International Research Networks: Determinants of Country Embeddedness

Full text

JENA ECONOMIC
RESEARCH PAPERS
# 2016 – 016
International Research Networks: Determinants of Country
Embeddedness
by
Holger Graf
Martin Kalthaus
www.jenecon.de
ISSN 1864-7057
The JENA ECONOMIC RESEARCH PAPERS is a joint publication of the Friedrich
Schiller University Jena, Germany. For editorial correspondence please contact
markus.pasche@uni-jena.de.
Impressum:
Friedrich Schiller University Jena
Carl-Zeiss-Str. 3
D-07743 Jena
www.uni-jena.de
© by the author.

International Research Networks: Determinants of Country
Embeddedness
Holger Graf *and Martin Kalthaus †
Friedrich Schiller University Jena, Department of Economics,
Carl-Zeiß-Straße 3, 07743 Jena, Germany.
*holger.graf@uni-jena.de
†martin.kalthaus@uni-jena.de
September 12, 2016
Abstract
We analyze the evolution of the international collaboration network in photovoltaic re-
search. Using data on scientific publications for the period 1980–2015, we apply social
network analysis to trace the evolution of the global network of countries and national re-
search networks of organizations. Our objective is to identify the determinants of countries’
international research embeddedness by looking at national policies and structural proper-
ties of the national research networks. We observe a steady increase of publications and
collaboration within the global research network. While there is a small group of countries
that remains central throughout all years, several countries emerge and catch up while others
lose their relative position.We find that cohesion and connectedness of the national system
positively affect research output as well as international embeddedness, whereas centralized
systems are less embedded. Policy, especially demand side instruments, has a positive effect
on publication output and embeddedness.
Keywords: International Collaboration; Research Network; Photovoltaics; Instrument
Mix; Bibliographic Data
JEL Classification: L14, O31, O38, Q42
1
Jena Economic Research Papers 2016 - 016

1 Introduction
The generation and diffusion of knowledge is an interactive process between a multitude of actors
(Dosi,1988;Powell et al.,1996). Connections to a diverse set of actors provide access to external
knowledge which is considered crucial for own research, successful innovation, and eventually
economic performance (Powell et al.,1999). However, access to external knowledge is not only
dependent on direct collaboration partners, but also influenced by indirect linkages, or more
generally, by the position in the knowledge network (Ahuja,2000;Schilling and Phelps,2007).
Geography is a relevant aspect in the analysis of collaborations and networks. While connections
to co-located actors might be more frequent and easier to establish, partners in distant locations
in the form of international collaborations are considered highly fruitful (Cantner and Rake,2014;
Herstad et al.,2014). Regarding the determinants of tie formation and network positions, several
theories have been developed in sociology, physics, management, or economics (for reviews see
Ozman,2009;Cantner and Graf,2011;Phelps et al.,2012;Hidalgo,2015). The decisions to
form ties in knowledge networks are typically determined by individual characteristics, such
as attractiveness in terms of capabilities, or by dyadic characteristics, such as geographical
distance. These decisions are taken by individuals or organizations where international linkages
connect different countries to form an international research network (Owen-Smith et al.,2002).
Knowledge development is an increasingly global phenomenon, so that it is economically relevant
for any country to be integrated in international research networks. Before this background,
policy can create an environment conducive to international collaboration which might lead to a
central position of the country within the international research network. Since the determinants
of international research embeddedness are not well understood, we want to fill this gap and
provide some novel perspectives on the analysis of international research networks.
We seek to identify the determinants of international embeddedness for the case of photo-
voltaic (PV) research. Given the rising awareness of climate change, research and development
(R&D) in environmentally friendly technologies is increasing, even though R&D in this field faces
several disadvantages (Rennings,2000;Jaffe et al.,2005). Renewable energies and especially PV
are generally seen as promising technologies to mitigate climate change. Since renewable ener-
gies have been – and still are – competing with existing technologies which produce electricity
at lower costs, many governments decided to support the development of renewable energies
by fostering R&D activities, providing investment subsidies, and/or promoting their diffusion
(Jaffe et al.,2002;Kemp and Pontoglio,2011;Groba and Breitschopf,2013). While there is
a growing literature evaluating the effect of policies on innovation and diffusion in PV (e.g.
Watanabe et al.,2000;Johnstone et al.,2010;Peters et al.,2012;Wangler,2013;Polzin et al.,
2015;Cantner et al.,2016), there are hardly any studies dealing with the influence of different
policy measures on scientific performance and the position in the international collaboration
network.
We argue that the position of a country in the international research network is influenced
by two driving forces. First, international embeddedness should be affected by the functionality
of its innovation system (Nelson,1993;Lundvall,1992;Carlsson and Stankiewicz,1991). We
focus on one particular aspect of the innovation system, namely its interaction structure, which
is highly relevant for knowledge diffusion within the system (Cowan and Jonard,2004;Schilling
2
Jena Economic Research Papers 2016 - 016

