OECD 2017 The Links Between Global Value Chain

the links between
global value
chains and
global innovation
networks
an exploration
OECD SCIENCE, TECHNOLOGY

AND INNOVATION
POLICY PAPERS
April 2017 No. 37
THE LINKS BETWEEN GLOBAL VALUE CHAINS AND GLOBAL INNOVATION NETWORKS: AN EXPLORATION
The paper was written by Koen De Backer, Timothy Destefano and Laurent Moussiegt of the
OECD Secretariat. The Committee on Industry, Innovation and Entrepreneurship (CIIE) approved and
declassified this report on 5 November 2016 as part of its work on “GVC upgrading and extensions”.
Note to Delegations: This document is also available on OLIS with the cote:
DSTI/IND(2016)8/FINAL
This document, as well as any data and any map included herein, are without prejudice to the
status of or sovereignty over any territory, to the delimitation of international frontiers and boundaries
and to the name of any territory, city or area.
The statistical data for Israel are supplied by and under the responsibility of the relevant Israeli
authorities. The use of such data by the OECD is without prejudice to the status of the Golan Heights,
East Jerusalem and Israeli settlements in the West Bank under the terms of international law.
© OECD 2017
You can copy, download or print OECD content for your own use, and you can include excerpts from
OECD publications, databases and multimedia products in your own documents, presentations, blogs,
websites and teaching materials, provided that suitable acknowledgment of OECD as source and
copyright owner is given. All requests for commercial use and translation rights should be submitted
to [email protected].
2 OECD SCIENCE, TECHNOLOGY AND INDUSTRY POLICY PAPERS

FOREWORD
Economic globalisation has dramatically increased during the past decades and as a result,
economic activities increasingly stretch out across OECD and emerging economies. Goods, services,
capital, people, technology and knowledge are easily transferred across national borders. Following
the international fragmentation of production, goods and services are nowadays produced and heavily
traded in international production networks or Global Value Chains (GVCs). More recently,
innovation activities have also become increasingly internationalised thereby giving rise to global
innovation networks (GINs). The networks typically consist of own R&D facilities abroad as well as
collaborative arrangements with external partners and suppliers.
The nexus between these two types of networks in the global economy has not really been
explored, although strong interdependencies between GVCs and GINs are likely. For example, do the
two types of networks show a similar geographic and industry concentration; more specifically, are
hubs and nodes of GVCs the same as in GINs? Do countries that trade with each other within GVCs
show also a propensity to cooperate in GINs? Are countries that are involved in trade and production
networks more likely to innovate with each other? Does the co-operation in innovation and GINs
result in better economic performance in GVCs?
This paper takes a first attempt to analyse the linkages between both types of networks and
identify a number possible government implications. The motivation for this analysis is that concerns
are raised in policy discussions that countries are not able to capture the value of their innovative
activities.
OECD SCIENCE, TECHNOLOGY AND INDUSTRY POLICY PAPERS 3

ABSTRACT
Economic globalisation has given rise to two types of networks that stretch out across OECD and
emerging economies. At the one side, Global Value Chains (GVCs) can be thought of as the
“material” transfers of goods and services (final as well as intermediate) across borders. At the other
side, Global Innovation Networks (GINs) refer to the transfers of intangibles and immaterial assets
between countries. Concerns are increasingly raised in policy discussions that countries are not able to
capture the value of their innovative activities, hence the clear need to better understand the
interdependencies between these two types of networks.

TABLE OF CONTENTS
FOREWORD ................................................................................................................................ 3
ABSTRACT .................................................................................................................................. 3
THE LINKS BETWEEN GLOBAL VALUE CHAINS AND GLOBAL INNOVATION
NETWORKS: AN EXPLORATION ............................................................................................ 6
1. Introduction: the globalisation of economic activities .......................................................... 6
2. Mapping of GVCs and GINs ................................................................................................ 7
2.1. Empirical measures of GVCs and GINs ......................................................................... 7
2.2. The geography of co-invention and GINs ..................................................................... 12
3. From GVCs to GINs: does trade promote co-invention? ................................................... 21
4. From GINs to GVCs: does co-invention induce GVC upgrading? .................................... 25
5. Conclusions and some policy implications ........................................................................ 28
NOTES ........................................................................................................................................ 30
REFERENCES ............................................................................................................................ 33
ANNEX 1: CO-INVENTION AND TRADE INTENSITY, BY COUNTRY, PARTNER AND
INDUSTRY................................................................................................................................. 36
ANNEX 2: CO-INVENTION GRAVITY MODEL WITH DIFFERENT TIMING OF TRADE72
ANNEX 3: CO-INVENTION GRAVITY MODEL WITH VALUE ADDED TRADE ............ 73
ANNEX 4: UPGRADING MODEL FOR INDIVIDUAL INDUSTRIES ................................. 74
Boxes
Box 1. Some tentative conclusions about open innovation and GINs ...................................... 9

THE LINKS BETWEEN GLOBAL VALUE CHAINS AND GLOBAL INNOVATION

NETWORKS: AN EXPLORATION
1. Introduction: the globalisation of economic activities
During the past decades, economic globalisation has dramatically progressed with economic
activities increasingly stretching out across OECD and emerging economies. Goods, services, capital,
people, technology and knowledge are increasingly moving across national borders. Multinational
companies (MNEs) have been a driving force of growing globalisation, but more purely “domestic’
actors like SMEs, universities, public research organisations, etc. have also expanded the geographical
reach of their activities. Globalisation increasingly affects how a growing number of economic actors
operate, compete and innovate, both at home and abroad.
The international trade of goods and more recently also of services has been the most prominent
channel of economic globalisation. For decades, final products have been produced in one country and
were then exported to other countries according to the principle of comparative advantage. In more
recent decades international trade has grown further in importance as individual stages in the
production process are increasingly dispersed across different countries. This has resulted in the
growing international trade of intermediate goods within global value chains (GVCs), in addition to
trade of final and capital goods. Trade and investment are closely linked in GVCs as MNEs have been
the first to internationalise their production networks with some affiliates having a world or regional
mandate for the production of a specific product/service. OECD (2013a) describes in detail the
emergence of these international production networks and gives an overview of their implications on
different policy domains.
More recently, innovation activities have also become increasingly internationalised giving rise to
global innovation networks (GINs); nevertheless, it should be noted that the majority of R&D
investments is still concentrated in the home country near the headquarters (OECD, 2008a; Belderbos
et al., 2016). Following the offshoring of production activities, companies have also started to
offshore R&D activities with MNEs increasingly setting up R&D activities abroad. To support their
international production and distribution activities; these R&D investments abroad were in the
beginning mainly designed to adapt products and process to local market demands. But R&D
investments have recently also been motivated to tap into foreign knowledge, technology and human
capital, and as such leverage the innovation activities which remain concentrated in R&D headquarters
at home.
In addition, a trend of “open” innovation has been observed with companies opening their
innovation process to external partners and outsourcing innovation activities, often abroad (OECD,
2008b). Owing to more intense and global competition and technological progress, product life cycles
have been drastically shortened, forcing companies to innovate more quickly and develop products
and services more efficiently. Moreover, the growing integration of different technologies has made
innovation more costly and riskier. The greater the need for interdisciplinary cross-border and cross-
sector research, the less a single company has the capability to innovate successfully. Companies
increasingly look for partners with complementary expertise to obtain access to different technologies
and knowledge quickly.

These GINS include own R&D facilities abroad as well as collaborative arrangements with
external partners and suppliers, in which firms depend in various ways on the expertise of the different
partners. Companies link into these GINs with people, institutions (universities, government agencies,
etc.) and other companies in their own or different countries to solve problems, source knowledge and
generate ideas. GINs influence national and regional innovation systems. MNEs’ ecosystems or
networks of innovation often represent “nodes” linking regional/national systems of innovation across
borders and therefore various S&T actors in different countries.
Until now, and maybe surprisingly, not much analysis has been undertaken to explore the nexus
between these two types of networks 1, one reason being the lack of accurate and complete data (see
below). In order to identify the likely interdependencies between GVCs and GINs – the first one
representing more “material” transfer across borders, while the second refers more to the transfer of
intangibles and immaterial assets, this paper takes a first attempt to analyse the nexus between both
types of networks and identify a number possible government implications:
− First, how do GVCs and GINs map onto each other? Do the two types of networks show
a similar geographic and industry concentration; more specifically, are hubs and nodes of
GVCs the same as in GINs?
− Second, do countries that trade with each other within GVCs show also a propensity to
cooperate in GINs? Are countries that are involved in trade and production networks
more likely to innovate with each other? If so, non-technological policies focusing on
trade promotion may also help to encourage international co-operation in innovation;
− Third, does the co-operation in innovation and GINs result in better economic
performance in GVCs? It has been shown that co-operation in innovation has positive
effects on technological outcomes and in some cases enhances economic performance
(e.g. productivity), but does it also have an impact on countries’ performance in GVCs,
in particular on the upgrading trajectories within GVCs?
The motivation for casting light on these questions relates to concerns raised in policy discussions that
countries are not able to capture the value of their innovative activities.
In providing a (partial) answer to these questions, this paper complements other CIIE work focusing
on the link between Knowledge Based Capital (KBC) and GVCs. That work aims to identify the
interdependencies between KBCs and GVCs in a broad sense, covering different forms of KBC
(R&D, patents, organisational capital, etc.). By focusing on GINs, this paper focuses specifically on
the importance of international co-operation in innovation.
2. Mapping of GVCs and GINs
2.1. Empirical measures of GVCs and GINs
OECD work in recent years, and particularly the construction of the Trade in Value Added
(TiVA) database 2, has provided the evidence base necessary for the analysis of GVCs. Before this, the
emergence of international production networks was mainly described through case studies and
company/industry surveys. The new TiVA data, however, document the participation and upgrading
of countries in GVCs across industries and over time. Among other contributions, the work has shown
that GVCs have become more important and complex, with more and more countries participating in

