnol
scientific publications are mapped out, two differ ogy are close to their respective NARS, which tend
ent thresholds are used to identify international and to be located in farming areas. The CIMMYT in
national clusters. For international clusters, only Texcoco is approximately one hour away from Mexico
foreign-oriented patent families are considered, City, while Buenos Aires is home to the National
in combination with published scientific articles. Agricultural Technology Institute (INTA) of Argentina.
These patents must be either filed in an IP office The International Crops Research Institute for the
different from the applicant’s residence or in at least Semi-Arid Tropics, a CGIAR institute, is located in
one foreign IP office, for example, at a national IP Patancheru, close to Hyderabad, in India, while the
office and a foreign IP office. Patents filed at an IRRI in Los Baños, the Philippines, is around one hour
international patent office, such as the European from Dasmarinas City. In Brazil, the national clusters
Patent Office or through the Patent Cooperation coincide with the locations of EMBRAPA centers. All
Treaty (PCT), are also considered as foreign-oriented of these NARS are within a 50-km radius of national
patent families. crop biotechnology clusters.
101
Innovation takes place far away from cropland

Figure 4.6 Location of plant biotechnology innovation centers and croplands
INTERNATIONAL PLANT BIOTECH CLUSTERS NATIONAL PLANT BIOTECH CLUSTERS
Source: WIPO based on PATSTAT, PCT and Web of Science data (see Technical Notes). Cropland data from Ramankutty et al. (2008).
Note: Green areas represent croplands and pastures circa 2000.
The presence of these agricultural institutions is likely human capital – the fact that knowledge, as opposed
to create regional ecosystems that are conducive to to information, does not transfer that easily between
startups as well as R&D facilities of companies in the locations – as well as the localized nature of knowledge
industry. Samad and Graff (2020) show that the single spillovers (see Chapter 1).
most important determinant of the number of inven
tions to come from a given region is the number of Second, as noted, most of the international plant
inventions that have come from that region in the past. biotechnology clusters are concentrated in metro
This relationship represents the “sticky” nature of fixed politan areas. Figure 4.7 plots the location of these
investments in regional knowledge infrastructure and international clusters alongside global innovation
102
Innovative activities tend to cluster, particularly in metropolitan areas

Figure 4.7 Worldwide distribution of innovation (GIHs, SNCs and
international plant biotechnology clusters)
GIH SNC AGBIO CLUSTER
hotspots (GIHs) and specialized niche clusters (SNCs), on patented inventions (left) and scientific publica
as defined in Chapter 2 of this report. tion (right) for 2010–2017. These links are based on
co-inventorship and co-authorship across regions.
Strong agglomerating forces determine where the inno The size of the bubbles in the figure represents the
vative regions of plant biotechnology are located. By volume of patented inventions (or scientific publication)
co-locating in regions where there are strong innovative in that particular cluster, while the thickness of the lines
activities, whether plant biotechnology specific or not, represents the frequency of the interactions between
researchers in both public and private sectors are able them. The colors of the bubbles indicate the countries
to benefit from the knowledge spillover (see Chapter 2). to which the clusters belong.
For example, they can profit from the presence of other
related innovating industries and specialized skilled The U.S., Canada, Europe, particularly Germany, France,
workers, some of which could be relevant and useful Netherlands, Denmark and the U.K., and countries in
and facilitate new technological advances in the crop East Asia (Japan, the Republic of Korea and China) are
biotechnology industry.53 home to most of the international clusters for patented
innovation in crop biotechnology.54 As in the case of
biotechnology, distance is not necessarily the main
4.3 The innovation network criterion for connecting to clusters.
of plant biotechnology
For example, inventors in the two largest international
The main innovation clusters in agricultural biotechnol clusters, San Jose and New York City (nearly 4,724 km
ogy are found, not surprisingly, in the leading countries apart), interact more frequently than San Jose with
that invest in agricultural R&D. San Diego (approximately 739 km apart). Inventors in
Rotterdam, the Netherlands, co-invent more frequently
Figure 4.8 provides rough illustrations of how the top with inventors in San Diego than with their compatriots
30 international clusters connect to one another, based in Eindhoven.
103
More openness in scientific publications than patenting? Collaboration

between plant biotechnology clusters are more frequent and denser
in scientific publications than in patenting activities
Figure 4.8 Linkages between the top 30 international biotechnology clusters
based on patent filings (left) and scientific publications (right), 2010–2017
Note: Size of the bubbles corresponds to the relative volume of patent and scientific publications, respectively.
Specialized researchers in agricultural biotechnology tend to come from the U.S.

Figure 4.9 Comparison of top 10 percent applicant–inventor ties of foreign-oriented patents,
1970–1999 (left) and 2000–2017 (right)
Applicant Inventor
China
Germany
Switzerland
France
Belgium Switzerland
Germany India
Canada
Austria
Other
U.S.
Denmark
Australia Belgium
Other
U.K. U.S.
Netherlands
Canada Netherlands
Spain
France Japan
Japan Australia
Israel U.K.
Denmark Italy
Note: Only patent families with foreign-orientation are shown. In addition, applicant-inventor ties where the applicant has a different residence than
the inventor are displayed.
104
The picture for international clusters based on published research avenues and hence applications for plant
scientific articles follows a similar pattern. However, the biotechnology. The adaptation of CRISPR-Cas9 is likely
size of the clusters and their interactions with one another to reinvigorate research on the genetic improvement
are more diverse and denser. The two biggest clusters of crops and livestock. Moreover, as this technol
based on publication are Beijing and Tokyo. U.S. clusters ogy becomes more affordable, it has the potential to
do not figure as prominently as they do for patenting. “democratize” innovation in agricultural biotechnology. 55
Combined with the rising role of developing economies
Nevertheless, the U.S. has by far the most international in such innovation, this latest advance could lead to the
clusters based on both measures of innovation: 16 clus global innovation network being more evenly dispersed.
ters using patents and eight using scientific publication. Clusters in different parts of the world could soon be
It is followed by Germany with three international clus making important contributions that will enhance food
ters, as measured by patents, and China with six inter security in an efficient and sustainable manner.
national clusters, as measured by scientific publication.
In addition, new applications of sensors and artificial
Both measures of internationally comparable crop intelligence to systematize the quantification of an
biotechnology clusters point to the U.S. as central to inno organism’s phenotype and physical traits could enable
vation in plant biotechnology. One reason for the U.S.’s much more powerful and precise connections to be
importance in international clusters of plant biotechnol drawn between genotype, genetic traits and phenotype
ogy is the quantity and quality of its specialized inventors than was previously possible. With the combined abili
and researchers. When looking at where most inventors ties to “read,” “write” and “edit” nucleotide sequences,
reside, especially when it is different to that of the patent new technological opportunities are possible for the
applicant, we see the overwhelming centrality of the U.S. genetic improvement of crops and livestock.
as the place to find crop biotechnology researchers.
The second development that may change the global
Figure 4.9 illustrates the location of researchers in plant innovation landscape and improve developing econo
biotechnology by exploiting the different locations of mies’ participation in the global innovation network is
the applicant (left) and the inventor (right) of a particular the recent shift in CGIAR’s stance on IP rights.56 In
patent. The left panel provides the links between appli the past, CGIAR had been committed to ensuring
cant–inventor pairs in the years 1970–1999, while the that its members’ work could be shared and would
right panel paints the picture for the years 2000–2017. be easily accessible to all; it had shied away from the
The lines connecting the applicant to the inventor are exclusionary properties of IP rights. This stance has
proxies for the strength of the relationship: the thicker changed. The CGIAR has recognized the importance
the line the more frequent the interaction. of collaborating with the private sector and has begun
using IP rights as an incentive for such collaboration
In both periods, many patent applicants outside the and partnerships and to encourage innovation.
U.S. search for U.S. researchers and scientists. The
fact that many of the important discoveries in agri Finally, in July 2018, the European Court of Justice (ECJ)
cultural biotechnology came from U.S. universities ruled that plants engineered using gene-editing tech
and public institutions is one reason why U.S. scien nologies, such as CRISPR-Cas9, would be subject to
tists and researchers are highly sought after. Another the same regulations as those applied to GMOs. The
explanation is that private companies in the U.S. were CRISPR-Cas9 technique changes a plant’s make-up,
often the first to invest strategically in the exploration its DNA, but without introducing any foreign mate
of commercial applications of biotechnology in plants. rial, and arguably might have been exempt from the
These factors combine to increase the weight of the regulations. However, the ECJ ruled that the technique
U.S. in the crop biotechnology innovation network. was still subject to the European Commission direc
tive. Scientists and researchers argue that the ruling
could result in a further exodus of plant biotech R&D
4.4 Future of plant biotechnology outside Europe. If proven true, then the ruling will
further change the innovation landscape and networks
Three new developments in plant biotechnology could of plant biotechnology.
transform the current global innovation network. Recent
breakthroughs in molecular biology are opening new
105
Notes
1 This chapter draws on Graff and 11 See Alvarez-Morales (2000). 24 For the list of regulations
Hamdan-Livramento (2019). 12 Patents are territorial in nature. and directive on GMOs
2 Other traditional ways include This means that a patent granted visit ec.europa.eu/food/plant/
hybridization as well as grafting. in one country or jurisdiction gmo/legislation_en
3 The United Nations Convention is not necessarily enforceable 25 The European countries included
on Biodiversity (CBD) definition. in another. Inventors who want in the figure are: EU-28 (except
It differs slightly from that of to ensure that their invention is for missing data from Malta,
the Biotechnology Innovation protected from imitation across Bulgaria and Poland). Portugal,
Organization (BIO), a major countries would have to file Spain and the U.K. are included
industry association. BIO defines patent applications for the same in the list of countries in Europe
biotechnology as “technology invention in those jurisdictions. still farming transgenic crops.
based on biology – [it] harnesses 13 Other forms of IP protection 26 USDA Foreign Agricultural
cellular and biomolecular on plants are plant varieties Service (2018).
processes to develop and plant patents (specific to 27 ISAAA (2017). In 2012, BASF
technologies and products that the U.S.). However, these two announced that it was closing
help improve our lives and the IP instruments are outside the its SunGene, its main plant
health of our planet” (www.bio. scope of this chapter and not biotechnology activity in
org/what-biotechnology). addressed here. Gatersleben, Germany, to
4 Graff et al. (2003). 14 Barton (2000). concentrate on the North and
5 FAO (2003). 15 See Eisenberg (1996) and Heller South American markets by 2013.
6 The term “agricultural and Eisenberg (1998). Visit www.sungene.de
biotechnology” differs from 16 Barton and Berger (2001). 28 See European Commission (2004,
“plant biotechnology” in that 17 Brennan (1980). 2009); National Research Council
the former refers to the general 18 Another instrument which (1987, 1998); FAO (2003, 2004).
industry, while the latter protects innovation in plants is 29 “CRISPR” stands for clustered
applies to a particular field the International Union for the regularly interspaced short
of agricultural biotechnology. Protection of New Varieties of palindromic repeats while
“Plant biotechnology” is Plants (UPOV) system of plant “Cas9” refers to CRISPR-
used interchangeably with varieties rights. This chapter associated protein 9.
“crop biotechnology.” does not touch on this right. 30 Wright (2012). See Alston et al.
7 The first licensed drug using 19 Figueiredo et al. (2019). See (2010) and Olmstead and Rhode
rDNA technology was the human Teece (1986) and Rothaermel (2011) on how these Land-Grant
insulin drug, produced by (2001) for importance of colleges have been useful for the
Genetech and licensed to Eli Lilly complementary assets. agriculture industry in the U.S.
and Company (Johnson, 1983). 20 See Eckerstorfer et al. (2019) 31 See Chapter 2 of FAO (1996)
8 Kenny (1988). for an overview of the different for further details on agro-
9 The proximity between humans existing regulatory frameworks. ecological conditions.
and animals. Humans fall under 21 See Glowka (2003) and 32 Byerlee and Fischer (2002) and
the category of mammals in the Komen (2012). FAO (2004).
animal kingdom, which allows for 22 See Graff and Hamdan- 33 Barry and Hosch (2000).
easier transition between human Livramento (2019), and Brenner 34 See FAO (2004), and Serageldin
and animal health. and Komen (1994). and Persley (2000).
10 Carrer et al. (2010). The terms 23 See WTO dispute settlement 35 Bijman and Tait (2002).
genetically engineered-, case DS291: European 36 Clancy et al. (2016).
genetically modified- and Communities — Measures 37 Komen and Persley (1993),
genetically improved-organisms Affecting the Approval Persley (2000) and Fukuda-
are used interchangeably and Marketing of Biotech Parr (2006).
throughout this chapter. Another Products (www.wto.org/english/ 38 Kalaitzandonakes and Bjornson
term that is used alongside these traptop_e/dispu_e/cases/e/ (1997) calculated the number
is transgenic crops. ds291_e.htm). of mergers, acquisitions and
106
strategic alliances between 46 Pinstrup-Andersen and contributions, are excluded.