and Phelps,2007;Cantner and Graf,2011;Herstad et al.,2014). This argument is related to
the links between micro, meso, and macro levels of economic analysis (Dopfer et al.,2004).
Here, the structure of the national networks, i.e. the functionality and the way the research
system is set up, is a determinant of international collaboration behavior. In the empirical
analysis of this relationship, we exploit the multimodal structure in publication data and link
the national research network structure to the position of a country in the international research
network. We expect that different national strategies concerning its openness, connectedness,
and centralization should more or less conducive to international collaboration. Second, since a
great share of research is publicly financed, countries’ embeddedness is influenced by their policies
towards PV. The strategies towards PV differ sufficiently between countries in terms of activities
and timing which enables us to identify their impact on international embeddedness. A recently
emerging literature on the policy mix convincingly shows the relevance of various dimensions of
policy making (Flanagan et al.,2011;Rogge and Reichardt,2016). In this paper, we focus on
the instrument mix, covering direct R&D funding, demand pull instruments, but also a general
commitment to tackle climate change by the ratification of the Kyoto Protocol. Furthermore,
we extend this common set of policy variables by accounting for public procurement (proxied by
the cumulative number of satellites), which is especially relevant for research activities in early
phases of technology development (Geroski,1990;Edler and Georghiou,2007;Aschhoff and
Sofka,2009;Guerzoni and Raiteri,2015). We test the effect of the national network structure
and policy interventions on the embeddedness by OLS panel regression for the periods from
1980 until 2015.
We observe a steady increase of publications and collaboration within the global research
network. While there is a small group of countries that remains central throughout all years,
there are some countries catching up while others lose their relative position. We find that
cohesion and connectedness of the national system positively affect research output as well as
international embeddedness, whereas centralized systems are less embedded. Policy, especially
demand side instruments, have a positive effect on embeddedness.
The paper proceeds as follows. We review the literature and derive hypotheses in Section 2.
In Section 3, we first describe the publication data and then the international as well as the
national collaboration networks. In part 4, we present the econometric study where we estimate
the effects of the national network structure and different policies on the embeddedness of
countries. Our results are discussed in Section 5, Section 6concludes.
2 Literature review and research objectives
2.1 Networks of scientific collaboration
Knowledge generation is an interactive process in which the relationship between actors is key for
knowledge exchange and diffusion (Dosi,1988;Powell et al.,1996;Ahuja,2000). During the last
decades, collaboration in research has steadily increased and it has been shown to lead to more
valuable output than individual research (Adams et al.,2005;Wuchty et al.,2007;Adams,2013).
Increasing specialization and division of labor leads to larger teams and competence building and
sharing. However, researchers who collaborate, as documented e.g. by co-authorship, do not just
3
Jena Economic Research Papers 2016 - 016

add their individual expertise for a joint output but also exchange information and learn from
each other (Breschi and Lissoni,2004). As such, it is very common to treat authors as nodes
connected by joint publications in so called knowledge-networks. Such and similar networks are
frequently analyzed in the social sciences or in physics to identify universal structures, such as
small world properties, or test hypotheses regarding processes of network formation, such as
preferential attachment or homophily (Newman,2001;Barabasi et al.,2002).
Besides their structural properties, networks are also of interest because they provide infor-
mation about the position of individual nodes among a group of actors. Central positions might
indicate importance or power in a network by controlling information flows between otherwise
unrelated actors (Freeman,1979). In the above mentioned knowledge-networks, some positions
within the network might be better to access novel, external knowledge than others. Given
that external knowledge is a highly valuable input for processes of invention and innovation,
the question if and how network positions influence performance is widely studied especially in
management and economics. Based on various types of networks, this field of research produced
substantial empirical evidence showing that direct but also indirect connections matter for inno-
vation performance (for reviews see Ozman,2009;Cantner and Graf,2011;Phelps et al.,2012;
Hidalgo,2015).
2.2 Networks as multimodal structures
While interaction and learning takes place among individuals, these networks can be analyzed
at more aggregated levels to study interaction between groups of actors, such as organizations,
industries, or geographical levels. A critical assumption is that knowledge and information
are transmitted within nodes of a higher level of aggregation. At the organizational level,
one is interested in collaborations between organizations (affiliations of the researchers) while
knowledge flows within these organizations are assumed to be existent but usually not explicitly
taken into account (Cantner and Graf,2006;Adams et al.,2005;Guan et al.,2015a). Aggregation
can also account for the geographical dimension as in studies on international collaboration,
shedding light on knowledge flows between different regions (Wanzenb¨ock et al.,2014,2015) or
countries (Owen-Smith et al.,2002;Wagner and Leydesdorff,2005;Cantner and Rake,2014).
Figure 1displays the different levels or modes of networks that are used in the present study.
Raw publication data is on the micro level and provides information about co-authorship between
individuals. Information on the affiliations of the researchers is used to aggregate them to the
meso level. These networks between organizations will be studied for each country separately
to characterize the national research and innovation systems. By using information on the
location of organizations, we reconstruct global networks – the macro level – of international
collaboration. The position of countries within these networks provides valuable information
about international embeddedness and (potential) access to global knowledge flows.
Our research aims at explaining differences between countries in terms of international em-
beddedness by looking at the structure of the national research systems as well as various other
policies towards PV employed by national governments. As such, we contribute to the literature
by linking meso and macro structures (Dopfer et al.,2004). The relationships and interactions
between different levels of aggregation have recently been empirically tested. The underlying
4
Jena Economic Research Papers 2016 - 016