these international production networks (OECD, 2013). De Backer and Miroudot (2013) map GVCs
across countries, industries and time and show that the length of GVCs has significantly increased
over time with different roles for different countries. In addition, they show how the geography of
GVCs has changed over the years with emerging economies taking up a growing role in a number of
industries.
The evidence on GINs is not as widely available, and furthermore GINs have largely been
discussed in terms of case studies typically in high-technology sectors. Recently also more aggregate
indicators based on R&D data, patent data, innovation surveys have increasingly been used to describe
the growing importance of open innovation 3 – for an overview see OECD (2008b).
− Data on R&D investments are internationally comparable, but only provide rather
indirect evidence on open innovation. For example, specific information on public-
private funding of R&D reveals some of the interactions and collaboration between
government and the business sector. Government-financed business R&D is typically
more related to direct government funding, than to actual collaboration between
government and business. In contrast, business funding in the higher education and
government sectors (e.g. research centres) often indicates close collaboration between
public and private entities.
− Innovation surveys are increasingly used in OECD and in many non-member countries to
better understand the role of innovation and the characteristics of innovative companies.
The latest surveys include information on innovation linkages, including collaboration on
innovation. Collaboration, defined as the “active participation in joint innovation projects
with other organisations” may involve the joint development of new products, processes
or other innovations with customers and suppliers, as well as horizontal work with other
enterprises or public research bodies. Therefore, more direct evidence on open
innovation and specifically on the sourcing of innovation can be derived from innovation
surveys.
− Patent data are considered a unique, broadly available and reliable source of statistical
material and are increasingly used to study different aspects of the innovation process
since they represent an output of knowledge creation and innovation (Griliches, 1990;
OECD, 2009). 4 Patent data provide detailed information such as the nature of the
invention as well as the actors involved in the innovative process. Patent documents
report the inventor(s), the applicant(s) – i.e. the owner of the patent at the time of
application – along with their addresses and countries of residence. Furthermore, the
time series nature of patent data allows for an analysis of innovative activities over time.
The main disadvantage of patent statistics is that they fail to capture all innovative
activity, as not all innovations are patented and not all patents lead to innovations.
Typically, patent information on different inventors, different co-assignees or owners,
differences between inventors and assignees are used to analyse open innovation and
identify GINs.

Box 1. Some tentative conclusions about open innovation and GINs
Indicators on open innovation based on large-scale data suggest that companies increasingly innovate together
with external and international partners. The industry distribution shows that collaboration on innovation is
significant in manufacturing as well as in services, although certain industries (chemicals, pharmaceuticals, ICT,
including software) typically have higher levels of open innovation. While co-operation in innovation is on the rise,
the data show clearly that larger firms innovate more openly than SMEs. These results suggest that limited
resources may prevent SMEs to deploy open innovation practices more broadly and on an international scale.
Large companies are much more active in public research although there is more cross-country variation for large
firms than for SMEs.
Companies collaborate on innovation with suppliers and customers more than with universities and government
research institutions, at least in terms of numbers of collaborations. This may be because public research focuses
more on upstream research and exploration activities which may be a small part of overall innovation.
The empirical data also show that despite globalisation, geographic proximity still matters in open innovation.
Companies were found to collaborate more with geographically close external partners, although it should be
noted that the data measure the number of interactions and not the intensity and quality of collaboration.
Additional evidence suggests that proximity may matter somewhat less than good connectivity with external
partners.
Source: OECD (2008b); see also OECD Science, Technology and Industry Scoreboard 2013 and 2015.
This paper uses patent data and in particular information on international co-invention to identify
and analyse GINs. It is clear from the start however that this information – just like all indicators
described above - does not allow for the identification of all types of GINs across countries and
industries. International co-invention refers to patents with multiple inventors who reside in different
countries. The patent data used in this paper refer to patent applications filed under the Patent Co-
operation Treaty (PCT) at the EPO, in order to guarantee the international comparability across
countries 5. Addresses of the inventors are used to assign the patent applications to countries while
PCT patents are assigned to industries based on the applicant information 6 using the Harmonized
Applicants’ Names (HAN) database. This database provides a depository of the names connected to
the applicant (and not the inventor) and is expanded upon with business register data.
After dropping equivalent PCT patents – i.e. through the identification of patents with the same
inventors and same priority application date – almost 2 million PCT patents are obtained for the period
between 1995 and 2013. Amongst these, 137,758 international co-inventions were identified based on
the addresses of inventors. Figure 1 shows the level of international co-operation in innovation across
national borders between 1995 and 2013. As can be seen clearly, the trend has generally been
increasing, at the time of the economic crisis in 2008-2009 international co-invention decreased
globally before recovering (see also Guellec and Van Pottelsberghe, 2001 for a discussion of earlier
years). Moreover, the co-invention intensity (i.e., the number of co-inventions over the total number
of PCT patents) has actually decreased since the economic crisis after a gradual rise since 1995.

Figure 1. International co-invention, absolute numbers and intensity, 1995-2013
absolute numbers %
16 000 8
Number of international co-inventions (left scale)
Intensity - international co-invention (right scale)
14 000 7
12 000 6
10 000 5
8 000 4
6 000 3
4 000 2
2 000 1
0 0
Note: Intensity is measured as the number of international co-inventions in percentage of total number of PCT applications.
Source: calculations based on OECD patent database.
Important differences in co-invention exist across industries reflecting the fact that some
industries show a higher propensity to patent and a higher proportion of international co-invention. A
large number of international co-inventions are observed in chemicals, electronics and business
services (Figure 2); other industries with a significant number of international co-inventions are
machinery, whole/retail trade, hotels and restaurants and other services. Intensities (international co-
inventions as percentage of PCT patent) show a somewhat different pattern with high intensities in a
number of industries with a relatively small number of PCT patents: agriculture, mining, food and
financial intermediation. For the other industries, the intensity varies in general between 4 and 10%.

Figure 2. International co-invention across industries, absolute numbers and intensity, 1995-2013
absolute numbers %
30000 14
Number of international co-oinventions (left scale)
25000 Intensity - international co-invention (right scale) 12
20000 10
15000 8
10000 6
5000 4
0 2
Source: calculations based on OECD patent and HAN databases.
Given the central role of MNEs in cross-border activities in general and GINs in particular, it can
be expected that co-invention patterns across countries are partially driven by the activities of MNEs.
Based on the applicants’ information in terms of ownership structure 7, PCT patents and co-inventions
are allocated to MNEs and non-MNEs. 8 The results clearly show the importance of MNEs in both
patenting in general, as well as international co-invention (Figure 3): more than 60% of all PCT patent
applications are found to be related to MNE activities (i.e. where an MNE –headquarter or affiliate –
was the only or one of the applicants of the patent).
For co-inventions, the importance of MNEs is even more marked with two-thirds of the co-
inventions that can be directly linked to MNE activities. In particular, more than 50% concern co-
inventions between inventors in different countries but with the same MNE as applicant (i.e.
headquarters and/or affiliates). Thirteen percent concern co-inventions between at least one MNE and
at least one non-MNE, while 5% concern co-invention between different groups of MNEs. 9

Figure 3. The importance of MNEs in patenting and international co-invention, 1995-2013
PCT patents international co-inventions
25.8% 23.4%
no info
non-MNE
MNE - non-MNE
50.1%
54.3% multi MNE 9.6%
12.1%
same MNE
13.3%
6.0%
1.8% 3.6%
Note: The percentages present the share of each group in the total number of PCT applications and international co-inventions
Source: calculations based on OECD patent, HAN and ORBIS databases.
Figure 4 presents the number of PCT patents and co-inventions according to the nationality of the
MNEs (Top 15 countries, based on the global ultimate owner). 10 MNEs with headquarters in Japan,
the United States and Germany are observed to have most of PCT patent applications between 1995
and 2013 across the globe. 11 But also smaller countries like Sweden, Switzerland and the Netherlands
– i.e. countries with a large presence of domestic MNEs – occupy an important position. In terms of
international co-inventions, MNEs from the United States, Germany and Switzerland make up the top-
3 largest applicants; the large majority of these applications are held within the same MNE group.
Figure 4. Top 15 MNE home countries in patenting and international co-invention, 1995-2013
PCT patents international co-inventions
absolute numbers absolute numbers