startups and MNCs at 167 Cohen (2003). 50 See Griliches (1990).
between 1981 and 1985, and 801 47 See International Potato 51 Samad and Graff (2020) also
mergers between 1991 and 1996. Center (1995). find this urban–rural divide when
39 See Kalaitzandonakes (2000); 48 Co-applications refer to patent looking at the innovation centers
Fulton and Giannakas (2001); Tait applications where there are at of agriculture biotechnology
et al. (2002); and OECD (2018). least two listed applicants on regions in the U.S.
40 Howard (2015). the document. 52 See Graff and Hamdan-
41 See OECD (2018) and Fuglie 49 It is plausible that there are Livramento (2019) for more
et al. (2012). The OECD (2018, more regions that should be information on the procedure for
p. 104) reviewed the empirical included rather than those farming transgenic crops.
literature on concentration in displayed in this chapter. In other 53 Hermans et al. (2008).
the seed industry and impact on words, the regions covered 54 See annex in Graff and Hamdan-
innovation. The study concludes by scientific publication may Livramento (2019) for a detailed
that there is little evidence for the be an underestimate. This is list of the top 30 clusters by
adverse impact of concentration because identifying articles on patents and scientific articles
on innovation based on plant biotechnology is sensitive published, respectively.
historical data. to the method used. Here it is 55 See Mahfouz et al. (2014) and
42 Fukuda-Parr (2006). based on the top journals in plant Shwartz (2018).
43 Huang et al. (2002). biotechnology. Other journals 56 See CGIAR (2006, 2013).
44 Byerlee and Fischer (2002). that are not as well known, but
45 See Barton and Berger (2001). which may have equally relevant
107
References
Alston, J.M., M.A. Andersen, J.S. Bijman, J. and J. Tait (2002). Public CGIAR (2013) Implementation
James and P.G. Pardey (2010). policies influencing innovation in Guidelines for the CGIAR
Persistence Pays: US Agricultural the agrochemical, biotechnology Principles on the Management of
Productivity Growth and the and seed industries. Science and Intellectual Assets.
Benefits from Public R&D Spending, Public Policy, 29(4), 245–251. doi.
Natural Resource Management org/10.3152/147154302781780895 Clancy, M., K. Fuglie and P. Heisey
and Policy, Zilberman, D., R. Goetz (2016, November 10). U.S. Agricultural
and A. Garrido (eds), Vol. 34. New Brennan, A.A. (1980). Patentability R&D in an era of falling public funding.
York: Springer. of micro-organisms: Diamond v. Amber Waves. www.ers.usda.gov/
Chakrabarty, 100 S. Ct. 2204. Akron amber-waves/2016/november/
Alvarez-Morales, A. (2000). Mexico: Law Review, 14(2), 341–349. us-agricultural-rd-in-an-era-of-falling-
ensuring environmental safety while public-funding
benefiting from biotechnology. In Brenner, C., and J. Komen
Persley, G.J. and M.M. Lantin (eds), (1994). International Initiatives in Eckerstorfer, M.F., M. Engelhard,
Agricultural Biotechnology and the Biotechnology for Developing A. Heissenberger, S. Simon and H.
Poor: Proceedings of an International Country Agriculture: Promises and Teichmann (2019). Plants developed
Conference, Washington, D.C., 21–22 Problems. OECD Development by new genetic modification
October 1999. Washington, D.C.: Centre Working Papers, No. 100. techniques – comparison of existing
Consultative Group on International Paris: OECD Publishing. doi. regulatory frameworks in the EU
Agricultural Research. org/10.1787/257557587410 and non-EU countries. Frontiers in
Bioengineering and Biotechnology,
Babinard, J. (2001). A short history Brookes, G. (2018). The farm level 7. doi.org/10.3389/fbioe.2019.00026
of agricultural biotechnology. In economic and environmental
Nelson, G.C. (ed.), Genetically contribution of Intacta soybeans in Eisenberg, R.S. (1996). Intellectual
Modified Organisms in Agriculture. South America: the first five years. property issues in genomics. Trends
London: Academic Press, GM Crops & Food, 9(3), 140–151. doi. in Biotechnology, 14(8), 302–307. doi.
271–274. doi.org/10.1016/B978- org/10.1080/21645698.2018.1479560 org/10.1016/0167-7799(96)10040-8
012515422-2/50029-1
Byerlee, D. and K. Fischer (2002). European Commission (2004).
Barry, G. and R. Horsch (2000). Accessing modern science: Plants for the Future: A 2025 Vision
Evolving role for the public and private policy and institutional options for European Plant Biotechnology,
sector in agricultural biotechnology for agricultural biotechnology EUR 21359 EN. Brussels: European
in developing countries. In in developing countries. World Commission Directorate-General for
Persley, G.J. and M.M. Lantin (eds), Development, 30(6), 931–948. doi. Research Food Quality and Safety.
Agricultural Biotechnology and the org/10.1016/S0305-750X(02)00013-X
Poor. Washington D.C.: Consultative European Commission (2009).
Group on International Agricultural Carrer, H., A.L. Barbosa and D.A. Communication from the Commission
Research, 183–185. Ramiro (2010). Biotechnology in to the Council, the European
agriculture. Estudos Avançados, Parliament, the European Economic
Barton, J.H. (2000). Intellectual 24(70), 149–164. doi.org/10.1590/ and Social Committee and the
property, biotechnology, and S0103-40142010000300010 Committee of the Regions: a mid-term
international trade: two examples. assessment of implementing
World Trade Forum, 3, 1–15. CGIAR (2006). CGIAR Research the EC Biodiversity Action Plan.
Strategies for IPG in a Context of IPR: Journal of International Wildlife
Barton, J.H. and P. Berger (2001). Report and Recommendations Based Law & Policy, 12(1–2), 108–120. doi.
Patenting agriculture. Issues in on Three Studies. Rome: Consultative org/10.1080/13880290902938435
Science and Technology, 17(4). issues. Group on International Agricultural
org/barton Research. ispc.cgiar.org/sites/default/
files/ISPC_IPGStrategiesIPR.pdf
108
FAO (1996). Agro-Ecological Graff, G. and I. Hamdan-Livramento Huang, J., R. Hu, Q. Wang, J. Keeley,
Zoning Guidelines, FAO Soils (2019). The Global Roots of Innovation and J.F. Zepeda (2002). Agricultural
Bulletin No. 73. Rome: Food and in Plant Biotechnology. WIPO biotechnology development, policy
Agriculture Organization of the United Economic Research Working Paper and impact in China. Economic and
Nations. www.fao.org/3/w2962e/ No. 59. Geneva: World Intellectual Political Weekly, 37(27), 2756–2761.
w2962e00.htm#P-2 Property Organization.
International Potato Center (1995).
FAO (2003). World Agriculture: Graff, G., A. Heiman, C. Yarkin and Program Report 1993-1994. Lima,
Towards 2015/2030. An FAO D. Zilberman (2003). Privatization Peru: International Potato Center.
Perspective. Rome: Food and and Innovation in agricultural
Agriculture Organization of the biotechnology. ARE Update, 6(3), 5–7. ISAAA (2017). Global Status of
United Nations. Commercialized Biotech/GM Crops
Graff, G. and D. Zilberman (2007). in 2017: Biotech Crop Adoption
FAO (2004). The State of Food and The political economy of intellectual Surges as Economic Benefits
Agriculture 2003–2004. Agricultural property: re-examining European Accumulate in 22 Years (ISAAA Briefs
Biotechnology: Meeting the Needs Policy on plant biotechnology. No. 53). Ithaca, NY: International
of the Poor? Rome: Food and In J. Kesan (ed.), Intellectual Service for the Acquisition of Agri-
Agriculture Organization of the United Property Protection for Agricultural Biotech Applications.
Nations. www.fao.org/3/Y5160E/ Biotechnologies: Seeds of Change.
y5160e00.htm#TopOfPage Wallingford: CABI Press. Johnson, I. S. (1983). Human insulin
from recombinant DNA technology.
Figueiredo, L.H.M., A.G. Griliches, Z. (1990). Patent statistics Science, 219(4585), 632–637. doi.
Vasconcellos, G.S. Prado and as economic indicators: a survey. org/10.1126/science.6337396
M.F. Grossi-de-Sa (2019). An Journal of Economic Literature,
overview of intellectual property 28(4), 1661–1707. Kalaitzandonakes, N. G. and
within agricultural biotechnology B. Bjornson (1997). Vertical and
in Brazil. Biotechnology Research Heller, M.A. and R.S. Eisenberg horizontal coordination in the agro-
and Innovation, 3(1), 69–79. doi. (1998). Can patents deter innovation? biotechnology industry: evidence
org/10.1016/j.biori.2019.04.003 The anticommons in biomedical and implications. Journal of
research. Science, 280, 698–701. doi. Agricultural and Applied Economics,
Fuglie, K.O., J.L. King, P.W. Heisey org/10.1126/science.280.5364.698 29(1), 129–139. doi.org/10.1017/
and D.E Schimmelpfennig (2012). S1074070800029187
Rising concentration in agricultural Hermans, R., A. Löffler and S. Stern
input industries influences new (2008). Biotechnology. In J.T. Macher Kalaitzandonakes, N. G. (2000).
farm technologies. Amber Waves, and D.C. Mowery (eds), Innovation Agrobiotechnology and Competitive-
10(4). ageconsearch.umn.edu/ in Global Industries: U.S. Firms ness. American Journal of Agricultural
record/142404 Competing in a New World (Collected Economics, 82(5), 1224–1233.
Studies). Washington, D.C.: The
Fukuda-Parr, S. (ed.) (2006). The National Academies Press. doi. Kenny, M. (1988). Biotechnology: The
Gene Revolution: GM Crops and org/10.17226/12112 University-Industrial Complex. New
Unequal Development. London and Haven: Yale University Press.
Sterling: Earthscan. Howard, P.H. (2015). Intellectual
property and consolidation in the Klümper, W. and M. Qaim (2014). A
Fulton, M. and K. Giannakas (2001). seed industry. Crop Science, 55 meta-analysis of the impacts of
Agricultural biotechnology and (November–December). www. genetically modified crops. PLOS
industry structure. AgBioForum, apbrebes.org/files/seeds/files/ ONE, 9(11), e111629. doi.org/10.1371/
4(2), 137–151. Howard_seed_industry_patents_ journal.pone.0111629
concentration_2015.pdf
Glowka, L. (2003). Law and Modern
Biotechnology, FAO Legislative
Study. Rome: Food and Agriculture
Organization of the United Nations.
109
Komen, J. (2012). The emerging Persley, G.J. (2000). Agricultural Samad, G. and G.D. Graff (2020). The
international regulatory framework for biotechnology and the poor: urban concentration of innovation
biotechnology. GM Crops & Food, 3(1), Promethean science. In Persley, G.J. and entrepreneurship in agricultural
78–84. doi.org/10.4161/gmcr.19363 and M.M. Lantin (eds), Agricultural and natural resource industries. In
Biotechnology and the Poor. Iftikhar, M.N., J.B. Justice and D.B.
Komen, J. and G.J. Persley (1993). Washington D.C.: Consultative Audretsch (eds), Urban Studies and
Agricultural Biotechnology in Group on International Agricultural Entrepreneurship. Cham: Springer
Developing Countries: A Cross- Research, 3–21. International Publishing, 91–116. doi.
Country Review (ISNAR Research org/10.1007/978-3-030-15164-5_6
Report No. 2). The Hague: Persley, G.J. and J.N. Siedow (1999).
International Service for National Applications of Biotechnology to Serageldin, I. and G.J. Persley
Agricultural Research. Crops: Benefits and Risks. Council for (2000). Promethean Science:
Agricultural Science and Technology Agricultural Biotechnology, the
Mahfouz, M.M., A. Piatek and C.N. Issue Paper No. 12. Environment, and the Poor.
Stewart (2014). Genome engineering Washington, D.C.: Consultative
via TALENs and CRISPR/Cas9 Pinstrup-Andersen, P. and M.J. Group on International Agricultural
systems: challengesand perspectives. Cohen (2003). Biotechnology and the Research. documents.worldbank.org/
Plant Biotechnology Journal, 12(8), CGIAR. In Plenderleith, K. and P. De curated/en/698501468739325409/
1006–1014. doi.org/10.1111/pbi.12256 Meyer (eds), Sustainable Agriculture Promethean-science-agricultural-
in the New Millennium: The Impact biotechnology-the-environment-and-
National Research Council (1987). of Biotechnology on Developing the-poor
Agricultural Biotechnology: Strategies Countries. Brussels: Friends of the
for National Competitiveness Earth Europe. Sheldon, I.M. (2004). Europe’s
(Report of the Committee on a regulation of agricultural
National Strategy for Biotechnology Pray, C.E. and A. Naseem (2003). biotechnology: precaution or trade
in Agriculture). Washington, D.C.: The Economics of Agricultural distortion? Journal of Agricultural
National Research Council. Biotechnology Research. ESA & Food Industrial Organization,
Working Paper No. 03-07. Rome: The 2(2), 1–28.
National Research Council (1998). Food and Agriculture Organization
Designing an Agricultural Genome of the United Nations. www.fao.org/ Shwartz, M. (2018). CRISPR is a
Program. Report of the Board on tempref/docrep/fao/007/ae040e/ gene-editing tool that’s revolutionary,
Biology and Board on Agriculture. ae040e00.pdf though not without risk. Stanford
Washington, D.C.: The National Medicine, Winter. stanmed.stanford.
Academy Press. Ramankutty, N., A.T. Evan, edu/2018winter/CRISPR-for-gene-
C. Monfreda and J.A. Foley editing-is-revolutionary-but-it-comes-
Organisation for Economic (2008). Farming the planet: 1. with-risks.html
Cooperation and Development Geographic distribution of global
(OECD) (2018). Concentration in Seed agricultural lands in the year Tait, J., J. Chataway and D.
Markets: Potential Effects and Policy 2000. Global Biogeochemical Wield (2002). The life science
Responses. Paris: OECD Publishing. Cycles, 22(1), GB1003. doi. industry sector: evolution of agro-
org/10.1029/2007GB002952 biotechnology in Europe. Science and
Olmstead, A.L. and P.W. Rhode Public Policy, 29(4), 253–258.
(2011). Adapting North American Rothaermel, F.T. (2001).
wheat production to climatic Complementary assets, strategic Teece, D.J. (1986). Profiting from
challenges, 1839–2009. Proceedings alliances, and the incumbent’s technological innovation: implications
of the National Academy of Sciences, advantage: an empirical study of for integration, collaboration, licensing
108(2), 480–485. doi.org/10.1073/ industry and firm effects in the and public policy. Research Policy,
pnas.1008279108 biopharmaceutical industry. Research 15(6), 285–305. doi.org/10.1016/0048-
Policy, 30(8), 1235–1251. 7333(86)90027-2
110
United States Department of WIPO (2018). World Intellectual Zilberman, D., C. Yarkin and
Agriculture (USDA) Foreign Property Indicators. Geneva: World A. Heiman (1997). Agricultural
Agricultural Service (2018). EU-28: Intellectual Property Organization. Biotechnology: Economic and
Agricultural Biotechnology Annual, International Implications. Paper
GAIN Report No. FR1827. Global Wright, B.D. (2012). Grand missions presented at the International
Agricultural Information Network. gain. of agricultural innovation. Research Agricultural Economics Association,
fas.usda.gov/Recent%20GAIN%20 Policy, 41(10), 1716–1728. doi. Sacramento, California, August. are.
Publications/Agricultural%20 org/10.1016/j.respol.2012.04.021 berkeley.edu/~zilber11/yark.pdf
Biotechnology%20Annual_Paris_
EU-28_12-14-2018.pdf
111
Innovation is becoming
more collaborative as
technology becomes
more complex. For large
multi-skilled teams to
thrive, knowledge needs
to be able to flow freely
across borders.
Chapter 5
Policy perspectives:
the case for openness
Innovation has always spanned countries and As documented in this report, innovation today is
continents. At the turn of the 20th century, both highly localized and international at the same
the Wright brothers in the United States of time. Different agglomeration forces have favored the
America (U.S.) and Alberto Santos-Dumont formation of innovation hotspots that typically fall within
in Brazil invented the first airplanes to fly large metropolitan regions. A limited set of hotspots
successfully. Yet, the development of the modern lead the way and are at the center of global innovation
airplane owes much to scientific advances in networks. Various formal and informal links connect the
Europe that explained why heavier-than-air nodes of these networks, with multinational companies
machines could fly.1 The development and playing a key role within them. Evidence from patent
dissemination of the agricultural technologies and scientific publication records suggests that the
that unleashed the green revolution after the cross-border dimension of these links has increased
Second World War relied on partnerships over the past decades.
between the Ford and Rockefeller Foundations
in the U.S. and a large number of agricultural The growing internationalization of innovation owes
research institutes in developing economies.2 much to technology itself. Advances in information
Tim Berners-Lee invented the World Wide and communication technologies (ICTs), in particular,
Web at the European Organization for Nuclear have fueled the flow of knowledge over long distances.
Research (CERN) – a research consortium on the Crucially, however, the growth of global innovation
Franco–Swiss border sponsored by 23 (mostly) networks has relied on policies favoring openness
European countries.3 and international cooperation. Such an environment
of openness and cooperation should not be taken
for granted – especially, as public perceptions have
become more skeptical as to the benefits of globaliza
tion in general in recent years.
This closing chapter therefore reviews the case for