Holger Graf, Martin Kalthaus
Friedrich Schiller University Jena
24/06/16
9
Multilayer Networks
Co-authorship at the researcher level
International collaboration between countries
Micro
Collaboration between organizations
Meso
Macro
Figure 1: Multimodal structures
assumption of such analyses is that the network structures at different levels of aggregation in-
fluence each other (Gupta et al.,2007). For example, Guan et al. (2015b) analyze the influence
of countries’ positions in the global innovation network on the performance of actors in city level
networks. In a similar vein, Paruchuri (2010) shows that inventor performance is influenced by
the positions in intra- and interfirm networks.
2.3 Linking national research networks and global embeddedness
In the following, we derive hypotheses regarding the relation between the meso structures and
macro embeddedness. Research networks on the national level can be thought of represent-
ing countries’ research systems where different types of actors, such as universities, research
institutes, companies, or governmental agencies interact in different ways. Collaboration on
this level is determined by incentives, norms, or specific cultures towards collaboration which
might differ between research fields and/or technologies but also between countries (Lundvall,
1992;Malerba,2002;Wuchty et al.,2007). Despite the benefits of collaborating with partners
that speak the same language or are proximate with respect to geographical or institutional
dimensions (Boschma,2005), the reasons for collaboration with national or international part-
ners should be the same with a focus on the cognitive dimension. Therefore, if a country is
characterized by a high level of collaboration on the national level, we expect the likelihood to
cooperate on the international level to be higher as well.
Hypothesis 1 Countries that are characterized by high collaboration intensity within the na-
tional research network, also collaborate more with international partners than countries with a
low national collaboration intensity.
Countries might rely on few strong actors (national champions) to follow a mission oriented
national strategy to advance research in a specific field (Ergas,1987). If countries have such
strong leaders, it is often the strategic goal to advance knowledge mainly within the country
with a reluctance to share knowledge internationally. Furthermore, Owen-Smith et al. (2002)
5
Jena Economic Research Papers 2016 - 016

argue, that the decentralized organization of public research in the U.S. was relevant for their
central position within the international life sciences knowledge network. Therefore, we expect
centralized countries to be less open to international collaboration and less embedded in the
international research network.
Hypothesis 2 Countries with highly centralized national research networks are less embedded
within the global knowledge network than countries with decentralized, diffusion oriented national
networks.
We also expect that functioning national research systems are characterized by internal as
well as external openness due to a general, learned capability of collaboration and networking
(Bathelt et al.,2004;Graf,2011). Here, we assume that the functionality of a system in terms
of knowledge exchange and learning is better the larger the share of actors who are connected
to each other.
Hypothesis 3 Countries with national research networks characterized by high connectivity are
better embedded within the global knowledge network than countries with fragmented national
networks.
2.4 Policy influence on international embeddedness
PV is considered an environmentally friendly technology which generates electricity without
emitting CO2or other harmful substances. However, it was only until recently that PV became
cost competitive with conventional electricity generating technologies. Therefore, governments
intervene to increase the efficiency of the technology, to decrease production costs, and to foster
R&D in PV. In general, there are several approaches to support research activity and tech-
nological development which can be broadly categorized as demand pull or technology push
policies (Mowery and Rosenberg,1979;Jaffe et al.,2002;Groba and Breitschopf,2013;Rogge
and Reichardt,2016). There is a growing theoretical and empirical literature in innovation and
environmental economics which tries to understand how these policy interventions affect innova-
tive output, especially in environmentally friendly technologies (see Jaffe et al.,2002;Kemp and
Pontoglio,2011;Groba and Breitschopf,2013, for reviews). In the case of scientific research and
collaboration, evaluations of such interventions are scarce and focuses on direct funding only.1
In the following, we derive hypotheses regarding the influence of different policies towards re-
newable energies and PV in particular on the international embeddedness of countries in the
global research network.
Technology push instruments are motivated by positive externalities or technological spillovers
which lead to underinvestment in R&D. R&D subsidies are a classic example of such policies as
they foster research activities by public and private actors (Arrow,1962). Several studies in the
economics of innovation show that R&D subsidies help to increase inventive activity (Watanabe
et al.,2000;Johnstone et al.,2010;Peters et al.,2012;Wangler,2013) and networking (Cant-
ner et al.,2016) in PV research. Concerning general effects of technology push instruments on
1However, several studies focus on the micro (researcher) or meso (institute) level and find usually a positive
effect of funding on publication output (see Ebadi and Schiffauerova,2013, for a review).
6
Jena Economic Research Papers 2016 - 016