250000 25000
200000 one MNE other

20000 one-MNE other
150000 15000
100000 10000
50000 5000
0 0
Source: calculations based on OECD patent, HAN and ORBIS databases.
2.2. The geography of co-invention and GINs
The geography of co-invention has significantly changed over the past decades with GINs
becoming denser and increasingly stretching out over a growing number of countries, including
emerging economies. A first look at international co-invention across different regions shows the
importance of North America and Europe for international co-operation in innovation: together they
were responsible for more than 80% of all international co-inventions between 1995 and 2013 (Figure

5). This is due to their innovation links with other regions but also strong intra-regional co-invention
(particularly for Europe). 12 Between the two sub-periods (1995-2004 and 2005-2013), both regions
witnessed a strong increase in the number of international co-inventions, again both intra- and extra-
regionally. In addition, the growing importance of Asia is clearly observed between the two sub-
periods as Asian partners increasingly collaborate in innovation with North American and to a lesser
extent with European partners. The intra-regional co-inventions in Asia (i.e. international
collaborations among Asian countries) also increased over time, but less strongly than the extra-
regional links of Asia. This growing importance of Asia in co-invention patterns reflects the
importance of Asian markets for companies in Europe and North America as well as the growing
investments in Science, Technology and Innovation (STI) in these countries.
Figure 5. International co-invention by region (absolute numbers), 1995-2004 and 2005-2013
1995-2004 2005-2013
Moving to the individual country level, a country’s degree (or intensity) of international co-
invention can be measured as the number of patents invented by a country with at least one foreign
inventor in the total number of patents invented domestically (see also Hascic et al., 2012). As
suggested by Dernis and Khan (2004), co-invention for individual countries is calculated by fractional
counts depending on the number of inventors and their countries of residence in order to prevent
double counting. To illustrate this, imagine a hypothetical scenario where a patent consists of two
inventors in two different countries. In this case, each country will be allocated a share of 0.50. If
however there are three inventors all from different countries, each country would allocate a share of
0.33 and so on. In addition, the more inventors from a specific country, the greater their share for that
country becomes. For example, a patent with 4 inventors, two from the United States, one from
Germany and one from the United Kingdom, the countries shares are 0.5, 0.25 and 0.25, respectively.

Figure 6 presents the intensity of the different countries in the sample: in line with the growing
trend of international co-invention during the past decades, most of the countries have increased their
co-invention share between 1995 and 2013. Japan and Korea show very low international co-
invention shares indicating the domestic focus of their innovation process (at least in patenting). Often
smaller countries show relatively higher co-invention shares. Also remarkable is the strong co-
invention share of a number of emerging economies, which is again most likely due to the growing
attractiveness of these economies for other partners and the search of these economies for access to
foreign knowledge.
Figure 6. International co-invention intensity by country, 1995-2013
% 100
2005-2013 1995-2004
90
80
70
60
50
40
30
20
10
Analysing the above country co-invention measures by partner allows for a more detailed assessment
of changes in the geography of GINs; in addition similar trade intensity by partner have been
calculated to start to reveal the links between GINs and GVCs.
The following co-invention and trade intensity measures therefore attempt to illustrate the
innovation and trade networks between countries within industries overtime 13:
 Coinventionijk ,t + Coinvention jik ,t 

CoinvIntensity =   (1)
 ∑ Invention + ∑ Invention
 ikt jkt 
 Exportsijk ,t + Exports jik ,t 

TradeIntensity =   (2)
 ∑ Output + ∑ Output
 ikt jkt 

Figure 7 presents maps for three separate industries (one manufacturing industry with a low
number of international co-inventions – textiles; one manufacturing industry with a high number of
international co-inventions – electric and optical equipment and one services industry – real estate,
renting and business services). 14 The graphs for the other industries can be found in Annex 1. 15
Figure 7. Co-invention and trade intensity, by country, partner and industry 1995-2013
Textiles, textile products, leather and footwear (ISIC 17 to 19), 2000-05
1
International co-invention intensity
2
Trade intensity
1. Bilateral co-inventions as a share of all inventions by both countries.

2. Bilateral trade as a share of industry output of both countries.

1
2
Trade intensity


Electrical and optical equipment (ISIC 30 to 33), 2000-05
1
2
Trade intensity


1
2
Trade intensity


Real estate, renting and business activities (ISIC 70 to 74), 2000-05
1
2
Trade intensity


1
2
Trade intensity

Source: calculations based on OECD patent database

A number of common observations emerge from the figures:
a) GINs have become more important over time, as reflected in the growing links between
countries. In addition, a clear broadening of GINs across countries has taken place: the
number of countries engaging with multiple partners has substantially risen. In
particular, South East Asia experienced a significant increase in participation within
innovation networks. In contrast, countries in the African region and much of South
America have shown little increase in engagement in patent collaboration, aside from
Brazil and Chile.
b) The changes over time in international trade networks have overall been smaller,
probably since these networks were already more established.
c) Trade networks in goods are denser than in services, with more trade flows in
manufacturing than in services industries.16 The difference between manufacturing and
services seems to be less marked in innovation networks, with for example in the
services industry of ‘renting, real estate and business services’ a large and growing
number of international co-inventions.
d) When comparing both types of networks, trade networks appear to be more intensive
than GINs: trading relationships show more and ‘thicker’ links between countries than
co-invention networks. GINs have become more intensive in the second sub-period but
not up to the same level as GVCs.
e) Overall, GINs show a more concentrated pattern with a small number of partners
cooperating in innovation. GVCs are more dispersed and countries trade with more
partners than they co-invent. Both types of networks show a strong regional character
suggesting that distance (still) seems to matter in GVCs as well as GINs.
f) A significant geographic overlap exists between GINs and GVCs, taking into account
the differences discussed above. This reflects the strong concentration of international
co-inventions and trade across the large regions, with important hubs in the United
States, Europe (around Germany and France) and East Asia (around Japan, the People’s
Republic of China and Korea). The majority of intra- and extra-regional links are also
centred around these hubs
3. From GVCs to GINs: does trade promote co-invention?
Following the mapping of GVCs and GINs, a more formal analysis is proposed to assess the
potential interdependencies between both types of networks. In a first analysis, the effect of GVCs on
GINs is analysed: does co-operation (i.e. trading) in GVCs increase the propensity and extent of co-
invention between countries. More specifically: do countries that trade with each other in GVCs show
also higher co-invention rates with each other in GINs; do countries that start to trade more intensively
subsequently exhibit higher co-invention rates.? One reason for this link is that through establishing
trade relationships companies/countries may acquire contacts and get to know potential partners for
co-operating in innovation. In addition, working together in one type of transaction makes partners
more familiar with each other and may induce co-operation in other transactions, particularly in
innovation where familiarity and trust can be expected to be more important than for example in trade.

In order to confirm this type of link between GVCs and GINs, a gravity-type 17 model is estimated
explaining bilateral co-invention between countries at the industry level and relating this to a number
of similar factors as in the usual gravity model on the one hand, as well as the trade relations between
the countries on the other hand. This empirical strategy follows a number of papers that have analysed
technological collaboration in a gravity context (Guellec and Van Pottelsberghe, 2001; Maggioni et
al., 2007; Picci, 2010 and Montobbio and Sterzi, 2013). The dependent variable in the model is the
number of bilateral co-inventions between each pair of countries per industry and year. Because of the
large numbers of zeros – a considerable number of country pairs do not show international co-
inventions, thus the gravity model has been estimated using the Poisson Pseudo Maximum Likelihood
(PPML) estimator. The results are presented in Table 1. 18
Patent stocks of country pairs are used as ‘attractors’ in the model, assuming that countries will
collaborate more in innovation the larger their respective innovation activities in general.19 Because of
the limited availability of KBC variables - longitudinal at the industry level and for OECD as well as
emerging economies, patent stock is used as a – albeit imperfect – measure of the innovation activities
of countries. 20 The patent stock for each year has been calculated based on the number of new
inventions and, in line with the existing literature, applying a 15% depreciation rate on the existing
inventions.
Just like in the trade-gravity model, distance is assumed to moderate the probability and extent of
co-invention, GINs are more difficult to develop between partners further away from each other. This
paper uses a number of different ‘distance’ measures across models; in the first place, the traditional
measure of physical distance expressed as the number of kilometres between country/partner pairs.21
Because of the growing importance of communication technologies including broadband, satellite
communications but also social networks, it has been argued that physical distance has however
become less important. Instead, differences in time zones may still matter in today’s easy and low-cost
communications world, as this would complicate and prevent real-time communication between
partners. As such, variables are included on the hour difference in time zones, and a dummy variable
for being in the same time zone. 22
Also cultural and historical distances are taken into account as they have been found to affect the
probability and intensity of collaboration. More specifically, indicators have been included to reflect
whether or not countries speak the same language and if they share a national border or colonial
history.
In addition, a measure of technological distance is included since previous research has