openness in the pursuit of innovation. It does so primar
ily from an economic perspective. At times, whether
and how to partner with foreign innovators involves
questions of national security, which go beyond the
scope of the chapter’s discussion.
5.1 The economics of openness
Openness of national innovation systems entails the

free exchange of knowledge between economies.
Knowledge may flow across borders when research
ers communicate with each other, or when they read
scientific journals and patent documents published
abroad. It may also occur through international trade,
when knowledge is embedded in goods and services;
and it may occur through migration, when it is embed
ded in people.
113
How do restrictions on the international flow of knowl The notion of differentiated varieties, in turn, finds its
edge affect national economies and the world as a equivalence in the specialization of different innovation
whole? The answer depends crucially on which knowl clusters around the world. For example, there are many
edge flows the restrictions target, the capabilities of innovation clusters focusing on medical technology,
national innovation systems, patterns of production with each offering specialized knowledge not avail
and employment, and the nature of the economic able elsewhere. This generates bidirectional knowl
growth process. While not offering a definite conclu edge flows, even among otherwise similar economies.
sion, the economic literature offers some guidance on Global innovation networks act as a broker for such
the effects of knowledge flow restrictions, which this knowledge flows.
section strives to summarize.
Trade theory holds that there are mutual gains from
comparative advantage-based trade. These gains
Gains from specialization take the form of increased economic efficiency and
a wider variety of goods available to businesses and
One simplified way to approach the question is to view end-consumers. Given the highly specialized nature of
knowledge like any other good. Just as the production innovative output, the variety effect seems particularly
of cars requires capital and labor inputs, so does the important to knowledge trade.
production of new knowledge through innovation.
Notwithstanding these mutual gains, trade theory
Restricting international knowledge flows then affects also holds that open trade affects the distribution of
how economies allocate resources toward different incomes within economies. Such distributional effects
production activities. Viewed this way, the traditional are stronger if differences in capital and labor endow
predictions of international trade theory apply. Above ments give rise to international trade. In other words,
all, openness leads to production and trade patterns they are more important for trade between dissimilar
that allow economies to specialize based on their economies – notably between economies at different
comparative advantage. Trade economists usually levels of development. As will be further discussed
consider two forces that give rise to specialization:4 below, these distributional effects matter for policy.
• Differences in factor endowments. An economy