publications, Crespi and Geuna (2008) find that on the macro level expenditures on higher edu-
cation research and development increase research output, while Popp (2016) shows that direct
funding increases research output in energy research, especially in solar energy, but in both cases
with a considerable time lag. Concerning the effect of such policies on collaboration and network
structures, there is only limited evidence for the collaboration intensity at the micro (researcher)
level. Based on survey data, Bozeman and Corley (2004) and Lee and Bozeman (2005) find that
the availability of grants leads to larger researcher teams and more collaboration. In a similar
vein, Ubfal and Maffioli (2011) find that Argentinian researchers who received a grant are better
integrated in the scientific community. Adams et al. (2005) find that federally funded R&D,
increase the number of papers, team size per publication, as well as international cooperation
for US universities.
Hypothesis 4 International embeddedness increases with the amount of funding towards re-
search and development.
Demand pull policies increase demand by creating (niche) markets for new or infant tech-
nologies. Thereby, they attract companies to engage in production and benefit from economies
of scale and learning-by-doing effects. If firms are profitable, they generate internal funds to con-
duct research and inventive activities which also contribute to the advancement of a technology.
Investment subsidies, quota systems, or feed-in-tariffs are typical examples for such policies. In
the case of PV, countries implemented different approaches to support commercialization of PV
which in most cases also increased inventive activity (Johnstone et al.,2010;Peters et al.,2012;
Wangler,2013) and research collaboration (Cantner et al.,2016). Public procurement is another
form of demand pull policy which has shown positive effects on R&D activities (Geroski,1990;
Edler and Georghiou,2007;Guerzoni and Raiteri,2015). In the case of public procurement,
governments create demand for societal needs and acts as a lead user by asking for sophisti-
cated products with clearly defined characteristics. In the case of PV, the government was the
first customer for PV cells to power satellites and space applications (Oliver and Jackson,1999;
Petroni et al.,2010;West,2014), which can be considered public procurement. Since PV cells
for aerospace needed to be as efficient as possible, research was conducted to fulfill advanced
requirements and provide efficiency improvements until today.
Hypothesis 5 International embeddedness increases with the amount of effective demand pull
policies.
Besides these targeted instruments, the Kyoto Protocol can also be considered as a policy
instrument which should encourage research and inventive effort in PV. Ratifying the Kyoto
Protocol shows some commitment towards emission reduction and, especially for the Annex B
countries, it has binding targets (UNFCC,1997). Since one way to achieve these targets is PV,
countries might increase their research effort after ratifying the Protocol. Some studies show
indeed that the ratification of the Kyoto Protocol fosters inventive activity for PV (Johnstone
et al.,2010) and renewable energies in general (Nesta et al.,2014). Furthermore, the Kyoto
Protocol contains instruments which foster international collaboration and knowledge transfer
(Dechezleprˆetre et al.,2008). These instruments, namely the clean development mechanism
7
Jena Economic Research Papers 2016 - 016

and joint implementation, increase international collaboration and form networks of knowledge
transfer by itself (Kang and Park,2013) which can lead to scientific collaboration between
countries as well.
Hypothesis 6 International embeddedness is larger for countries after ratifying the Kyoto Pro-
tocol.
3 Scientific collaboration networks
3.1 Data: photovoltaic publications
Publications are frequently used to measure output and collaboration at early stages of the
research and innovation process. Several recent bibliometric studies have focused on renewable
energies (Suominen,2014;Guan et al.,2015a;Poirier et al.,2015;Popp,2016) and PV (Dong
et al.,2012;Huang et al.,2013;Du et al.,2014;Stek and van Geenhuizen,2015). We collect
data on photovoltaic publications from Thomson Reuters Web of Science Core Collection2.
The sample consists in total of 106,836 publications from 1946–2015 by 146 countries covering
various scientific fields. Figure 2a depicts the number of publications over time. An exponential
growth in the number of scientific publications which indicates the increased pervasiveness of
PV research during the last decades is evident.
In the following analysis, we restrict the sample to the years from 1980 until 2015 since there
are only few publications before 1980. Furthermore, policy makers started to put more emphasis
on PV research as a response to the oil crisis in the 1970s and research took off globally. In the
sample from 1980 to 2015 105,809 publications are included. We use information on affiliations
as provided by Web of Science to assign papers to organizations and countries. Most publications
are from China, the USA, and Japan (see Table 1) but also European countries are among the
top publishing countries3.
Concerning international collaboration, i.e. publications of co-authors with affiliations lo-
cated in different countries, there are on average 1.26 different countries involved in each pub-
lication. European countries, especially the United Kingdom, France, and Spain are frequently
involved in international collaboration whereas Asian countries, especially Taiwan and China
are less involved internationally. Concerning the development over time, depicted in Figure 2b,
there is a steep increase around 1996, which is most likely related to our original data source.
The information on author affiliations in the Web of Science is more reliable from 1996 onwards.
Keeping this potential problem in mind but in line with Adams et al. (2005), we observe an
increasing trend in international collaboration with some notable differences between countries.
Asian countries, especially Taiwan and China, do not collaborate extensively internationally
and stay roughly at the same level. European countries frequently engage in international col-
laborations and increase their international activity over time. This increase for the European
2The query is photovoltai* or solar cell* in the topic and title section on August 22nd 2016. Only articles,
proceedings papers, reviews or book chapters are considered.
3We do not calculate publication shares in case of international collaborations so the total number of publication
per country does not match the total number of publications. Furthermore, we do not control for the quality of
publications since our focus is on collaboration patterns and restricting the sample to some top journals would
not represent the whole collaboration network. We also do not limit the scope of papers to specific research fields,
since technological and social progress are interlinked.
8
Jena Economic Research Papers 2016 - 016