demonstrated that ‘technologically’ close partners seem to collaborate more. Again, several measures
have been proposed in the literature going back to the seminal work of Jaffe in 1989 with each
measure possessing strengths and weaknesses (for an overview, see Stellner, 2014). Technological
distance in this paper is calculated as the difference (in size) between the partners’ patent stocks;
because of the non-directional dependent variable, the absolute value of this difference in included.
Last but not least, the trade relationship between partners is included as an additional measure of
distance to analyse the interdependencies between GVCs and GINs. The hypothesis is that partners
will collaborate more in innovation if they also collaborate in trade networks as they are more familiar
with each other. ‘Trade’ distance is measured by the bilateral trade (exports + imports) from one
country to the other country (at the industry level). As this variable also contains many zero values; a
dummy variable is include in order to assess the effects of zero trade links on co-invention:

a 0 + a1 patstockist + a 2 patstock jst + a 3distanceij + a 4tdiffij +

coinvijst =
a 5 notdffij + a 6 contiguityij + a 7languageij + a 8colonyij + a 9techdistijst + (3)
a10 mneist + a11tradeijst + a12 notradeijst + a13δ ijst + e
with i = country i, j = partner country j, s = industry and t = year
Table 1. Co-invention gravity model: econometric results

g y
Base model a) Base model b) Base model c) Model 1 Model 2 Model 3 Model 4
Patent Stock country 0.510*** 0.537*** 0.474*** 0.474*** 0.473*** 0.469*** 0.484***
(0.01) (0.01) (0.03) (0.03) (0.03) (0.03) (0.06)
Patent Stock partner 0.452*** 0.477*** 0.274*** 0.274*** 0.274*** 0.271*** 0.373***
(0.01) (0.01) (0.04) (0.04) (0.04) (0.04) (0.06)
Physical distance -0.108*** -0.109*** -0.079*** -0.035* -0.035* -0.037
(0.01) (0.01) (0.01) (0.02) (0.02) (0.03)
Time difference -0.067*** -0.041** -0.041** 0.032
(0.01) (0.02) (0.02) (0.04)
Same time zone 0.161*** 0.159*** 0.159*** 0.102*
(0.03) (0.03) (0.03) (0.05)
Contiguity (dummy) 0.327*** 0.328*** 0.381*** 0.407*** 0.378*** 0.370*** 0.371***
(0.04) (0.04) (0.03) (0.03) (0.03) (0.03) (0.05)
Language (dummy) 0.947*** 0.965*** 0.707*** 0.669*** 0.670*** 0.665*** 0.600***
(0.03) (0.03) (0.02) (0.03) (0.03) (0.03) (0.04)
Colony (dummy) 0.053 0.024 -0.105*** -0.124*** -0.118*** -0.118*** -0.032
(0.04) (0.04) (0.03) (0.03) (0.03) (0.03) (0.04)
Technological distance 0.015*** 0.001
(0.01) (0.01)
MNE Turnover 0.015**
(0.01)
Bilateral trade 0.282*** 0.264*** 0.275*** 0.271*** 0.267*** 0.269*** 0.334***
(0.01) (0.01) (0.01) (0.01) (0.01) (0.01) (0.02)
No bilateral trade -2.967*** -2.893*** -3.879*** -3.843*** -3.850*** -3.847***
(0.71) (0.71) (0.75) (0.75) (0.75) (0.75)
Control Variables
Year      
Origin/Destination fe's     
R-squared 0.650 0.664 0.889 0.891 0.891 0.891 0.926
Observations 199,534 199,534 174,132 174,132 174,132 173,91835,670
Note: No bilateral trade is a dummy which equals to one when there is zero trade (imports + exports) from a country to a
partner in a particular industry and zero otherwise. Level of significance * p<0.10, ** p<0.05, *** p<0.01
The base co-invention model shows similar results as in the typical trade-gravity equations 23: Co-
invention is proportional to the innovative activities of the partners and inversely proportional to the
distance between the countries. The size of the innovative activities reflected in the patent stock of the
countries is observed to be an important determinant of international knowledge links (see also
Montobbio and Sterzi, 2013). Also historical and cultural ties play a role: speaking the same language

and sharing a common border is positively related to global co-invention. The effect of colonial
history between countries on co-invention indicates a negative effect which may be due to the different
levels of economic development between country pairs.
Consistent with much of the literature on distance and knowledge diffusion, there appears to be
less international collaboration the greater the distance between countries; both the physical distance in
kilometres (base model and model 2) as well as the difference in time zones (models 1 and 2)
negatively relates to the propensity and extent of co-invention. On the basis of the size of the
estimated coefficients, the results suggest that time difference is somewhat more important than
geographic distance. Furthermore, operating in the same zone strongly and positively correlates to co-
invention between partners. In sum, both face-to-face interaction costs (as proxied by geographic
distance) and virtual interaction costs (as proxied by time distance) seem to matter for GINs
(Montobbio and Sterzi, 2013).
In contrast to previous research on co-invention, the results for technological proximity/distance

(model 3) indicate that co-invention is more likely the larger the technological gap between the
partners. Technology sourcing may be one explanation for this result, with less innovative countries
purposefully setting up technology links with stronger partners in order to access their knowledge and
learn from these more knowledgeable partners. More research is needed in order to confirm this
motivation as the chosen indicator – difference in size of patent stocks – may be an imperfect measure
of technological distance/proximity.
The results demonstrate, unlike Montobbio and Sterzi (2013), a strong and positive correlation
between bilateral trade and co-invention suggesting the interdependence between GVCs and GINs at
the country level. In addition, having no trade relationship with a partner country strongly decreases
the propensity and extent of co-invention with that partner. The results thus demonstrate that being in
a trade network – be it through exporting and/or importing - provides countries with information on
potential partners and/or makes them more familiar with these partners; this positively impacts the
likelihood of cooperating in innovation.
The positive effect of trade on co-invention is also prominent in all the different models when
accounting for the importance of MNEs across industries and countries. As MNEs have been
observed to play an important role for the international co-invention links (see above) as well as for
the trade relationship and specialisation of (some) countries (see. Caves, 1996; Sleuwagen and De
Backer, 2001), the link between GVCs and GINs could be solely driven by MNE activities. This does
not seem to be the case given that the export variable still exhibits a significant effect even after
controlling for MNE presence across industries and countries, 24 confirming the importance of the link
between GVCs and GINs.
However, the inclusion of the MNE variable seems to change the results of some variables. In
particular, the distance measures and the technological distance become insignificant. However, this
may be due to changes in the samples since there is limited availability of MNE data across countries
and industries in the OECD AMNE database. The AMNE information is typically more available for
EU countries and the United States, for which on average the variability in the above mentioned
variables is much smaller 25.
In order to allow for potential temporal effects, the trade variables have been lagged with one to
three years as one could argue that familiarity in trading networks is not instantaneous and takes some
time before partners getting to know each other better. The results are presented in Annex 2 and
confirm the importance of trade relationship for collaboration in innovation across different time
periods. However, contrary to what could be expected, the effect of trade on co-invention seems to

become somewhat less important over time. One reason for this result could be the changes in trade
relationships that happen over time with partners setting up new trade links and stopping existing
ones. 26
Lastly, estimations have been also undertaken using value added exports instead of gross exports: if
value added trade would reflect a more intensive relationship between partners than gross trade, a
stronger effect co-invention could be expected. Again, the results confirm the link between GVCs and
GINs but no significant differences are found in the two types of trade relationships. The results are
presented in Annex 3.
4. From GINs to GVCs: does co-invention induce GVC upgrading?
The links between GVCs and GINS may also go in the opposite direction due to the existence
international spillovers of technology and knowledge. The endogenous growth literature has discussed
the role of international knowledge flows for economic growth: spillovers from investments and
activities (in knowledge) in one location and by one party may result in beneficial effects for other
economic actors in other locations. Different channels of international spillovers have been analysed,
amongst which most prominently trade and foreign direct investment (FDI). Empirical research has
shown that imports and exports generate international spillovers as products and services embody
technological knowledge (Coe et al., 1997; Keller, 2004). Likewise, the growing trade of
intermediates within GVCs reflects the embodiment of the skills, factors and technologies used to
produce them and therefore can be a form of technology transfer from abroad to local firms (Amiti and
Konings, 2007; Bas and Strauss-Kahn, 2015; Halpern et al., 2015).
International investment may also act as a channel of technology/knowledge diffusion since local
firms may learn about advanced technologies or good organisational and managerial practices of
foreign MNEs (Caves, 1996). A large empirical literature has investigated FDI spillovers largely
showing that positive spillovers emerge particularly through backward linkages; the evidence on
spillovers through other MNE linkages is more mixed (see Alfaro (2014) for an overview of the
evidence and issues at stake).
A disembodied channel of technology diffusion is the (international) collaboration in innovation