richly endowed with capital will specialize in and Innovation as a global public good
export goods that are capital-intensive to produce.
Conversely, an economy richly endowed with labor Viewing knowledge like any other good helps explain
will specialize in and export goods that are labor- important aspects of the global innovation landscape.
intensive to produce. However, it is a highly simplified view that fails to
• Differentiated varieties and economies of scale. account for the unique characteristics of knowledge
Where goods come in differentiated varieties – say, production and knowledge consumption.
different car brands – and production of those
varieties entails economies of scale, economies Above all, knowledge has attributes of what economists
will specialize in and export some varieties and refer to as a public good: many people can use it at the
import others. same time, without diminishing the use of the knowl
edge by those who produce it.6 For example, the basic
These predictions can shed light on important facets science behind artificial intelligence emerged from a
of the global geography of innovation. Innovation limited number of scientific organizations, yet a large
requires highly skilled labor, which explains why most number of innovations employ this science for a wide
innovative activity takes place in high-income countries variety of applications around the world.7
where such labor is relatively abundant. At the same
time, the decision by multinational companies to locate In practice, there are limits to how widely knowledge
some research and development (R&D) activities in can be shared. In fact, a central tenet of economic
certain developing economies, such as China and geography research is that knowledge does not flow
India, reflects the availability of highly skilled labor at freely within and across economies; knowledge flows
lower wages – fully in line with patterns of compara have distinctive geographical patterns and biases.8 One
tive advantage.5 reason is that absorbing and applying cutting-edge
114
5 Policy perspectives: the case for openness
knowledge often requires highly specialized skills that are successful at innovating will, in the long term,
that are in scarce supply.9 Moreover, for some forms experience faster overall economic growth than those
of knowledge to flow, human interaction is required, that fail to do so. However, this does not necessar
which is precisely a key reason for innovative activity ily mean that one economy’s success constrains
to agglomerate (see Chapter 1).10 another economy from being equally success
ful. In fact, the public good nature of knowledge
Yet, to the extent that knowledge lives up to its public suggests that innovation can contribute to productivity
good potential, does this change the case for open growth everywhere.
ness? In fact, it strengthens it. If knowledge outflows
generate economic benefits abroad without diminishing Notwithstanding this general optimism, as national
the use of knowledge at home, there are bound to be innovation performance shapes patterns of production
mutual gains from openness. and trade, it is conceivable for one economy to end up
specializing in activities that put it on a permanently
faster or slower growth path. Strategically restricting
Innovation and growth trade and knowledge flows could then tilt production
patterns in such a way as to favor faster growth at
Innovation differs in another important way from other home. Box 5.1 summarizes theoretical research that
goods produced in the economy. Through innova identifies the conditions in which such “zero-sum”
tion, companies can create a competitive edge over outcomes can arise.
their rivals. A successful innovator can gain market
share at the expense of a company that fails to be Whether such conditions prevail in practice is ulti
cutting-edge. Competition based on innovation, in mately an empirical question. Rigorously answering it
turn, drives productivity enhancements and long-term is not easy, given that one does not know how differ
economic growth. ent economies would fare under different trade and
knowledge-flow policies. However, one can look at the
As companies compete on the global stage, commen actual growth experience of economies around the
tators have applied the same logic to economies as a world over the past decades. One important pattern is
whole. Accordingly, those economies that are success that today’s high-income economies have experienced
ful at innovating grow faster at the expense of econo remarkably similar growth over the past 40 years.
mies that do not innovate successfully.11 In such a zero- Before 1980, per capita incomes of poorer high-income
sum world, restricting knowledge outflows would help economies saw faster growth than those of richer high-
economies retain an innovative edge and avoid “falling income economies. But this convergence process
behind” other successfully innovating economies. eventually slowed (Figure 5.1). While differences in per
capita incomes persist, the most advanced economies
At the outset, the international economics literature have grown largely at a similar pace since the 1990s
would dismiss such “simplistic” zero-sum scenarios. (Figure 5.2). This may suggest that new technologies
Economies as a whole differ from companies in impor have spread seamlessly across the set of economies
tant ways. For one, economies as a whole cannot go already at the technology frontier and they have stimu
bankrupt. If companies in a particular sector exit the lated growth in comparable magnitudes.
market or lose market share due to foreign competition,
they free up labor and capital that can be deployed Beyond the group of high-income economies, the
elsewhere in the economy. growth experience has been mixed. For a long time,
incomes across the world diverged.12 In 1870, the
The reverse happens in sectors gaining international gross domestic product (GDP) per capita of the richest
market share – they attract labor and capital from else economy was around 10 times that of the poorest one;
where in the economy. In addition, faster productivity by 2008 this gap had widened to a factor of 126.13 For
growth in successfully innovating economies enlarges a very long time, poorer economies did not grow any
their size and increases demand for foreign products. faster than richer ones. More recent data – starting
from the 1990s – suggest a reversal of this trend, with
Overall, innovation leads to adjustments in prices, incomes converging across economies. In other words,
wages and exchange rates, which prompt shifts in since the 1990s, poorer economies have, on average,
production and trade patterns. Clearly, economies grown faster than richer ones.14
115
In the presence of such forces, strategic trade

Box 5.1 Theoretical foundations of and related policies could well reshape patterns of
strategic trade policy production and alter an economy’s growth path. In
practice, successfully implementing such policies is
A branch of trade theory in the 1980s and early 1990s difficult. The choice of policy instruments depends
was devoted to analyzing the circumstances in which critically on initial conditions, the evolving nature of
departures from free-trade policies may be welfare competition and technological opportunities. Given
enhancing. Many underlying models focused on that the future path of technology and its implica
imperfectly competitive markets and trade policies tions for markets are highly uncertain, choosing
that might increase the share of excess economic the right policy mix in a forward-looking way is a
profits flowing to the domestic economies.15 Some formidable challenge.
more complex theories also accounted for the
role of innovation in driving long-term growth. The
book by Gene Grossman and Elhanan Helpman
(1991) provides the most detailed treatment of these
latter theories.
Notwithstanding this trend reversal, average conver
In relevant models, firms invest in R&D with the gence does not mean universal or automatic conver
prospect of reaping economic rents in imperfectly gence. Some poorer economies have been more
competitive product markets. Competitive market successful at catching up to the richer ones than
forces, in turn, sustain incentives to continuously others. Developing countries in East Asia and, more
invest in R&D, thereby generating the productivity recently, India, have been particularly successful
gains that sustain growth in the long run. Mindful at doing so. Given their central role in the growth
that companies compete in a global arena, the process, knowledge flows and innovation must be
models then analyze the interdependence of growth part of the explanation behind these trends. However,
processes in different countries. which precise structural forces and economic policies
have favored catch-up growth remains the subject of
The predictions stemming from these models considerable debate.16 A pessimistic view is that the
confirm, first of all, the general optimism expressed historical concentration of innovative activities in a
in the text: global interactions generate forces that limited set of economies and the strong agglomera
accelerate growth in every country. But they also tion forces associated with such activities reinforce a
point to reasons why this may not always be the global core–periphery division. Even if policies do not
case. For example: restrict knowledge flows, this division fosters diverg
ing development paths. A more optimistic view is that
• Suppose that an economy has a comparative innovation eventually spreads beyond the core group
disadvantage in research due to limited high- of innovators; with the right policies, economies in the
skilled labor. Integration with the rest of the world periphery can absorb foreign knowledge and catch up.
could then lead it to specialize in more stagnant
activities, with overall output growing more slowly. In conclusion, the economic literature offers good
• Suppose that knowledge does not easily flow reasons why openness is bound to be beneficial in
across borders, because it is difficult to reverse the pursuit of innovation. Theoretically, there may
engineer or it requires critical skills not available well be circumstances in which strategic restrictions
in recipient countries, as described in the text. on trade and knowledge flows could alter the growth
Integration may then lead economies that are paths of economies. However, it is difficult to trans
small in size – or that historically have conducted late this theoretical possibility into concrete policy
little research – to specialize in manufacturing proposals. As pointed out in Box 5.1, adopting the
activities, preventing the onset of innovative right policy instruments in a forward-looking way is
activity. In fact, small differences in initial condi a formidable challenge. Practically, it may be difficult
tions between economies can lead to perpetual to prevent knowledge from flowing abroad, without at
differences in productivity growth. the same time restricting knowledge circulating within
economies. In addition, one economy’s policy choices
may prompt policy responses from other economies.
116
High-income economies Such policy reciprocity may well undermine the case
grow at a similar pace for strategically limiting openness. Finally, the growth
experience of high-income economies over the past
Figure 5.1 Gini coefficient, real GDP per
Gini coefficient, real GDP per capita, decades
Real GDPsuggests
per houran overall
worked, positive-sum
relative to U.S. impact of
capita, group
group of high-income
of high economies
income economies new technologies.
0.30 1.4
1.25.2 Openness in an age of

0.25
falling R&D productivity
1.0
0.20 The case for openness becomes even stronger when
0.8considering the context in which innovation takes
0.15
place today. Continuously pushing the technological
0.6frontier is becoming exceedingly difficult. Evidence
suggests that achieving the same level of technological
0.10
0.4progress as in the past requires more and more R&D
effort. For example, Gordon Moore – the co-founder
0.05 0.2of Intel – famously predicted in 1975 that the number
of transistors on a computer chip would double every
0.0
two years. What came to be known as Moore’s Law has
0.00
1953 1963 1973 1983 1993 2003 2013 roughly
1950 held
1960 up until
1970 today.
1980 Notably
1990 however,
2000 to double
2010
chip density today requires 18 times more researchers