Table 1: Number of publications and international collaboration by country from 1980 until 2015
Country Publications Share International collaboration
per publication
China 21,380 16.7% 1.266
USA 18,790 14.6% 1.451
Japan 9,196 7.2% 1.329
South Korea 8,985 7.0% 1.319
Germany 8,648 6.7% 1.662
India 5,728 4.5% 1.344
Taiwan 4,787 3.7% 1.214
United Kingdom 4,688 3.7% 1.837
France 3,851 3.0% 1.828
Spain 3,447 2.7% 1.739
Rest of World 38,843 30.3%
Total 128,343 100,0% 1.256
countries could be related to the common labor market and the EU-Framework Programmes,
which require pan-European collaboration.
0 5000 10000 15000
Year
Publications per year
1946 1951 1956 1961 1966 1971 1976 1981 1986 1991 1996 2001 2006 2011
(a) Annual PV publications.
1.0 1.2 1.4 1.6 1.8 2.0
Year
Average Country Collaboration
1980 1983 1986 1989 1992 1995 1998 2001 2004 2007 2010 2013
China
USA
Japan
Germany
South Korea
India
Taiwan
United Kingdom
France
Spain
Average
(b) Average country-collaboration per pub.
Figure 2: Overview of global PV publications.
3.2 International research network
As pointed out above, we are interested in the structure and dynamics of scientific collaboration
between countries. We employ methods of social network analysis (see Wassermann and Faust,
1994) to elaborate on the countries’ collaboration pattern and embeddedness in the interna-
tional research network. To analyze the networks over time, we use three-year moving windows.
Thereby we account for persistence and decay of collaboration, since the date of publication
is just a point in time, while the actual collaboration existed before and maybe persisted after
the publication (Fleming et al.,2007;Schilling and Phelps,2007). We reconstruct undirected
international research networks using publications from 1980 until 2015, i.e. the first network
covers the period 1980 to 1982 and the last network covers 2013 to 2015 leading to 34 overlap-
ping observation periods. Three of these reconstructed international networks are illustrated in
Figure 3.
9
Jena Economic Research Papers 2016 - 016

We calculate several indicators to describe the development of the international collaboration
network over time (see Figure 4). The number of nodes (i.e. countries), which indicates the size of
the network, increases steadily (see Figure 4a). The mean degree measures the average number
of connections of a node, i.e. the number of distinct co-authoring countries. Here, we see a
steady increase, indicating that on average countries become increasingly embedded within the
global network. The declining number of components also shows that the countries are getting
increasingly interconnected and hardly any country performs research without international
collaboration by the end of our observation period. This can also be seen in the share of
isolates, countries which are not connected to another country, which diminishes drastically (see
Figure 4b).
Concerning the importance of different countries in the network, we use the concept of net-
work centralization. These measures are less concerned with the overall connectedness but rather
with the specific structure of relations and relative positions of nodes. We use two centralization
measures to account for the concentration of linkages on few nodes (degree centralization) and
the dependence on nodes that connect many other nodes (betweenness centralization) proposed
by Freeman (1979). Both measures are equal to 1 in a star network, in which all nodes are
connected to one central node but not among each other, and take a value of 0 for networks
without prominent positions, such as a ring or a complete graph. In Figure 4b, we present
degree and betweenness centralization for the network. The degree centralization increases con-
stantly over time, indicating that there are some countries that are way more interconnected
than the average. The development of betweenness centralization shows that the concentration
of knowledge flows increases during the early periods but diminishes throughout the last periods.
Additionally, transitivity indicates the likelihood that adjacent nodes of a node are connected.
For the global network, we see that except for the early phase transitivity increases constantly.
Apparently, countries increasingly form densely connected clusters. Network density, which is
the share of all present connections in all possible connections, increases despite network growth,
indicating an over-proportional increase in linkage formation.
Regarding countries’ positions within the global network, we focus on four measures of
research performance and international embeddedness: Number of publications, degree, flow
betweennes, and k-core per country. The number of publications per country indicates the
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
(a) Period 1980–1982.
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●●●
●
●
●
●●
●
●
●●
●
●
●
●
●●
●●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●●●●
(b) Period 2003–2005.
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●●
●
●
●
●
●
●
●
●●
●
●
●
●
(c) Period 2013–2015.
(Colored nodes refer to the countries presented in Figure 6)
Figure 3: International research network for three periods.
10
Jena Economic Research Papers 2016 - 016