between partners. Technological knowledge often includes non-codified and tacit knowledge which is
overall more difficult and costly to transfer. Close interactions and face-to-face contacts facilitate the
transfer of knowledge and this is largely the reason why spillovers tend to be localised and often
require absorptive capacity (Cohen and Levinthal, 1989; Griffith et al., 2004; Lyachagin et al., 2016).
Knowledge may also diffuse more rapidly when interpersonal links through joint work and co-
operation create opportunities for learning which go beyond the exchange of codified information.
Recent research shows that companies and countries that are connected in international collaboration
are more innovative (Breschi and Lissoni, 2009; Singh, 2005; Hoekman et al. (2009).
The question addressed in this paper is then whether GINs and co-inventions in patents help
countries to innovate and upgrade in GVCs. Will the GVC activities of countries benefit from their
interactions in GINs, as collaboration in innovation provides access to foreign knowledge and
technology and thus may give rise to international spillovers? The value created in GVCs is unevenly
distributed and depends on the ability of participants to supply sophisticated and hard-to-imitate
products and services. Increasingly, such products or services stem from forms of KBC such as
brands, basic R&D, intellectual property like patents, etc. OECD (2013b) has shown that KBC is an
important driver of GVC upgrading, which enables companies and countries to create and capture
more value and improve their position in GVCs. This paper aims to extend this discussion and include

the effect of international knowledge spillovers (through co-invention patterns) – in addition to own
investments and activities in knowledge - on the upgrading of countries within GVCs.
Innovation and KBC are closely linked to each of the different patterns of upgrading – process,
product, functional and chain upgrading (OECD, 2013a). In general, upgrading trajectories at the
country level become apparent when looking at the domestic value added content of countries’ export.
This indicator 27 provides insights into how much value an economy creates from its exports and is
overall the result of the different upgrading strategies of companies located within its borders. The
base empirical model used in this paper links then the change in the domestic content of exports to
own investments in KBC (proxied by own patent stock) and international knowledge spillovers (which
in line with the empirical spillover literature are calculated as the weighted average of the foreign
partners’ patent stocks, with the weights being the co-invention patterns across partners 28):
dva coinvijst
∆ a 0 + a1 pstockist + a 2 ∑ (
= * pstock jst ) + e (4)
exp ortsist coinvist
with i = country i, j = partner country j, s = industry and t = year
The results presented in Table 2 primarily demonstrate the importance of own investments in
KBC of countries for GVC upgrading; this is in line with OECD (2013a) which showed the positive
effect of countries’ KBC on the export competitiveness and upgrading of these countries within
GVCs. On the contrary, the results do not seem to indicate any effect of GINs via international
spillovers on the GVC upgrading of countries. Moreover, after taking into account a number of factors
that have been discussed in the spillover literature to matter and mediate international knowledge
flows, no real evidence is found for the contribution of co-invention networks on GVC upgrading.
First, correcting for the physical and time distance between partners in collaborations (models 1 and 2)
– as the previous analysis has shown the importance of both types of distance for co-invention
patterns, does not significantly change the results. Second, the types of partners do not appear to play
a major role (model 3); international spillovers do not seem to differ between international
collaborations with stronger partners and those with more similar partners. Third, taking into account
the importance of absorptive capacity (model 4) – through own investments - does neither show a
significant effect from GINs to GVCs. 29
Annex 4 presents the econometric results for the individual industries of chemicals (ISIC 23-26),
electronics (ISIC 30-33) and business services (ISIC 70-74), i.e. industries with a significant number
of co-inventions. Results for the chemicals and business services industries are in line with the
reported results above finding no evidence of an effect of co-invention on the GVC-upgrading in
countries. Results for the electronics industry show - surprisingly - rather a negative effect of co-
invention on GVC upgrading; more research is needed to identify the reasons behind this, like for
example the results that co-invention is more likely between more technologically distant partners

Table 2. Upgrading model: econometric results
Own Knowledge 0.001 0.013*** 0.009** 0.009** 0.009** 0.008* 0
(0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (.)
Coinvention * Partner Knowledge 0.002** 0.001 -0.002 -0.012**
(0.00) (0.00) (0.00) -(0.01)
Co-invention*Partner Knowledge*Distance -0.001
(0.00)
Coinvention* Partner Knowledge * Time difference -0.001
(0.00)
Coinvention * Partner Knowledge * (Partner/Own knowledge) -0.001
(0.00)
Coinvention * Partner Knowledge*Own Knowledge 0.009**
(0.00)
Control Variables
Country*sector      
Year     
R-squared 0.004 0.43 0.439 0.439 0.439 0.439 0.439
Observations 1,607 1,607 1,607 1,607 1,607 1,607 1,607
Note: level of significance * p<0.10, ** p<0.05, *** p<0.01
Overall the empirical findings in this paper do not support the importance of co-invention within
GINs for success in GVCs. However, when interpreting these results a number of important points
have to be made, suggesting the need for further research to disentangle the complex
interdependencies between GINs and GVCs:
a) the empirical literature on spillovers is broad and extensive and has, to say the least,
produced diverse results. Spillovers along the different channels discussed above range
from negative in a number of analyses (usually in the case of MNE spillovers in some
developing economies) to positive in relatively more studies. A publication bias in
spillover results has however been observed with a tendency to publish results if studies
are able to reject the null hypothesis, that is, when the investigations produce positive
and statistically significant findings (Havranek, 2012; Demena and van Bergeijk, 2016).
b) most of the empirical research on knowledge and technological spillovers has focused
on productivity (particularly via trade and FDI) and innovative outcomes (for
disembodied channels) while studies on the impact of spillovers on GVC upgrading are
non-existent. Indeed, the concept of GVC upgrading is rather broad given the different
upgrading patterns; consequently, empirical indicators on GVC upgrading are not
uniformly defined 30 and have been used only recently.
c) the spillover literature has shown that timing issues matter in assessing the beneficial
effects of international knowledge flows with some spillovers only materialising after
quite some time. A general empirical strategy has not been established in order to
capture these timing effects, but the availability of TiVA data has somewhat limited the
empirical analysis in this paper. Furthermore, the TIVA data encompass relatively
different time periods for GVCs, with the periods 1995-2000 and especially 2000-2005
showing a significant expansion of GVCs while the economic crisis has resulted in a
consolidation of GVCs in the period 2005-2010.

d) investment in knowledge – own as well as partners’ – in this analysis are proxied by

patent stocks; as mentioned already above, recent OECD work has shown the need to
analyse KBC from a broader perspective beyond R&D and intellectual property. Also
the traditional disadvantages of patent data – e.g. differences in patent propensity across
industries and countries, have to be taken into account when interpreting the results.
e) the analysis in this paper has been undertaken at the country-industry level which
overall may be too broad to capture spillover effects; recent research has shown the
importance of firm-level characteristics for assessing the effects of international
collaboration in innovation.
5. Conclusions and some policy implications
In exploring the interdependencies between GVCs and GINs at the country level, this paper has
reported on a number of clear links between both types of networks. At the same time, as it became
clear in the discussion of the results, the analyses in this paper should be regarded as only a first step
and more research is needed in order to better and fully understand the linkages between GVCs and
GINs. Several directions for future research can be proposed: complementing the aggregate analysis
with more firm-level analysis; discussing different types of GINs (the above-mentioned work on the
links between GVCs and KBC goes in this direction); and, the discussion of timing issues, etc.
Despite the limitations, the paper has provided evidence on the growing importance of
international linkages in the innovation landscape of countries; companies are increasingly opening
their innovation process and collaborating with other partners across borders. The number of
international co-inventions in patent data shows a gradual increase over the past two decades, although
the economic crisis of 2009 has slowed this internationalisation somewhat. Nevertheless, in
developing an effective innovation strategy, policy makers will have to increasingly take into account
this international perspective; an overly inward-looking perspective in innovation may isolate
domestic companies in the global innovation process.
While previous OECD work had already pointed to the growing internationalisation of
innovation, this paper additionally shows the importance of MNEs in this development. As almost
two-thirds of international co-inventions during the period 1995-2013 are directly linked to MNE
activities – and more than 50% of the international co-inventions are held by one MNE, it is clear that
GINs are to a large extent structured within and around MNE networks. Also this is an important
point to take into account when developing and implementing government policies; connection to such
MNE networks by attracting foreign MNEs and helping to grow incumbent companies to become
MNEs, needs to be an essential part of an effective government policy.
Another important observation emerging from this paper, is that distance is not dead yet. Even in
a time with extensive and advanced communication possibilities, distance still seems to matter. GVCs
and GINs both show a strong regional character with trade and co-invention relationships still strongly
concentrated within supra-national regionals (North America, Europe and Asia), notwithstanding the
growing importance of extra-regional links. Internationalisation strategies and policies will have to
take into account both the costs of distance due to face-to-face interaction costs (as reflected by
geographic distance) and virtual interaction costs (as reflected by time distance).
Despite the importance of distance, the geography of both GVCs and GINs has been changing
with networks becoming denser and linkages stretching across a growing number of countries
including emerging economies. While there are clearly some parallels to be drawn between the
development of GINs and GVCs in recent years, there are nevertheless a number of clear differences