AUSTRALIA CANADA SWITZERLAND GERMANY
thanFRANCE
it did in theUK
early 1970s. 17
ITALY JAPAN
per capita, Real GDP per hour worked, relative to U.S.
Source: Figure
Penn World5.2 Real
Table, GDP
version per hour
9.0, available worked,
at www.ggdc.net/pwt
Note: The Gini coefficient measures the distribution of incomes on a scale of 0 to 1, with theOther
lower thefields of technology
value, the show
greater the equality. GDPsimilar
per capitasigns
ratios of slowing
relative
in right panel
1.4 are to
based U.S.
on constant 2011 U.S. dollar real GDP figures, with 1.0 representing parity with the U.S. Values greater than 1.0 mean that a country's
GDP per hour worked exceeds that of the U.S. The group of high-income countries includesR&D productivity:
Australia, it takes
Austria, Belgium, Canada,multiple timesFrance,
Denmark, Finland, as much medi
Germany, Israel, Ireland, Italy,Japan, Netherlands, Norway, New Zealand, Republic of Korea, Spain, Sweden, Switzerland, United Kingdom, U.S.
cal R&D to achieve similar increases in life expectancy
1.2 as in the past; investments in agricultural R&D have
grown more rapidly than increases in agricultural crop
1.0 yields.18 More generally, most high-income economies
have seen a gradual decline in the growth of economic
0.8 productivity over the last half century. Economist
Robert Gordon has prominently attributed this decline
0.6 to innovations of the recent past boosting productivity
growth by less than innovations of the more distant
0.4 past.19 In particular, he argues that the innovations asso
ciated with the second industrial revolution supported
0.2 fast productivity growth in high-income economies
until the 1970s; the innovations associated with the
0.0 third (digital) industrial revolution have not been able
1950 1960 1970 1980 1990 2000 2010
to sustain such fast productivity growth.
AUSTRALIA CANADA SWITZERLAND GERMANY
AUSTRALIA CANADA SWITZERLAND
1993 2003 2013 FRANCE UK ITALY JAPAN
GERMANY FRANCE U.K. Policies cannot alter opportunities for technological
t www.ggdc.net/pwt progress. However, policies shape to what extent those
n of incomes on a scale of 0 to 1, with theITALY JAPAN
lower the value, the greater the equality. GDP per capita ratios
real GDP figures, with 1.0 representing parity with the U.S. Values greater than 1.0 mean that a country's opportunities are realized. They determine how much
e group of high-income countries includes Australia, Austria, Belgium, Canada, Denmark, Finland, France,
Note: Spain,
Norway, New Zealand, Republic of Korea, The Gini coefficient
Sweden, measures
Switzerland, theKingdom,
United distribution
U.S.of incomes on a resources are invested in R&D, how R&D is performed
scale of 0 to 1; the lower the value, the greater the equality. GDP per and how innovations find their way into the economy.
capita ratios in figure 5.2 are based on constant 2011 U.S. dollar real
GDP figures, with 1.0 representing parity with the U.S. Values greater Falling R&D productivity calls for constantly increasing
than 1.0 mean that a country’s GDP per hour worked exceeds that of
investments in innovation – both scientific research and
the U.S. The group of high-income countries includes Australia, Austria,
Belgium, Canada, Denmark, Finland, France, Germany, Ireland, Israel, applied R&D. It also calls for collaboration and open
Italy, Japan, Netherlands, New Zealand, Norway, Republic of Korea,
Spain, Sweden, Switzerland, U.K., U.S.
ness. Finding solutions to increasingly complex techno
Source: Penn World Table, version 9.0, available at www.ggdc.net/pwt logical problems requires larger teams of researchers
117
(see Chapter 2) and greater specialization in research. and the Hague System – facilitate the acquisition of IP
Openness and international collaboration promote rights in many countries by filing a single international
such specialization and can thus help slow declining application; the ultimate grant of IP rights remains a
R&D productivity. national decision.
For openness to work, policymakers need to go beyond Finally, governments can pool resources and fund
simply dismantling border barriers. There is an impor large-scale scientific projects that go beyond the
tant role for international cooperation to support open envelopes of national budgets or require technical
ness. Equally important, policymakers need to address knowledge available in different countries. CERN –
regional imbalances that openness may partly foster. mentioned at the outset of this chapter – is a good
The final part of this chapter looks at these two criti example of such cooperation. The International Space
cal dimensions. Station is another one. It is a joint project between the
national space agencies of Canada, Japan, the Russian
Federation and the U.S., as well as the European Space
Fostering international cooperation Agency. Launched in 1998, it has hosted more than 200
visitors from 18 different countries.20
International cooperation in relation to innovation has
many dimensions. An important one is to promote
incentives for innovation investments that reflect the Addressing regional imbalances
demands and size of the global economy. Setting
international rules for the protection of intellectual As discussed in Chapter 1, one worrying trend of
property (IP) rights serves this purpose. In practice, the past few decades is the increasing inter-regional
international treaties on IP typically establish the prin polarization of incomes, innovative activity, high-skilled
ciple of nondiscrimination, namely that national laws employment and wages within countries. Up to the
treat domestic and foreign IP owners equally. They 1980s, most high-income economies saw a steady
also set certain standards for the protection of differ convergence of incomes across regions.21 Poor regions
ent types of IP – for example, which inventions should of countries caught up with rich ones. Since then, inter-
be eligible for patent protection or how long copyright regional convergence has slowed and, in some cases, it
should last. At the same time, these standards do not has even reversed. In the U.S., the convergence process
fully harmonize IP protection across the world and slowed markedly starting in the 1990s.22 European
leave room for national policies to tailor IP protection economies have similarly seen slowing regional conver
to national needs. gence and, since the onset of the great recession in
2008, outright divergence. A few “champion” regions
A second important dimension is to promote the ease within European economies with already high levels
of doing business internationally. Innovating compa of income have seen substantially faster growth than
nies and knowledge workers face a variety of regula many of the poorer regions.23
tory measures when operating in markets around the
world. Promoting the compatibility of national regula There are many reasons for the polarization of econom
tory systems can help reduce the costs of regulatory ic activities within countries. The declining impor
compliance. For instance, some level of recognition of tance of agriculture and mining activities in economic
foreign regulatory standards can reduce the wasteful output has long favored a gravitational pull toward big
duplication of product testing and associated paper cities. In a knowledge-based and services-dominated
work, without necessarily compromising regulatory economy, businesses have strong incentives to locate
objectives. Recognition of foreign qualifications in line within large metropolitan areas. Openness arguably
with domestic standards can help facilitate the interna strengthens the gravitational pull toward champion
tional mobility of knowledge workers. Regular dialogues regions. The most vibrant innovation hotspots, which
between national regulatory agencies underpin such are embedded in global innovation networks, tend to
recognition frameworks. Similarly, setting technical be located in what already are the richest metropolitan
standards at the international level can avoid the costly agglomerations within countries. Their international
adaptation of products to different markets. In the area success reinforces their domestic lead. As described
of IP, WIPO’s international filing treaties – in particular, in Chapter 1, successful innovation agglomerations
the Patent Cooperation Treaty, the Madrid System may also see diverging incomes within them, with
118
fast growth of high-skilled jobs placing pressure on Box 5.2 Israel’s thriving innovation system:
disposable income in low-skilled occupations. Israel startup nation or startup region?
offers a good example of how burgeoning innovation
activities have raised concerns about a dual track Israel has a thriving innovation economy. Relative to
economy (see Box 5.2). the size of its GDP, no other country spends more on
R&D and attracts more venture capital investments.
Addressing such rising regional imbalances is one of Most of the world’s leading technology companies
the most difficult challenges for policymakers. Trying have established R&D centers in Israel to draw on
to reverse the gravitational pull of successful regions the skills and experience available in the country’s
may be neither feasible nor desirable. Restricting dynamic research community. In many fields –
participation in global innovation networks, in particu notably cybersecurity – Israeli companies set the
lar, would undercut an economy’s ability to generate trend. Its lively startup scene has earned Israel the
cutting-edge innovations. In any case, openness is but nickname “Startup Nation.”
one contributing factor to regional imbalances.24 The
long-term structural transformation of economic activity Israel’s vibrant innovation economy has been a
is arguably the fundamental driving force behind such key driving force behind the growth of the overall
imbalances. Internal migration from lagging to thriving economy. From 2008 to 2018, Israel’s economy
regions only offers a partial solution to regional diver grew by an average annual rate of 3.5 percent –
gence. Individuals may not have the capacity or willing again, far surpassing most developed economies.27
ness to move. High housing prices in thriving regions Unemployment fell to a record low of 4 percent
alone pose a significant barrier to internal migration.25 in 2018.28
Policy can play an important role in supporting regions Yet, the nickname masks the high geographical
whose fortunes have fallen behind. Development concentration of innovation activity in Israel. The
support for weaker regions has, of course, a long Tel Aviv metropolitan area stands out as the clear
history, with mixed success. A full review of historical champion region. It accounts for 77 percent of all
policy initiatives is beyond the scope of this report. startups and 60 percent of all high-tech jobs.29 It
Nonetheless, recent research points to a few consid hosts more than half of Israel’s inventors listed in
erations that are important when designing regional patent applications (see Figure 5.3).
support policies:26
Wages in the peripheral regions are around 35
• Ideally, regional development strategies should seek percent lower than in Central Israel. Tel Aviv’s domi
to build on existing capabilities and advantages nance has even intensified in recent years. The region
of regions and aim at amplifying them through was responsible for more than two-thirds of the
investments in infrastructure, education and tech growth in high-tech employees between 2015 and
nology. Existing capabilities and advantages can 2017.30 Tel Aviv is also highly connected to leading
take the form of relatively cheap land and labor innovation hotspots around the world, offering, for
and prevailing industrial capabilities, as well as example, nonstop flights to San Francisco.
reputational assets.
• Policy formulation should identify the key barriers As in other global innovation hotspots, Tel Aviv has
toward growing existing capabilities and rely on the seen rising concerns that the expansion of technol
input of all relevant local stakeholders. ogy companies is driving up housing prices and
• Resulting development policies should undergo widening income disparities.31
regular evaluation. The resulting evidence should
guide the adaptation of future policies. The Government of Israel recognizes that the gravi
tational pull of the Tel Aviv region reflects relative
While not reversing the gravitational pull of successful regional advantages and natural agglomeration
regions, such policies can ensure that innovation-driven forces. Yet it also realizes that this regional imbal
growth benefits economies as a whole. As such, they ance creates economic and social challenges. As
critically underpin the openness of national innova a result, Israel’s Innovation Authority has adopted
tion systems. a Strategy for an Innovation-Driven Economy in
the Periphery.
119
This strategy has four central pillars:32 The greater Tel Aviv area hosts
• Connecting human capital in the periphery to most of Israel’s inventors
leading high-tech companies;
Figure 5.3 Heat map of inventors listed
• Promoting technological innovation in the
in patent applications, 2008–2018
periphery in the manufacturing, agriculture and
food sectors;
• Encouraging entrepreneurship that draws on local
academic institutions and other sources of home-
grown knowledge and industrial expertise; and
• Strengthening the high-tech ecosystem in
those regions – namely, Haifa, Jerusalem and
Beersheba – that have the essential foundations
for such an ecosystem.
These pillars seek both to reduce a growing shortage