0 20 40 60 80 100 120
Period
Network Size
1980−1982 1985−1987 1990−1992 1995−1997 2000−2002 2005−2007 2010−2012
Mean degree
Number of nodes
Number of components
(a) Network size.
0.0 0.2 0.4 0.6 0.8 1.0
Period
Network Characteristics
1980−1982 1985−1987 1990−1992 1995−1997 2000−2002 2005−2007 2010−2012
Degree Centralization
Betweenness Centralization
Density
Transitivity
Share of Isolates
(b) Network Characteristics.
Figure 4: Evolution of the international research network.
3
1
1
1
2
4
2
1
1
A
B
C
D
E
F
Figure 5: Example network.
Table 2: Example data.
Node Degree Flow Betweenness K-Core
A 4 18 3
B 4 36 3
C 4 12 3
D 2 4 2
E 3 10 3
F 1 0 1
research output of a country. Degree, flow betweenness, and k-core are different concepts of
centrality, all related to the number of connections. Degree is a simple count of the number
of connections irrespective of their intensity, while flow betweenness considers the intensity but
also the relative position within the whole network (Freeman et al.,1991). The k-core of a graph
is the maximal subgraph in which every node has at least degree k (Seidman,1983). Figure 5
and Table 2show a simple example to point out the differences between the three concepts.
Nodes A and B in the example have the same degree, both are connected to four other nodes.
But if we consider flow betweenness, we see that node B is much more central than A. B is
better connected to its neighboring nodes than A which puts B a better position in the network
to access external knowledge. However, it has to be noted that degree is limited by the number
of nodes in the network, while flow betweenness is more or less unrestricted. This measure not
only accounts for the number of collaboration partners (A still has more access to knowledge
than the other nodes) but also for the quality of cooperation partners. The k-core tells us if a
node is member of the network core or rather of its periphery. Here, we see that nodes A, B, C,
and E form the core in which every node has a degree of at least three, while D and F are in a
more peripheral position.
Figure 6depicts the development of these four measures for the top ten countries over time.
The number of publications was highest in the USA until the last five periods, when China took
over the lead. There is a strong increase in the number of publications from Asian countries.
Besides China, also South Korea, India, and Taiwan are catching up. Japan was among the most
publishing countries from early on, but is eventually outmatched by South Korea and Germany.
The degree shows an interesting development over time (the maximum for degree is limited
by the size of the network, see Figure 4a). Surprisingly, Spain has the highest degree in some of
the early periods but was again overtaken by the USA, which together with Germany has most
11
Jena Economic Research Papers 2016 - 016

0 500 1500 2500 3500
Period
Publications per period
1980−1982 1987−1989 1994−1996 2001−2003 2008−2010
China
USA
Japan
South Korea
Germany
India
Taiwan
United Kingdom
France
Spain
0 20 40 60 80
Period
Degree per period
1980−1982 1987−1989 1994−1996 2001−2003 2008−2010
China
USA
Japan
South Korea
Germany
India
Taiwan
United Kingdom
France
Spain
0 10000 30000 50000
Period
Flow Betweenness per period
1980−1982 1987−1989 1994−1996 2001−2003 2008−2010
China
USA
Japan
South Korea
Germany
India
Taiwan
United Kingdom
France
Spain
0 5 10 15 20 25 30
Period
K−Core per period
1980−1982 1987−1989 1994−1996 2001−2003 2008−2010
China
USA
Japan
South Korea
Germany
India
Taiwan
United Kingdom
France
Spain
Figure 6: Network measures for top ten publishing countries.
connections over time. Both are connected to about 70% and 60% respectively of all countries in
the last period. Furthermore, the USA and European countries have a higher degree than Asian
countries for most of the time, and especially Taiwan is lagging behind. A similar pattern can
be observed for flow betweenness, where the USA and Germany are on top. However, in the last
periods China caught up and rages among the top three countries. This indicates that China,
even though it has a lower degree than European countries, is connected to well embedded actors
and has better access to knowledge. However, again, Taiwan is least embedded among the top
ten countries, surpassed by India and Japan. The k-core shows no surprising development. Over
time all countries join the core group within the network. There is very low variation over time
and besides Taiwan, all countries quickly connect to the central core.
So far, we exemplified general trends of network development by looking at the top ten
publishing countries. To analyze the underlying dynamics for all countries, we compare their
relative position in the network over time. We rank all countries according to their degree in
period 2003–2005 and compare this ranking with the periods 2008–2010 and 2013–2015. This
gives us a Salter-Curve like representation of the dynamics in the network (see Figure 7). We
see that at the top of the ranking the changes are marginal, while there is quite some turbulence
in the middle. Among the top actors, especially Mexico is loosing its position, while most of
the other countries stay rather stable. Qatar, the United Arab Emirates, Serbia and Malaysia
are the countries which improve the most. Some other Arab countries improve their position
as well. The top 15 as well as the 15 countries with the highest movement in the ranking are
shown in Appendix A.1.
12
Jena Economic Research Papers 2016 - 016