between both types of networks. First, GINs are not as intensive as GVCs; second, GVCs exhibit a
more stable pattern as the economic globalisation through trade has already been in place for many
decades.
In general, GVCs and GINs show a strong overlap in geographic concentration with hubs in
GVCs often also being hubs in GINs. Questions have been raised in policy discussions about the
localised impact of STI, i.e. countries are not always benefitting from the advances in STI (including
the government support for R&D). The aggregate analysis (on the supra-region and country level) in
this paper does not seem to support this claim very strongly. That said, it is clear that a more
disaggregated analysis (on the sub-regional and company level) is desirable and necessary to address
this important policy issue in more detail.
One link between GVCs and GINs clearly demonstrated in this paper is the strong positive effect
reported going from GVCs towards GINs. Partners who are trading with each other – be it via exports
or imports - are found to cooperate more intensively also in GINs. Collaborating within a trade
relationship offers the possibility to identify new partners and get more familiar with these partners.
This directly means that apart from S&T policies, also trade policies can have a role to play for the
integration of countries in GINs; more research is needed to assess the relative importance of different
government policies.
Indeed, previous OECD work has shown that a broad range of STI policies can promote the
(international) co-operation in science, technology and knowledge. As STI policies can no longer be
designed solely in a national context, government support for R&D innovation, strong universities and
other public research institute, a modern IPR system, etc. all should reflect the increasing open and
global character of technology and knowledge. In contrast, the paper was not able to report evidence
of the effect of GINs on GVCs; the fact that a country is heavily involved in GINs does not directly
help that country to upgrade in GVCs. In the first place, this shows the broader and complex
determinants of GVC upgrading in which GINs could play a role. A number of shortcomings have
been discussed largely reflecting the difficulty often encountered in empirical work to demonstrate the
effects of STI on economic performance and the impacts of international spillovers. As such, the (lack
of) evidence in this paper should not be interpreted as an indication of the importance of GINs for the
economic performance of companies, industries and countries. Rather, it demonstrates a potential area
for future research because its direct importance for policy.

NOTES
1
The first type might be easily assimilated with the material flows of intermediate and final products,
while the second one focuses more on the immaterial flow of ideas, knowledge and technology.
However, GVCs are broader than supply chains, including also the upstream activities of R&D,
design and so on; explaining to some extent the different performance of countries in GVCs.
2
Based on the OECD Inter-Country Input-Output (ICIO) system which links national IO tables of 61
countries across 34 industries.
3
Data on licensing can also be used to document the trend to open innovation but are overall rather
limited and lack uniformity. As most patent licensing is based on private contracts that are subject to
confidentiality agreements, robust statistics on technology licensing are not available. Furthermore,
accounting rules do not require firms to disclose patent licensing revenues as a separate item in
corporate reports.
4
Patents protect technological inventions, i.e. products or processes that provide new ways of doing
something or new technical solutions to problems.
5
Patent indicators constructed on the basis of information from a single patent office typically suffer
from a “home advantage” bias in proportionate to their inventive activity, domestic applicants tend to
file more patents in their home country (or region) than non-resident applicants
6
Inventors are physical persons from which it is not clear in which company and industry they work.
7
The reason for using applicants’ information is that inventors are typically physical persons and no
direct information is available if they work for or are employees of a MNE; instead, if the applicant is
found to be part of a multinational group, the PCT application and co-invention is assumed to belong
to MNEs.
8
This has been done by linking information on the ownership structure – available in the ORBIS
database –with the HAN database.
9
For about one quarter of the PCT applications as well as international co-inventions, no information
on the applicants was included in the HAN and ORBIS; given that both databases are more
representative for larger companies including MNEs, one could assume that this category of “no
information” concerns relatively more non-MNEs.
10
Some caution has to be taken into account when interpreting these data as the information on the
global ultimate owner may not always be available and accurate for some MNEs.
11
Abstraction made from the group of PCT patent that cannot be assigned to one country.
12
Which is of course due to the smaller size of the countries (relative to the United States for example).
13
Co-invention is calculated by adding the sum of co-invention shares between country-pairs divided by
the sum of domestic inventions in these respected countries for industry i at time t . Similarly, trade
intensity is estimated by adding the exports between respected trading partners out of the sum of
output for specific industries at a particular time. While these measures are largely similar, it should
be taken into account that co-invention is per se non-directional and largely symmetrical (apart from

the factional counts when more inventors in one country), exports are directional and non-
symmetrical.
14
The time periods considered are 2000-2005 and 2008-2011; the choice for the last period is to prevent
a possible bias in the data because of the economic crisis in 2008.
15
The co-invention intensity and trade intensity by partners are presented only at the industry level;
because of the multitude of bilateral relations, a similar graph on the economy-wide level is not
feasible.
16
OECD work has showed that services are to a large extent included in the exports of goods, i.e. the
gross value of manufactures exports includes a large share of value added that has been created in
services.
17
The gravity model is the workhorse model of international trade, going back to a contribution by
Tinbergen (1962); see Anderson and van Wincoop (2003) for a discussion of the theoretical
foundations of gravity modelling.
18
Although OLS estimations are less appropriate given the large number of zeros in the dependent
variable, the empirical model has also been estimated using OLS as robustness test. The OLS results
(not reported here) are largely in line with the PPML results and subscribe in particular the importance
of trade relations between countries for co-invention.
19
Using patent stocks as indicator proxy for countries’ innovation activities represents only a (rough)
proxy as recent OECD work on Knowledge Based Capital (KBC) has shown that innovation is much
broader than intellectual property (OECD, 2013b).
20
Another problem is directly linked to the disadvantages of patent data as patenting is not important in
all industries; R&D could be used as an alternative indicator of KBC but also R&D data are less
available for emerging economies.
21
Source: CEPII database based on km between capitals.
22
As the variables are expressed in logarithms, a dummy variable is included to retain observations
between country pairs in the same time zone.
23
The different relationship appear robust to the inclusion of both year and importer (partner-industry)
and exporter (country-industry) fixed effects
24
Because of more limited data availability of MNE presence, the analysis takes necessarily into
account a smaller number of observations.
25
Also the no-trade variable is dropped from the model as all observations with positive MNE turnover
show also positive bilateral trade.
26
Nevertheless, previous research has shown that changes in trade happen largely at the intensive
margin (changes in exports/imports volumes with existing partners) rather than at the extensive
margin (new partners).
27
It should however be stressed that this is only one possible indicator of GVC upgrading; there is a
broad discussion how upgrading can be assessed by which indicators and at what level of analysis
(country, industry, company, etc.).

28
More specifically, the weights – calculated at the industry level - are the shares of the co-inventions of
the country in question with each partners divided by all the co-inventions of that country across
partners. Because of the availability of the TiVA data, variables have been aggregated over 5 years
(1995-2000, 2000-2005 and 2005-2010).
29
Taking into account the individual and interaction effects simultaneously.
30
Additional estimations were undertaken with the growth in domestic value added in exports (in
absolute terms) and the growth in value added (irrespective if the value added was created in
exporting activities or activities for the domestic market); the results of international knowledge flows
for GVC upgrading were however also for these dependent variables, non-significant.

REFERENCES
Alfaro, L. (2014), Foreign Direct Investment: Effcets, Complementarities and Promotion, Harvard
Business School Working Papers, No. 15-006
Amiti, M. and J. Konings (2007), “Trade Liberalization, Intermediate Inputs, and Productivity:
Evidence from Indonesia”, American Economic Review, Vol. 97, No. 5, pp. 1611–38
Anderson, J.E. and E. van Wincoop (2003), “Gravity with Gravitas: A Solution to the Border Puzzle”,
American Economic Review, Vol. 93 (1), pp. 170-192
Bas, M. and V. Strauss-Kahn (2015), “Input-trade liberalization, export prices and quality upgrading”
Journal of International Economics, Vol. 95, No. 2, pp. 250-262
Belderbos, R., Sleuwaegen, L., Somers, D. and K. De Backer (2016), “Where To Locate Innovative
Activities in Global Value Chains? Does Co-Location Matter?”, OECD Science, Technology
and Industry Policy Papers, No. 30, OECD Publishing, Paris;
http://dx.doi.org/10.1787/5jlv8zmp86jg-en
Breschi, S. and F. Lissoni (2009), “Mobility of Skilled Workers and Co-Invention Networks: An
Anatomy of Localized Knowledge Flows”, Journal of Economic Geography, 9 (4), pp. 439–468
Caves, R.E. (1996), Multinational Enterprise and Economic Analysis, Cambridge University Press
Coe, D., E. Helpman and A.W. Hoffmaister (1997), “North–South R&D Spillovers”, The Economic
Journal, 107 (440), pp. 134–149
Cohen, W.M. and D.A. Levinthal (1989), “Innovation and Learning: The Two Faces of R&D”,
Economic Journal, 99 (1989), pp. 569–596
De Backer, K. and S. Miroudot (2013), “Mapping GVCs”, OECD Trade Policy Papers No. 159,
OECD Publishing, Paris; http://dx.doi.org/10.1787/5k3v1trgnbr4-en
Demena, B.A. and P.A.G. van Bergeijk (2015), A Meta-Analysis of FDI and Productivity Spillovers
in Developing Countries, Journal of Economic Surveys, forthcoming.
Dernis, H. and M. Khan (2004). "Triadic Patent Families Methodology," OECD Science, Technology
and Industry Working Papers 2004/2, OECD Publishing, Paris.
Griffith, R., S. Redding, J. Van Reenen (2004), “Mapping the Two Faces of R&D: Productivity
Growth in a Panel of OECD Industries”, Review of Economics and Statistics, 86 (4) (2004), pp.
883–895