of high-skilled workers in the innovation economy
and to promote the development of regions that
currently are lagging to produce more balanced
national growth.
120
Notes
1 See WIPO (2015). 7 See WIPO (2019). 22 See Ganong and Shoag (2017).
2 See the case study on 8 See Crescenzi et al. (2019). 23 See Alcidi et al. (2018).
agricultural biotechnology 9 See Cohen and Levinthal 24 In reviewing two decades of
in Chapter 4. (1989) for an early contribution research, Helpman (2018)
3 Visit home.cern. Israel is the only on the importance of concludes that globalization is
CERN member from outside absorptive capacity. responsible for only a small rise
of Europe. 10 See von Hippel (1994). in inequality within nations.
4 See Krugman et al. (2018). 11 Such arguments first became 25 See Ganong and Shoag (2017).
Differences in productivity levels prominent in the 1980s 26 See Foray (2015) and Rodríguez-
between countries are a third when the rapid growth of Pose (2018).
force driving specialization. East Asian economies was 27 Based on constant 2010 U.S.
5 Differences in factor endowments perceived to threaten the dollar GDP values, as reported
can also explain patterns of technological dominance of by the World Bank.
international migration. Thus, Western economies (see, e.g., 28 As per International Labour
highly skilled workers – say Tyson, 1984). Organization country profile for
software engineers from India 12 Pritchett (1997) famously Israel.
– tend to move to high-income characterized the long-term 29 See Israel Innovation Authority
economies where they are paid historical trend as “divergence, (2019).
higher wages (see Krugman big time.” 30 See Israel Innovation Authority
et al., 2018). Admittedly, lower 13 See WIPO (2015). (2019).
wages of R&D personnel are but 14 See Patel et al. (2018). 31 See Srivastava (2018).
one motivation for multinational 15 See Brander and Spencer (1985) 32 See Israel Innovation Authority
companies to locate R&D for a seminal contribution. (2019).
activities in developing 16 See WIPO (2015).
economies; the growth potential 17 See Bloom et al. (2019).
of local markets is often another 18 See Bloom et al. (2019). The
important factor (see Thursby authors also document declining
and Thursby, 2006). R&D productivity when analyzing
6 Noble prize-winning economist firm-level data across the
Kenneth Arrow first observed the U.S. economy. In addition,
public good nature of knowledge they consider and reject the
(Arrow, 1962). In addition to possibility that the emergence of
being non-rival in consumption, new technologies compensates
knowledge producers cannot – for declining R&D productivity in
without intellectual property (IP) existing technologies.
protection – exclude others from 19 See Gordon (2018).
using knowledge communicated 20 Visit en.wikipedia.org/wiki/
to the public. See WIPO (2011) for International_Space_Station.
further discussion. 21 See Crescenzi et al. (2019).
121
References
Alcidi, C., J.N. Ferrer, M. Di Salvo, R. Ganong, P. and D. Shoag (2017). Why Srivastava, M. (2018). Israel’s tech
Musmeci and M. Pilati (2018). Income has regional income convergence in expansion stokes glaring inequality
Convergence in the EU: A Tale of the U.S. declined? Journal of Urban in Tel Aviv. Financial Times,
Two Speeds. Commentary, January Economics, 102, 76–90. December 10.
9. Brussels: Centre for European
Policy Studies. Gordon, R.J. (2018). Declining Thursby, J. and M. Thursby (2006).
American economic growth despite Here or There? A Survey of Factors
Arrow, K. (1962). Economic welfare ongoing innovation. Explorations in in Multinational R&D Location.
and the allocation of resources Economic History, 69, 1–12. Washington, D.C.: The National
for invention. In Nelson, R.R. (ed.), Academies Press.
The Rate and Direction of Inventive Grossman, G.M. and E. Helpman
Activity: Economic and Social Factors. (1991). Innovation and Growth in the Tyson, L.D. (1984). Who’s
Princeton, NJ: Princeton University Global Economy. Cambridge, MA: Bashing Whom: Trade Conflict
Press, 609–626. MIT Press. in High-technology Industries.
Washington, D.C.: Institute for
Bloom, N., C.I. Jones, J. Van Reenen Helpman, E. (2018). Globalization and International Economics.
and M. Webb (2019). Are Ideas Inequality. Cambridge, MA: Harvard
Getting Harder to Find? NBER University Press. von Hippel, E. (1994). ‘Sticky
Working Paper Series, No. 23782. information’ and the locus of problem
Cambridge, MA: National Bureau of Israel Innovation Authority (2019). solving: implications for innovation.
Economic Research. State of Innovation in Israel Management Science, 40, 429–439.
2018. www.innovationisrael.org.il
Brander, J.A. and B.J. Spencer WIPO (2011). World Intellectual
(1985). Export subsidies and Krugman, P.R., M. Obstfeld and M. Property Report 2011: The Changing
international market share rivalry. Melitz (2018). International Economics: Face of Innovation. Geneva: WIPO.
Journal of International Economics, Theory and Policy (11th edition).
18(2), 83–100. Boston, MA: Pearson Education. WIPO (2015). World Intellectual
Property Report 2015: Breakthrough
Cohen, W.M. and D.A. Levinthal Patel, D., J. Sandefur and A. Innovation and Economic Growth.
(1989). Innovation and learning: two Subramanian (2018). Everything Geneva: WIPO.
faces of R&D. Economic Journal, you know about cross-country
99, 569–596. convergence is (now) wrong. Realtime WIPO (2019). Technology Trends 2019:
Economic Issues Watch, October 15. Artificial Intelligence. Geneva: WIPO.
Crescenzi, R., S. Iammarino, C. Peterson Institute for International
Ioramashvili, A. Rodríguez-Pose and Economics. www.piie.com/blogs/
M. Storper (2019). The Geography of realtime-economic-issues-watch/
Innovation: Local Hotspots and Global everything-you-know-about-cross-
Innovation Networks. WIPO Economic country-convergence-now-wrong
Research Working Paper No. 57.
Geneva: WIPO. Pritchett, L. (1997). Divergence,
big time. Journal of Economic
Foray, D. (2015). Smart Specialisation: Perspectives, 11(3), 3–17.
Opportunities and Challenges
for Regional Innovation Policy. Rodríguez-Pose, A. (2018). The
Abingdon: Routledge. revenge of the places that don’t
matter. VOX, February 6. voxeu.org/
article/revenge-places-dont-matter
122
Technical notes
Technical notes
Country income groups Statistical Database (PATSTAT, April 2019) and WIPO’s

This report uses the World Bank income classification to Patent Cooperation Treaty (PCT) collections. In the
refer to particular country groups. The classification analyzed period (1970–2017), these sources account
is based on gross national income per capita in 2018 for 49,286,675 first patent filings and 26,626,660 subse
and establishes the following four groups: low-income quent patent filings, totaling 75,913,335 patent applica
economies (USD 1,025 or less); lower middle-income tions from 168 different patent offices.
economies (USD 1,026 to USD 3,995); upper middle-
income economies (USD 3,996 to USD 12,375); and The main unit of analysis is the first filing for a set of
high-income economies (USD 12,376 or more). patent applications filed in one or more countries
and claiming the same invention. Each set containing
More information on this classification is available one first and, potentially, several subsequent filings
at data.worldbank.org/about/country-classifications is defined as a patent family. The analysis also distin
guishes foreign-oriented patent families – also referred
Country region groups to as international patent families – from domestic-
The country regions used in this report are closely based only ones. Foreign-oriented patent families concern
on the geographic regions from the Standard Country or those inventions for which the applicant has sought
Area Codes for Statistics Use, 1999 (Revision 4) known patent protection beyond its home patent office.
as M49 and published by the Statistics Division (UNDS) This definition includes also patent applications by
of the Department of Economic and Social Affairs, applicants filing only abroad, filing only through the
United Nations (UN). The full methodology can be PCT system or filing only at the EPO. Reciprocally,
found at unstats.un.org. domestic-only patent families refer to those patent
applications filed only at the applicant’s home office –
To simplify the analysis, some changes are introduced regardless of how many filings in the home office there
to this methodology. These are the following: Western are within the same family – without any subsequent
Europe includes Andorra, Austria, Belgium, Denmark, foreign filing though the Paris or PCT routes. Likewise,
Finland, France, Germany, Greece, Iceland, Ireland, patent applications with applicants of more than
Italy, Liechtenstein, Luxembourg, Malta, Monaco, one origin are by definition foreign-oriented patent
Netherlands, Norway, Portugal, San Marino, Spain, families. In addition, about 30 percent of the patent
Sweden, Switzerland, and United Kingdom. Central families relate only to utility model protection, which
and Eastern Europe includes all countries in the are mostly domestic only.
M49’s Northern and Southern Europe regions not
included in Western Europe. The geographical Unless otherwise stated, the report makes use of
subregions Southern Asia, Central Asia and South- international patent families only as the unit of analysis
eastern Asia are grouped in one category, which also for all patent statistics reported. This relates mostly
includes Mongolia. to the incomplete coverage of the domestic-only
patents (and utility models) of many national collec
Scientific publication data tions in PATSTAT. While the top national and interna
The scientific publication data used in this report comes tional offices are usually well covered – namely U.S.
from 27,726,805 records published from 1998 to 2017 in Patent and Trademark Office (USPTO), Japan Patent
the Science Citation Index Expanded (SCIE) of the Web Office (JPO), Korean Intellectual Property Office(KIPO),
of Science (WOS), the citation database operated by National Intellectual Property Administration of the
the Clarivate Analytics company. The analysis focuses People’s Republic of China (CNIPA), EPO and WIPO –
on 23,789,354 observations referring only to scientific some other offices have limited coverage in PATSTAT.
articles, conference proceedings, scientific abstracts For instance, the coverage in PATSTAT of national
and data papers. Scientific articles constitute the bulk collection data from some top 20 patent offices –
of the resulting dataset. such as India, Indonesia, Iran (Islamic Republic of),
Mexico and Turkey – is limited. As a result, the report
Patent data makes use of the information of 8,955,990 interna
The patent data used in this report are from the tional patent families containing 35,582,650 different
European Patent Office’s (EPO) Worldwide Patent patent applications.
123
Geocoding areas to be used in place of administrative ones (see