Rank of countries' degree
Germany
France
USA
United Kingdom
Italy
Japan
Netherlands
Spain
Sweden
Switzerland
Russia
Belgium
Australia
China
Austria
Mexico
India
Israel
Poland
Finland
Canada
Greece
Ukraine
Czech Republic
South Korea
Brazil
Cyprus
Egypt
Romania
Belarus
Argentina
Lithuania
Hungary
Bulgaria
Slovakia
Cuba
Algeria
Denmark
Ireland
Turkey
New Zealand
Morocco
Taiwan
Portugal
Saudi Arabia
Slovenia
Estonia
Thailand
Vietnam
Chile
Iran
Jordan
Tunisia
Singapore
South Africa
Sri Lanka
Bangladesh
Colombia
Lebanon
Latvia
Pakistan
Moldova
Ethiopia
Armenia
Indonesia
Uruguay
Uzbekistan
Tajikistan
Peru
Oman
Kuwait
Togo
Venezuela
Liechtenstein
Azerbaijan
Kenya
Libya
Zambia
Tanzania
Serbia Monteneg
Malaysia
Norway
Croatia
Nigeria
Senegal
Iraq
Bahrain
Syria
Iceland
Brunei
Yugoslavia
Serbia
Luxembourg
Cameroon
Ivory Coast
United Arab Emirates
Turkmenistan
Qatar
Kazakhstan
Philippines
Costa Rica
Ghana
Burkina Faso
Nepal
Niger
Sudan
Uganda
Malawi
Reunion
Jamaica
Zimbabwe
Yemen
Ecuador
Botswana
Fiji
Barbados
Bosnia & Hercegovina
Bolivia
Lesotho
Malta
Mauritania
Mauritius
Mongolia
Palestine
Benin
Cape Verde
Macedonia
Myanmar
Nicaragua
Georgia
Albania
Montenegro
Paraguay
130
120
110
100
90
80
70
60
50
40
30
20
10
1
Rank in Period 2003−2005
Rank in Period 2008−2010
Rank in Period 2013−2015
Figure 7: Rank of the degree of countries.
3.3 National research networks
In the following, the structure of interaction within each country is analyzed. Information on
author affiliations allows us to reconstruct national research networks. Here, nodes represent
different organizations, such as universities, research institutes, or companies and edges represent
joint publications of researchers with different affiliations4. We reconstruct national research
networks for all countries in our sample5. Again, we present network measures for the top
ten publishing countries in Figure 8to illustrate the general patterns of research activity and
network development.
We observe an exponential increase in network size, indicating that more organizations
emerge and engage in PV research. But there are notable differences between countries. While
China and India experienced vast growth especially in the last periods, other countries, most
notably the United Kingdom, show hardly any increase in the number of actors. Concerning
the connections among these actors in the research system, mean strength (degree, weighted
by the intensity of the connection) is increasing in all countries. Especially Taiwan and South
Korea are very well connected at the national level. This is remarkable, since they are not that
well connected internationally, as shown above (Table 1and Figure 6). Another interesting case
is India, which shows a very large increase in the number of nodes, but not with respect to
mean strength, which indicates that there might be some deficits in domestic collaboration. In
general, Asian countries seem to have a higher internal connectivity than European countries in
the last periods.
Further indicators add to our understanding of the development of structural differences
between national research networks. The share of actors in the main component takes the size
of the largest component over the size of the network.6This measure increases in all countries
4Since we are interested in the structure of national research systems (and use its structural properties to
explain global network positions, i.e. international collaboration in Section 4), we exclude cooperation partners
in foreign countries
5Since the affiliation data is quite noisy, we consider only the organization name and neglect information about
departments or other subsidiary information.
6The share of actors in the main component is sensitive for small networks and can lead to extreme values as
13
Jena Economic Research Papers 2016 - 016

0 500 1000 1500
Period
Number of nodes in each country
1980−1982 1987−1989 1994−1996 2001−2003 2008−2010
China
USA
Japan
South Korea
Germany
India
Taiwan
United Kingdom
France
Spain
0 5 10 15 20
Period
Mean Strength in each country
1980−1982 1987−1989 1994−1996 2001−2003 2008−2010
China
USA
Japan
South Korea
Germany
India
Taiwan
United Kingdom
France
Spain
0.0 0.1 0.2 0.3 0.4 0.5 0.6
Period
Degree Centralization in each country
1980−1982 1987−1989 1994−1996 2001−2003 2008−2010
China
USA
Japan
South Korea
Germany
India
Taiwan
United Kingdom
France
Spain
0.0 0.2 0.4 0.6 0.8 1.0
Period
Share in main component in each country
1980−1982 1987−1989 1994−1996 2001−2003 2008−2010
China
USA
Japan
South Korea
Germany
India
Taiwan
United Kingdom
France
Spain
Figure 8: National network properties of for top ten publishing countries.
from the mid 1990s onwards indicating that the networks become less fragmented over time with
the potential for knowledge flows between an increasing number of actors. Degree centralization
accounts for the concentration of links in the network. It does not show a clear trend as the other
structural measures and there is quite some variation between countries. Especially Taiwan,
China, and South Korea appear to have a more centralized research systems in PV than e.g.
Germany, India, the USA, or France.
4 Explaining embeddedness in the international research net-
work
The embeddedness of a country in the international research network might be influenced by
different objectives, economic circumstances, strategic decisions, or geographic location. We
are interested in the effect of two particular influencing factors; i) the structural properties
of the national research network and ii) national policies towards PV, e.g. by introducing and
supporting research activities. In this section, we test the hypotheses derived in Sections 2.3
and 2.4. The variables are defined in Section 4.1, followed by the description of our estimation
strategy in Section 4.2. The results are presented in Section 4.3 and their robustness is checked
in Section 4.4.
seen in the first periods.
14
Jena Economic Research Papers 2016 - 016