Griliches, Z. (1990). “Patent Statistics as Economic Indicators: A Survey”, Journal of Economic

Literature, 28(4), 1661-1707
Guellec, D. and B. Van Pottelsberghe (2001), “The Internationalisation of Technology Analyzed by

Patent Data, Research Policy, 30 (8), pp. 1253–1266
Halpern, L., M. Koren and A. Szeidl (2015), “Imported Inputs and Productivity”, American Economic
Review, Vol. 105, No. 12, pp. 3660-3703
Havranek, T. and Z. Irsova (2012). "Survey Article: Publication Bias in the Literature on Foreign
Direct Investment Spillovers," Journal of Development Studies, vol. 48(10), pages 1375-1396,
Haščič, I., N. Johnstone and N. Kahrobaie (2012), "International Technology Agreements for Climate
Change: Analysis Based on Co-Invention Data," OECD Environment Working Papers, No. 42,
OECD Publishing, Paris; http://dx.doi.org/10.1787/5k9fgpw5tt9s-en
Hoekman, J., K. Frenken, R. Tijssen (2010), “Research Collaboration at a Distance: Changing Spatial
Patterns of Scientific Collaboration within Europe”, Research Policy, 39, pp. 662–673
Jaffe, A. B. (1989). “Characterizing the ‘Technological Position’ of Firms, with Application to

Quantifying Technological Opportunity and Research Spillovers”, Research Policy, 18(2), 87-
97
Keller, W. (2004), “International Technology Diffusion”, Journal of Economic Literature, 42 (3), pp.
752–782
Lychagin, S., J. Pinkse, M.E. Slade and J. Van Reenen (2016), “Spillovers in Space: Does Geography
Matter?”, The Journal of Industrial Economics, Vol. 64 (2), pp. 295–335
Maggioni, M.A., M. Nosvelli and T.E. Uberti (2007), ”Space vs. Networks in the Geography of
Innovation: A European Analysis”, Papers in Regional Science, 86 (3) (2007), pp. 471–493
Montobbio, F. and V. Sterzi (2013), “The Globalisation of Technology in Emerging Markets: A

Gravity Model on the Determinants of International Patent Collaborations”, World
Development, Vol. 44, pp. 281-299
Montobbio, F. and V. Sterzi (2011), “Inventing Together: Exploring the Nature of International
Knowledge Spillovers in Latin America”, Journal of Evolutionary Economics, 21 (1) (2011),
pp. 53–89
OECD (2013a), Interconnected Economies: Benefitting From Global Value Chains, OECD
Publishing, Paris; http://dx.doi.org/10.1787/9789264174443-zh
OECD (2013b), Supporting Knowlegde Based Capital, Growth and Innovation, OECD Publishing,
Paris;http://dx.doi.org/10.1787/9789264193307-en
OECD (2009), OECD Patent Statistics Manual, OECD Publishing, Paris;

http://dx.doi.org/10.1787/9789264056442-en
OECD (2008), Open Innovation in Global Networks, OECD Publishing, Paris;

http://dx.doi.org/10.1787/9789264047693-en

OECD (2008), The Internationalisation of Business R&D: Evidence, Impacts and Implications, OECD
Publishing, Paris. http://dx.doi.org/10.1787/9789264044050-en
Picci, L. (2010), “The Internationalisation of Inventive Activity: A Gravity Model Using Patent Data”,
Research Policy, 39 (8) (2010), pp. 1070–1081
Singh, J. (2005), “Collaborative Networks as Determinants of Knowledge Diffusion Pattern”,

Management Science, 51 (5), pp. 756–770
Sleuwaegen, L. and K. De Backer (2001), “Multinational firms, market integration, and trade
structure: What remains of the standard-goods hypothesis?”, Weltwirtschaftliches Archiv, 137
(3), pp. 379-403
Stellner, F. (2014), Technological Distance Measures: Theoretical Foundation and Empirics Paper
Presented at the DRUID Society Conference 2014

ANNEX 1: CO-INVENTION AND TRADE INTENSITY, BY COUNTRY,

PARTNER AND INDUSTRY
Agriculture, hunting, forestry and fishing (ISIC 01 to 05), 2000-05
International co-invention intensity1
Trade intensity2


Agriculture, hunting, forestry and fishing (ISIC 01 to 05), 2008-11
Trade intensity2


Mining and quarrying (ISIC 10 to 14), 2000-05
Trade intensity2


Mining and quarrying (ISIC 10 to 14), 2008-11
Trade intensity2


Food products, beverages and tobacco (ISIC 15 to 16), 2000-05
Trade intensity2


Food products, beverages and tobacco (ISIC 15 to 16), 2008-11
Trade intensity2


Trade intensity2


Trade intensity2


Wood, paper, paper products, printing and publishing (ISIC 20 to 22), 2000-05
Trade intensity2


Wood, paper, paper products, printing and publishing (ISIC 20 to 22), 2008-11
Trade intensity2


Chemicals and non-metallic mineral products (ISIC 23 to 26), 2000-05
Trade intensity2


Chemicals and non-metallic mineral products (ISIC 23 to 26), 2008-11
Trade intensity2


Basic metals and fabricated metal products (ISC 27 to 28), 2000-05
Trade intensity2


Basic metals and fabricated metal products (ISC 27 to 28), 2008-11
Trade intensity2


Machinery and equipment, nec (ISIC 29), 2000-05
Trade intensity2


Machinery and equipment, nec (ISIC 29), 2008-11
Trade intensity2


Trade intensity2


Trade intensity2


Transport equipment (ISIC 34 to 35), 2000-05
Trade intensity2


Transport equipment (ISIC 34 to 35), 2008-11
Trade intensity2


Manufacturing nec; recycling (ISIC 36 to 37), 2000-05
Trade intensity2


Manufacturing nec; recycling (ISIC 36 to 37), 2008-11
Trade intensity2


Electricity, gas and water supply (ISIC 40 to 41), 2000-05
Trade intensity2


Electricity, gas and water supply (ISIC 40 to 41), 2008-11
Trade intensity2


Construction (ISIC 45), 2000-05
Trade intensity2


Construction (ISIC 45), 2008-11
Trade intensity2


Wholesale and retail trade; Hotels and restaurants (ISIC 50 to 55), 2000-05
Trade intensity2


Wholesale and retail trade; Hotels and restaurants (ISIC 50 to 55), 2008-11
Trade intensity2


Transport and storage, post and telecommunication (ISIC 60 to 64), 2000-05
Trade intensity2


Transport and storage, post and telecommunication (ISIC 60 to 64), 2008-11
Trade intensity2


Financial intermediation (ISIC 65 to 67), 2000-05
Trade intensity2


Financial intermediation (ISIC 65 to 67), 2008-11
Trade intensity2


Trade intensity2


Trade intensity2


Community, social and personal services (ISIC 75 to 95), 2000-05
Trade intensity2


Community, social and personal services (ISIC 75 to 95), 2008-11
Trade intensity2


ANNEX 2: CO-INVENTION GRAVITY MODEL WITH DIFFERENT TIMING OF TRADE
Timing of trade t=1 t-1 t-2 t-3

Bilateral Trade 0.269*** 0.203*** 0.126*** 0.071***
(0.01) (0.01) (0.01) (0.01)
No trade -3.847*** -3.936*** -2.428*** -0.642
(0.75) (0.74) (0.52) (0.41)
Patent Stock country 0.469*** 0.453*** 0.465*** 0.492***
(0.03) (0.03) (0.03) (0.03)
Patent Stock partner 0.271*** 0.257*** 0.280*** 0.308***
(0.04) (0.04) (0.03) (0.03)
Physical distance -0.035* -0.051*** -0.071*** -0.087***
(0.02) (0.02) (0.02) (0.02)
Time difference -0.041** -0.070*** -0.102*** -0.125***
(0.02) (0.02) (0.02) (0.02)
Same time zone 0.159*** 0.155*** 0.151*** 0.149***
(0.03) (0.03) (0.03) (0.03)
Technological distance 0.015*** 0.013** 0.011* 0.008
(0.01) (0.01) (0.01) (0.01)
Contiguity (dummy) 0.370*** 0.392*** 0.416*** 0.434***
(0.03) (0.03) (0.03) (0.03)
Language (dummy) 0.665*** 0.689*** 0.716*** 0.736***
(0.03) (0.03) (0.03) (0.03)
Colony (dummy) -0.118*** -0.116*** -0.114*** -0.115***
(0.03) (0.03) (0.03) (0.03)
Control Variables
Year    
Origin/Destination fe's    
R-squared 0.891 0.889 0.888 0.888
Observations 173,918 173,918 173,918 173,918
Note: No trade is a dummy which equals to one when there is zero trade from a country to a
partner in a particular industry and zero otherwise. Significance: * p<0.10, ** p<0.05, *** p<0.01