The geocoding – i.e., attributing the latitude and longi Ester et al., 1996). A first set – named global innovation
tude to a given location – of the scientific publication hotspots (GIHs) – captures the most innovation-dense
and patent data was performed using all available geographical areas of the world in terms of scientific
information on addresses and already existing geocod articles or patent families per square kilometer (km).
ing exercises for these data. By definition, these areas are internationally compa
rable and geographically distinct. The same scientific
In the case of scientific publications, the report assumes publication or patent density determines the same
that research conducted for any publication takes hotspot anywhere in the world, although the threshold
place at the institutions and organizations to which the is different for scientific publication and patent data.
authors declare their affiliation. Ninety-seven percent of No patent or scientific publication address can be in
all the available affiliation addresses were geocoded at two hotspots at the same time.
the postal code or sub-city level. In the case of authors
with more than one affiliation in the same publication, A second set, named specialized niche clusters (SNCs),
all different addresses were considered. was created to avoid biases arising from some scientific
or technological fields being overrepresented in the
In the case of patents, 87 percent of the international scientific publication and patent data, respectively. The
patent families filed from 1976 to 2015 were geocod SNCs capture areas with high innovation density in one
ed. Most of the non-geocoded cases had no usable or more specific scientific publication or patenting fields,
address information. As far as possible, the geocod and that otherwise have not met the criteria to be a
ing was applied to the inventors’ addresses by using global innovation hotspot (GIH). The resulting clusters
the most complete and reliable data source available are also distinct geographical areas, as the overlapping
within each patent family. In addition, the data were clusters for different fields are consolidated into one
enriched with exiting geocoded patent data (see Yin cluster. But they are only internationally comparable
and Motohashi, 2018; Ikeuchi et al., 2017; Li et al., 2014; within their specific scientific or technological field
de Rassenfosse et al., 2019; Morrison et al., 2017). All (or fields).
these sources and WIPO’s geocoding were analyzed
and consolidated to get the best possible geocoded As a result, the report identifies 174 GIHs and 313 SNCs
data for each patent family. When there was more than worldwide. The detailed identification method is
one source for a given patent family, the following order described as follows:
of priority was given: (1) sources having information
from the inventor (inventor principle); (2) sources having First, the points within GIHs are identified using the
more inventors’ addresses covered (coverage prin Density-based spatial clustering of applications with
ciple); (3) sources with the best geocoding resolution noise (DBSCAN) clustering algorithm applied separately
(resolution principle); (4) sources closest to the address to the geocoded patent and scientific publication data.
country – e.g., entrusting Chinese addresses to CNIPA The DBSCAN method requires two parameters – mini
data, Japanese addresses to Japan Patent Office (JPO) mum radius and points – to establish the minimum
data, etc. (local principle); and (5) manual check and acceptable density to form a candidate area. These two
ad hoc selection when two or more sources were still parameters were set differently for patents and scien
available. As a result, many inventor’s addresses were tific publications. The radius for scientific publication
geocoded at a precise level – i.e., street or block – but data was set to 23 km, which is the average commuting
others only at the postal code or other sub-city level. distance to work in OECD countries. Given the more
Patent families containing more offices are more likely precise geocoding of patent data, and based on visual
to be geocoded and at higher quality. This is another inspection, the radius was set to the smaller value of
reason why the report relies only on international 13 km. The minimum points parameter was set to the
patent families. For more information, please refer to median patent and scientific publication density of all
Miguelez et al. (2019). possible circumferences given the radius of each data
set. As a result, the minimum patent density of GIHs
Measuring innovation agglomeration was set to 1,453 patents per 10 km² and the minimum
In order to handle the modifiable areal unit prob scientific publication density is 3,328 scientific publica
lem (MAUP) and its resulting statistical distortions, tions per 10 km².
this report created two sets of ad hoc comparable
124
Technical notes
Second, the resulting groups of points from DBSCAN The following IPC and CPC symbols were used to
are used to determine the candidate geographical determine the AV-related patents and are based on prior
areas – i.e., boundaries – of the GIHs. The borders of patent landscapes of the UKIPO, EPO and JPO. Some
each scientific publication and patent agglomeration of the CPC and IPC symbols were used in combination
are determined using the k-nearest neighbors concave only with some keywords.
hull approach for each patent and scientific publica
tion group of points (see Moreira and Santos, 2007). In Standalone symbols: G05D 1/0088; G05D2201/0207;
order to avoid abnormal polygon shapes, the concave G05D2201/0212; G08G 1/22; B60L2260/40%;
algorithm was set to have at least 75 percent of the B 6 0 L 223 0%; B 6 0 K31/0 0 0 8; B 6 0 K31/0 0 0 8;
convex area covered by all the outer points of a given B60K2031/0091; B60K31/0058; B60K31/0066;
group. In the handful of cases where the group had B 6 0 W 2 5 5 0 /4 0 ; B 6 0 W 2 6 0 0%; G 01S15 / 8 8 ;
less than three coordinates, the polygon was set to a G06K9/00791; G06T2207/30252; G08G1/096791;
circumference of 13 km radius. The overlapping poly G08G1/16; G08G1/22; H04L67/12; Y02P90/285.
gons are merged, keeping only the outer borders of all
concerned agglomerations. However, if the overlap was Symbols in combination with keywords: B60L%;
less than 5 percent of either polygon, these were manu B60W%; B60W2030/%; B60W2040/%; B60W2050/%;
ally inspected and corrected. All patents and scientific B60W30/%; B60W40/%; B60W50/%; B60Y%;
articles within the resulting polygons are considered B60Y2200/11; B62D%; G01S13/93; G01S13/931;
in the analysis, regardless whether they were or were G01S15/93; G01S15/931/%; G01S17/88; G01S17/93;
not part of the DBSCAN results. G01S17/936; G01S7/022; G01S7/4806; G05D1/02;
G05D1/021/%; G08G1/16%; Y02T10/%; Y02T90/%.
Third, the above method is repeated for 25 sub-
samples of the same publication and patent data, Keywords: (ground | car | cars | lorri | lorry | road | street
which refer to 12 scientific fields and 13 technological | highway | convoy | platoon | fleet), (autonomous |
ones, respectively. The radius parameters are again unmanned | driver[.]{0,}less | agv), and NOT (air | aer |
set as 13 km for patents and 25 km for scientific publi drone | flight | flies | fly).
cations. The minimum points are set to the median
patent density of each of the 13 technological fields In the case of scientific publication data, an iterative
and the median scientific publication density of each process was applied. First, a keyword-based strategy
of the 12 scientific fields of all possible circumfer was made on the abstracts of the WoS SCIE data by
ences given the radius of each dataset. From the combining the following two lists of terms: (1) auto
resulting groups of each of these 25 iterations, only mated, autonomous, self-driving, driverless, unmanned,
the points not contained within a GIH hotspot are robotic, pilotless and unpiloted; and (2) vehicle, car,
kept to compute the concave polygon areas. From the truck, taxi, shuttle, lorry, driving, transport(ation)
resulting polygons, the overlapping ones are merged and automobile.
in the same way as mentioned above.
Second, the tags declared by the authors of the result
Mapping strategies ing scientific articles were then manually inspected
The patent mapping strategy for each of the two to build a new list of the following 40 terms: adaptive
sectors – autonomous vehicles in Chapter 3 and plant cruise control; advanced driver assistance system;
biotechnologies in Chapter 4 – is based on prior stud automated driving system; automated lane change
ies and experts’ suggestions. Whenever possible, maneuver; automatic vehicle control; automatic vehicle
each strategy relied on and was compared to existing following; automotive radar; automotive sensors; auton
equivalent scientific publication and patent mapping omous mobile robots; autonomous navigation; autono
exercises. For more details please see Graff and mous valet parking; autonomous vehicular networks;
Hamdan-Livramento (2019) and Zehtabchi (2019). autonomous-vehicle lane; collision avoidance; crash
avoidance; DARPA; DARPA urban challenge; Defense
Autonomous vehicles (AV) Advanced Research Projects Agency (DARPA) urban
The AV mapping is based on a combination of patents challenge; drivable-region detection; intelligent cruise
in PATSTAT data and scientific articles in WoS SCIE control vehicles; intelligent unmanned autonomous
data sampled based on patent classifications, scientific system; LADAR; laser imaging detection and ranging;
subjects and keywords. These are detailed as follows. LIDAR; LIDAR object detection; light detection and
125
ranging (LIDAR); look-ahead sensing; moving vehicle The following IPC and CPC symbols were used to
detection; obstacle avoidance; obstacle detection; determine the patents on each crop biotech category
pedestrian detection; pedestrian-crossing detection; and the union of these constitute the total of crop
platoon; predictive cruise control; unmanned ground biotech patents:
vehicle; unmanned surface vehicles; vehicle automa
tion; vehicle detection; vision-based guidance; wheeled Crop genetic improvement: A01H1%; A01H3%;
robotic vehicle. A01H4%; A01H5%; A01H6%; A01H7%; A01H17%;
C12N5/04%; C12N5/14%; C12N15/05%; C12N15/29%;
Third, the 40 terms were used in the abstracts and C12N15 / 79%; C12N15 /8 2%; C12N15 /8 3%;
titles of articles to extract a new set. To avoid false C12N15/84%; (C07K14/415% but not A61K%).
positives, articles published in journals tagged in the
following WoS subjects were excluded: Anatomy/ Pest control in crops: A01N63%; A01N65%;
Morphology; Art; Astronomy/Astrophysics; Audiology/ C12N15/31%; C12N/32%; (C07K14/325% but
Speech-Language Pathology; Behavioral Sciences; not A61K%).
Biochemistry/Molecular Biology; Biodiversity/
Conservation; Biophysics; Biotechnology/Applied Soil fertility: C05F%.
Microbiology; Cardiovascular System/Cardiology; Cell
Biology; Chemistry; Crystallography; Developmental Climate change: Y02A40/146; Y02A40/162; Y0240/164.
Biology; Education/Educational Research; Emergency
Medicine; Endocrinology/Metabolism; Entomology; The scientific publications were extracted from top
Environmental Sciences/Ecology; Evolutionary plant biotechnology scientific journals and from the
Biology; Fisheries; Food Science/Technology; conjunction of top scientific journals for agriculture
Forestry; Gastroenterology/Hepatology; General/ biotechnology and keywords. These are:
Internal Medicine; Geochemistry/Geophysics;
Geography; Geology; Geriatrics/Gerontology; Health (1) All articles from the following top plant biotechnology
Care Sciences/Services; Immunology; Infectious journals: Agri Gene; Crop Science; Euphytica; Genetics,
Diseases; Information Science/Library Science; Life Selection, and Evolution; Journal of Experimental
Sciences/Biomedicine – other topics; Linguistics; Botany; Journal of Plant Physiology; New Phytologist;
Marine/Freshwater Biology; Medical Informatics; Physiologia Plantarum; Plant and Cell Physiology;
Medical Laboratory Technology; Meteorology/ Plant Cell; Plant Cell and Environment; Plant Cell
Atmospheric Sciences; Microbiology; Mineralogy; Reports; Plant Journal; Plant Molecular Biology; Plant
Mining/Mineral Processing; Neurosciences/Neurology; Physiology; Plant Physiology and Biochemistry; Plant
Nuclear Science/Technology; Nursing; Nutrition/ Science; Planta.
Dietetics; Obstetrics/Gynecology; Oceanography;
Ophthalmology; Orthopedics; Otorhinolaryngology; (2) Top agriculture biotechnology scientific journals
Pathology; Pediatrics; Pharmacology/Pharmacy; and keywords:
Physiology; Plant Sciences; Psychiatry; Psychology;
Public Environmental/Occupational Health; Radiology Top agriculture biotechnology scientific jour
Nuclear Medicine/Medical Imaging; Rehabilitation; nals: Biochemical and Biophysical Research
Research/Experimental Medicine; Respiratory System; Communications; Cell; Journal of Biological Chemistry;
Rheumatology; Social Sciences – other topics; Sport Journal of Biology; Journal of Cell Biology; Journal
Sciences; Surgery; Toxicology; Transplantation; of Molecular Biology; Journal of the American
Tropical Medicine; Urology/Nephrology; Veterinary Medical Association; Molecular and Cellular Biology;
Sciences; Water Resources; Zoology. Nature; Nature Biotechnology; New England Journal
of Medicine; PlosBio; Proceedings of the National
Crop biotechnologies Academy of Sciences of the USA; Science; The EMBO
The crop biotechnology mapping is based on a combi Journal; Theoretical and Applied Genetics.
nation of patents in PATSTAT data and scientific articles
in WoS SCIE data sampled based on patent classifi Keywords: abscisic acid; ACC oxidase; ACC synthase;
cations, scientific journals and keywords. These are aerenchyma; agrobacterium rhizogenes; agrobacte-
detailed as follows. rium tumefaciens; agrobacterium; alfalfa; ammonium;
anther culture; anthocyanins; apoplast; arabidopsis;
126
Technical notes
arbuscular mycorrhiza*; auxin; bacterial blight; banana; nicotiana tabacum; nitrogen fixation; orchid; oryza;
barley; beta vulgaris; rachypodium distachyon; bras oryza sativa; osmotic adjustment; osmotic poten
sica; bread wheat; breeding; breeding value; C-4 tial; pea; peach; pectin; pepper; perennial ryegrass;
photosynthesis; canola; capsicum annuum; carrot; phaseolus vulgaris; phenotyping; phloem transport;
cassava; chickpea; chinese cabbage; chlorophyll a physcomitrella patens; phytic acid; phytotoxicity; picea
fluorescence; chloroplast DNA; citrus; coffea arabica; abies; pinus; pinus pinaster; pinus taeda; pisum; plant
cold tolerance; common bean; conifer*; cotton; cross breeding; plant defence; plant regeneration; plant
breeding; cucumis melo; cucumis sativus; cytokinins; transformation; pollen development; pollen germina
cytoplasmic male sterility; daucus carota; defoliation; tion; pollen tube; potato; prunus persica; QTL*; QTL
distillers grains; doubled; downy mildew; drought analysis; QTL mapping; QTLs; quantitative trait loc*;
resistance; ectomycorrhizal; eucalyptus; flaxseed; rapeseed; resveratrol; RFLP; rice; root elongation;
forage; fructan; fruit development; fruit quality; fruit root exudates; rubisco activase; rye; sap flow; seed;
ripening; fusarium; fusarium graminearum; fusarium self-incompatibility; shoot regeneration; solanum lyco-
head blight; garlic; genome; genotype x environment persicum; solanum tuberosum; somaclonal variation;
interaction; genotype; germplasm; gibberellins; glycine somatic embryogenesis; sorghum; soybean; spinacia
max; gossypium hirsutum; grain; grain filling; grain oleracea; stomatal conductance; strawberry; sucrose
yield; grapevine; hairy root; haploid; hevea brasilien- synthase; sugar beet; sugarcane; sunflower; suppres
sis; high; hordeum vulgare; hypersensitive response; sion subtractive hybridization; tall fescue; thlaspi caer-
kiwifruit; leaf anatomy; leaf growth; leaf rust; legume; ulescens; tomato; transgenic plant*; transgenic rice;
linseed; lolium perenne; lycopersicon esculentum; transgenic tobacco; tritic*; triticum aestivum; vicia
maize; male sterility; marker; medicago truncatula; faba; vitis vinifera; water potential; water use efficiency;
methyl jasmonate; micropropagation; mycorrhiza*; wheat; winter wheat; xylem sap; zea may*.
References
de Rassenfosse, G., J. Kozak and Industry using Linked Dataset of Second International Conference
F. Seliger (2019). Geocoding of Science, Technology and Industry. on Computer Graphics Theory
worldwide patent data. papers.ssrn. RIETI Discussion Paper Series, and Applications (GRAPP 2007),
com/sol3/papers.cfm?abstract_ 17-E-056. www.rieti.go.jp/en/publi Barcelona, March 8–11. INSTICC
id=3425764 cations/summary/17030073.html Press. ISBN 978-972-8865-71-9,
pp. 61–68.
Ester, M., H.-P Kriegel, J., Sander Li, G.-C., R. Lai, A. D’Amour,
and X. Xu (1996). A density-based D.M. Doolin, Y. Sun, V.I. Torvik, Morrison, G., M. Riccaboni and F.
algorithm for discovering clus A.Z. Yu and L. Fleming (2014). Pammolli (2017). Disambiguation
ters in large spatial databases Disambiguation and co-authorship of patent inventors and assignees
with noise. Proceedings of the networks of the U.S. patent inventor using high-resolution geoloca
2nd International Conference on database (1975–2010). Research tion data. Scientific Data, 4. doi.
Knowledge Discovery and Data Policy, 43, 941–955. org/10.1038/sdata.2017.64
Mining (KDD-96), Portland, Oregon,
August 2–4, Menlo Park, CA: AAAI Miguelez, E., J. Raffo, C. Chacua, Yin, D. and K. Motohashi (2018).
Press, 226–231. M. Coda-Zabetta, D. Yin, F. Lissoni Inventor Name Disambiguation
and G. Tarasconi (2019). Tied with Gradient Boosting Decision
Graff, G. and I. Hamdan-Livramento In: The Global Network of Local Tree and Inventor Mobility in China
(2019). The Global Innovation Innovation. WIPO Working Paper (1985–2016), RIETI Discussion
Network of Plant Biotechnology. No. 58, November. Geneva: WIPO. Paper Series, 18-E-018.
WIPO Economic Research Working
Paper No. 59. Geneva: WIPO. Moreira, A. and M.Y. Santos (2007). Zehtabchi, M. (2019). Measuring
Concave hull: A k-nearest neigh Innovation in the Autonomous
Ikeuchi, K., K. Motohashi, R. bours approach for the computa Ve h i c l e Te c h n o l o g y. W I P O
Tamura and N. Tsukada (2017). tion of the region occupied by a Economic Research Working Paper
Measuring Science Intensity of set of points. In Proceedings of the No. 60. Geneva: WIPO.
127
Acronyms
ADAS advanced driver assistance systems GIH global innovation hotspot