ANNEX 3: CO-INVENTION GRAVITY MODEL WITH VALUE ADDED TRADE
Patent Stock country 0.500*** 0.527*** 0.468*** 0.468*** 0.468*** 0.464*** 0.477***
(0.01) (0.01) (0.03) (0.03) (0.03) (0.03) (0.06)
Patent Stock partner 0.443*** 0.468*** 0.269*** 0.269*** 0.269*** 0.266*** 0.367***
(0.01) (0.01) (0.04) (0.04) (0.04) (0.04) (0.06)
Distance -0.110*** -0.110*** -0.077*** -0.034* -0.035* -0.039
(0.01) (0.01) (0.01) (0.02) (0.02) (0.03)
Time difference -0.064*** -0.039** -0.039** 0.037
(0.01) (0.02) (0.02) (0.04)
Same time difference 0.164*** 0.161*** 0.162*** 0.104**
(0.03) (0.03) (0.03) (0.05)
Contiguity (dummy) 0.314*** 0.318*** 0.384*** 0.408*** 0.380*** 0.372*** 0.370***
(0.04) (0.04) (0.03) (0.03) (0.03) (0.03) (0.05)
Language (dummy) 0.934*** 0.953*** 0.704*** 0.666*** 0.667*** 0.662*** 0.599***
(0.03) (0.03) (0.02) (0.03) (0.03) (0.03) (0.04)
Colony (dummy) 0.042 0.014 -0.104*** -0.123*** -0.117*** -0.118*** -0.033
(0.04) (0.04) (0.03) (0.03) (0.03) (0.03) (0.04)
Technological distance 0.015** 0.001
(0.01) (0.01)
MNE Turnover 0.014*
(0.01)
DVA in Trade 0.300*** 0.280*** 0.278*** 0.274*** 0.270*** 0.273*** 0.339***
(0.01) (0.01) (0.01) (0.01) (0.01) (0.01) (0.02)
No Dva in Trade -3.053*** -2.985*** -3.988*** -3.951*** -3.957*** -3.954***
(0.71) (0.71) (0.75) (0.74) (0.75) (0.75)
Control Variables
Year      
Origin/Destination fe's     
R-squared 0.652 0.664 0.889 0.891 0.891 0.891 0.926
Observations 199,534 199,534 174,132 174,132 174,132 173,918 35,665
Note: No DVA Trade is a dummy which equals to one when there is zero DVA in trade between country and partner pairs in a particular
industry and zero otherwise. Significance: * p<0.10, ** p<0.05, *** p<0.01

ANNEX 4: UPGRADING MODEL FOR INDIVIDUAL INDUSTRIES
Chemicals
Own Knowledge 0.000 0.025* -0.015 -0.015 -0.014 -0.019 0
(0.00) (0.01) (0.02) (0.02) (0.02) (0.03) (.)
Coinvention*Partner Knowledge 0.015** 0.017 -0.001 0.014
(0.01) (0.01) (0.01) (0.02)
Coinvention*Partner Knowledge *Distance 0.000
(0.01)
Coinvention*Partner Knowledge *Time 0.001
(0.01)
Coinvention*Partner Knowledge *(Partner/Own
knowledge) -0.002
(0.01)
Coinvention*Partner Knowledge *Own knowledge -0.015
(0.02)
Control Variables
Country      
Year     
R-squared 0.031 0.327 0.371 0.371 0.371 0.372 0.371
Observations 135 135 135 135 135 135 135
Electronics
Own Knowledge 0.011** 0.080*** 0.065*** 0.068*** 0.066** 0.041 0.122***
(0.00) (0.02) (0.02) (0.03) (0.03) (0.03) (0.03)
Coinvention*Partner Knowledge 0.019 -0.039* -0.058** 0
(0.01) (0.02) (0.03) (.)
Coinvention*Partner Knowledge *Distance -0.008
(0.03)
Coinvention*Partner Knowledge *Time -0.018
(0.03)
knowledge) -0.027*
(0.01)
Coinvention*Partner Knowledge *Own knowledge -0.058**
(0.03)
Control Variables
Country      
Year     
R-squared 0.076 0.472 0.481 0.45 0.453 0.474 0.481
Observations 124 124 124 124 124 124 124

Business services
Own Knowledge 0.001 0.014* 0.010 0.011 0.010 0.004 0
(0.00) (0.01) (0.01) (0.01) (0.01) (0.01) (.)
Coinvention*Partner Knowledge 0.000 -0.008 -0.012 -0.022
(0.00) (0.01) (0.01) -0.01
Coinvention*Partner Knowledge *Distance 0.001
(0.01)
Coinvention*Partner Knowledge *Time -0.002
(0.01)
knowledge) -0.006
(0.00)
Coinvention*Partner Knowledge *Own knowledge 0.010
(0.01)
Control Variables
Country      
Year     
R-squared 0.002 0.401 0.404 0.39 0.391 0.402 0.404
Observations 152 152 152 152 152 152 152

OECD 2017 The Links Between Global Value Chain

Uploaded by

Copyright:

Available Formats

OECD 2017 The Links Between Global Value Chain

Uploaded by

Document Information

Original Title

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

OECD 2017 The Links Between Global Value Chain

Uploaded by

Copyright:

Available Formats

the links between

OECD SCIENCE, TECHNOLOGY

2 OECD SCIENCE, TECHNOLOGY AND INDUSTRY POLICY PAPERS

OECD SCIENCE, TECHNOLOGY AND INDUSTRY POLICY PAPERS 3

4 OECD SCIENCE, TECHNOLOGY AND INDUSTRY POLICY PAPERS

OECD SCIENCE, TECHNOLOGY AND INDUSTRY POLICY PAPERS 5

THE LINKS BETWEEN GLOBAL VALUE CHAINS AND GLOBAL INNOVATION

1. Introduction: the globalisation of economic activities

6 OECD SCIENCE, TECHNOLOGY AND INDUSTRY POLICY PAPERS

2. Mapping of GVCs and GINs

2.1. Empirical measures of GVCs and GINs

OECD SCIENCE, TECHNOLOGY AND INDUSTRY POLICY PAPERS 7

8 OECD SCIENCE, TECHNOLOGY AND INDUSTRY POLICY PAPERS

Box 1. Some tentative conclusions about open innovation and GINs

OECD SCIENCE, TECHNOLOGY AND INDUSTRY POLICY PAPERS 9

Figure 1. International co-invention, absolute numbers and intensity, 1995-2013

10 OECD SCIENCE, TECHNOLOGY AND INDUSTRY POLICY PAPERS

OECD SCIENCE, TECHNOLOGY AND INDUSTRY POLICY PAPERS 11

Figure 3. The importance of MNEs in patenting and international co-invention, 1995-2013

PCT patents international co-inventions

PCT patents international co-inventions

absolute numbers absolute numbers

200000 one MNE other

Source: calculations based on OECD patent, HAN and ORBIS databases.

2.2. The geography of co-invention and GINs

12 OECD SCIENCE, TECHNOLOGY AND INDUSTRY POLICY PAPERS

Figure 5. International co-invention by region (absolute numbers), 1995-2004 and 2005-2013

Source: calculations based on OECD patent database.

OECD SCIENCE, TECHNOLOGY AND INDUSTRY POLICY PAPERS 13

Figure 6. International co-invention intensity by country, 1995-2013

 Coinventionijk ,t + Coinvention jik ,t 

 Exportsijk ,t + Exports jik ,t 

14 OECD SCIENCE, TECHNOLOGY AND INDUSTRY POLICY PAPERS

Textiles, textile products, leather and footwear (ISIC 17 to 19), 2000-05

1. Bilateral co-inventions as a share of all inventions by both countries.

OECD SCIENCE, TECHNOLOGY AND INDUSTRY POLICY PAPERS 15

Textiles, textile products, leather and footwear (ISIC 17 to 19), 2008-11

1. Bilateral co-inventions as a share of all inventions by both countries.

16 OECD SCIENCE, TECHNOLOGY AND INDUSTRY POLICY PAPERS

Electrical and optical equipment (ISIC 30 to 33), 2000-05

1. Bilateral co-inventions as a share of all inventions by both countries.

OECD SCIENCE, TECHNOLOGY AND INDUSTRY POLICY PAPERS 17

Electrical and optical equipment (ISIC 30 to 33), 2008-11

1. Bilateral co-inventions as a share of all inventions by both countries.

18 OECD SCIENCE, TECHNOLOGY AND INDUSTRY POLICY PAPERS

Real estate, renting and business activities (ISIC 70 to 74), 2000-05

1. Bilateral co-inventions as a share of all inventions by both countries.

OECD SCIENCE, TECHNOLOGY AND INDUSTRY POLICY PAPERS 19

Real estate, renting and business activities (ISIC 70 to 74), 2008-11

1. Bilateral co-inventions as a share of all inventions by both countries.

Source: calculations based on OECD patent database

20 OECD SCIENCE, TECHNOLOGY AND INDUSTRY POLICY PAPERS

A number of common observations emerge from the figures:

3. From GVCs to GINs: does trade promote co-invention?

OECD SCIENCE, TECHNOLOGY AND INDUSTRY POLICY PAPERS 21

In addition, a measure of technological distance is included since previous research has

22 OECD SCIENCE, TECHNOLOGY AND INDUSTRY POLICY PAPERS

a 0 + a1 patstockist + a 2 patstock jst + a 3distanceij + a 4tdiffij +

with i = country i, j = partner country j, s = industry and t = year