ADS automated driving systems GIN global innovation network
AHS automated highway systems GIO genetically improved organism
AI artificial intelligence GM General Motors
AV autonomous vehicle GMO genetically modified organism
BIO Biotechnology IARC International Agriculture
Innovation Organization Research Center
Bt Bacillus thuringiensis ICT information and
CAAS China Academy of communication technology
Agricultural Sciences IP intellectual property
CBD Convention on Biodiversity IPC International Patent Classification
CEO chief executive officer IRRI International Rice Research Institute
CERN European Organization for IT information technologies
Nuclear Research MaaS Mobility-as-a-Service
CGIAR Consultative Group for International MAUP modifiable areal unit problem
Agricultural Research MIT Massachusetts Institute
CIMMYT International Maize and Wheat of Technology
Improvement Center MNC multinational company
CIP International Potato Center NARS national agriculture research systems
CMU Carnegie Mellon University NOAA National Oceanic and
CNRS Conseil National de Atmospheric Administration
Recherche Scientifique OECD Organisation for Economic
Commission European Commission Co-operation and Development
CPC Cooperative Patent Classification OEM original equipment manufacturer
CRISPR-Cas9 clustered regularly interspaced PCT Patent Cooperation Treaty
short palindromic repeats-CRISPR R&D research and development
associated protein 9 rDNA recombinant DNA
CSAIL MIT’s Computer Science and S&T Science and Technology
Artificial Intelligence Laboratory SCIE Science Citation Index Expanded
CTO chief technology officer SNC specialized niche cluster
C-V2X cellular vehicle-to-everything TRI Toyota Research Institute
DARPA Defense Advanced Research U.K. United Kingdom
Projects Agency U.S. United States of America
DNA deoxyribonucleic acid UN United Nations
ECJ European Court of Justice UPOV Union for the Protection of New
EMBRAPA Brazilian Agricultural Research Varieties of Plants
Corporation USDA U.S. Department of Agriculture
EPA U.S. Environmental Protection USPTO U.S. Patent and Trademark Office
Agency V2I vehicle-to-infrastructure
EU European Union V2V vehicle-to-vehicle
FAO Food and Agriculture Organization VW Volkswagen
of the United Nations WatCAR Waterloo Centre for
FCA Fiat Chrysler Automobiles Automotive Research
FDA U.S. Food and Drug Administration WIPO World Intellectual
FDI foreign direct investment Property Organization
GDP gross domestic product WTO World Trade Organization
GEO genetically engineered organisms
128
World Intellectual Property Organization
34, chemin des Colombettes
P.O. Box 18
CH-1211 Geneva 20
Switzerland
Tel: + 41 22 338 91 11

Fax: + 41 22 733 54 28
For contact details of WIPO’s External Offices WIPO Publication No. 944E/19
visit: www.wipo.int/about-wipo/en/offices ISBN 978-92-805-3095-7

Propiedad Intelectual 2019

Uploaded by

Copyright:

Available Formats

Propiedad Intelectual 2019

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Propiedad Intelectual 2019

Uploaded by

Copyright:

Available Formats

World Intellectual Property

The user is allowed to reproduce, distribute, adapt, translate

Suggested citation: WIPO (2019). World Intellectual Property

Adaptation/translation/derivatives should not carry any official

For any derivative work, please include the following disclaimer:

When content published by WIPO, such as images, graphics,

To view a copy of this license, please visit:

The designations employed and the presentation of material

The mention of specific companies or products of manufacturers Attribution 3.0 IGO

Acronyms 128 Chapter 2

Geography has always played a

intertwined However, they neither capture all such activity, nor

innovation among innovators. In addition, the dynamic forces

We hope that this report contributes to a greater appre­

The geography of innovation can

New players, particularly Asian

For most of the period from 1970 to 2000 only three

Inventors and scientists within hotspots and niche

The intensity of international collaboration varies

International cooperation in relation to innovation has

Falling R&D productivity reinforces

Continuously pushing the technological frontier is

Falling R&D productivity calls for constantly increasing

The changing global

Starting in the 1970s and 1980s, a third industrial revo­

From the mid-20th century until the Great Recession, Agglomeration

R&D offshoring Knowledge

back to home countries. At the same time, however,

The international and inter-regional mobility of skilled

These patterns seem to prevail in the largest global

The rising need for complex and specialized knowledge

Global innovation networks have been a key centrifugal

In view of these considerations, it is crucial to under­

knowledge-production centers or hubs results in an

publication, all different addresses are considered.

2.1 The two sides of global 60

The accelerating international 40

Where does knowledge production take place? Is

that of other economic activities? Empirical evidence

economies – particularly, the U.K., France, Switzerland

The rest of the world passed from less than 6 percent

Asia’s share in innovation rises strongly

Region Patents Publications

More the value, more the concentration

Shares of top innovation subnational regions within countries

Country Patents Publications

Baden-Württemberg Bayern Bayern Nordrhein-Westfalen

Île-de-France Île-de-France Île-de-France Île-de-France

Greater London Greater London Greater London Greater London

Maharashtra Karnataka Maharashtra Tamil Nadu

Tokyo Tokyo Tokyo Tokyo

California California California California

Source: WIPO based on PATSTAT, PCT and Web of Science data.

Population density does not ensure high innovation density

Population (ca. 2017)

Min: 10,852,000 Max: 37,800,000

Innovation density and urban density largely coincide

GIH SNC NIGHTLIGHT

agglomerations or hotspots – which are by definition

means the urban areas can be less dense.

We hope that this report contributes to a greater appre

Starting in the 1970s and 1980s, a third industrial revo

In view of these considerations, it is crucial to under