4.1 Variables
For the econometric analysis we use four sets of variables: dependent variables to describe
international embeddedness of countries in the global PV research network and independent
variables characterizing the national networks, national policies related to PV and renewable
energies as well as controls. We conduct the analysis for the period 1980–2015, a robustness
check for the sub-period 1997–2015 is discussed in Section 4.4. Since we use three-year moving
windows for international and national network measures, a period serves as an observation and
the starting year of the period refers to the year of observation. So the first period 1980–1982 is
the observation for 1980 and the second period, 1981–1983 is the observation for 1981. Summary
statistics of the variables are presented in Table 3. The correlations between the variables are
documented in the Appendix A.3.
Dependent Variables – International embeddedness: The four dependent variables pub-
lications,degree,flow-betweenness, and k-core (as discussed in Section 3.2) measure countries’
performance and international embeddedness. Publications accounts for the research output
and can be seen as a benchmark how the national research system and policy influence perfor-
mance. The three network variables emphasize different aspects of international embeddedness,
i.e. how well a country is connected to other countries and how important a country is in terms
of knowledge transfer between other countries.
National network variables: We use three properties of the national research networks as
explanatory variables to account for the characteristics of the respective innovation systems (see
Section 3.3). Mean strength measures the intensity of interaction, degree centralization indicates
the concentration of linkages, i.e. the importance of ’national champions’, and the share in main
component to account for the overall potential of knowledge flows inside the country if the
national network is well connected.
Policy variables: Several variables are used to operationalize national policies towards PV
and renewable energies in particular or climate change in general. To account for technology
push policies towards PV research, we use PV R&D expenditures by the government for PV in
Mio US$ (IEA,2016). However, this information is only available for some countries and not
for all years. Whenever only a few years of observation for a country are missing, we interpolate
R&D data and add a dummy to control for a possible effect of interpolation. Furthermore, we
use the logarithm of annually installed PV capacity in MW (IEA,2016), as a proxy for demand
pull. Since PV is only recently price competitive, any installation must be somehow subsidized
by the government. This measure is frequently used in the literature because it accounts for
the effectiveness of a variety of policy instruments in inducing demand (Johnstone et al.,2010;
Peters et al.,2012;Wangler,2013;Cantner et al.,2016). Additionally, we use data on satellites
to proxy public procurement in PV, since satellites were the first major application of PV and
require until today the highest efficiency which is achieved by constant research activity (Oliver
and Jackson,1999;Petroni et al.,2010;West,2014). We use the cumulated number of satellites
15
Jena Economic Research Papers 2016 - 016

Table 3: Variable descriptive statistics 1980–1982 until 2013–2015.
Min. Median Mean Max. SD Obs.
Dependent variables
Publicationst0.000 0.000 17.057 3371.000 111.068 4964
Degreet0.000 3.000 9.203 87.000 13.724 2488
Flow Betweennesst0.000 52.000 760.922 45521.000 2663.286 2488
K-Coret0.000 3.000 5.599 27.000 6.587 2488
National network variables
Mean Strengtht0.000 0.500 1.171 19.589 1.940 2488
Degree Centralizationt0.000 0.119 0.123 0.667 0.115 1937
Share in Main Componentt0.033 0.500 0.585 1.000 0.300 2488
National policy variables
Kyoto Ratificationt−10.000 0.000 0.276 1.000 0.447 4964
Cum. Number of Satellitest−10.000 0.000 31.311 3412.000 263.446 4964
Installed PV Capacityt−10.000 0.000 1.097 9.138 1.979 748
PV R&D Exp.t−10.000 2.604 17.135 395.660 38.755 731
PV R&D Exp. interp.
Dummyt−1
0.000 0.000 0.079 1.000 0.270 986
Controls
GDP per Capitat−1428.150 7392.135 13481.597 249579.559 17014.726 4440
EU Membershipt0.000 0.000 0.114 1.000 0.318 4964
Conventional R&D Exp.t−10.000 51.908 387.259 6110.350 818.237 695
deployed over time7to proxy the effort and commitment of a country towards the aerospace
sector.8Kyoto Ratification is a dummy variable which takes a value of 1 in each year in which a
country has ratified the Kyoto Protocol and 0 otherwise. It serves as an indicator for countries’
commitment towards emission reduction.
Control Variables: We use the GDP per Capita provided by the Penn World Table (Feenstra
et al.,2015) to account for countries’ general state of development. Furthermore, national
conventional R&D expenditures for fossil and nuclear energy research are employed to account
for research activity in the energy sector (IEA,2016). Since we expect that the common EU
research area fosters collaboration between European research partners (Defazio et al.,2009),
we control for EU Membership in all models, except for the short period full models in table A.4.
4.2 Estimation strategy
We conduct our analysis using unbalanced OLS-panel regression controlling for country and time
fixed effects to account for the differences between countries but also for time effects such as
general economic circumstances. Since we are interested in the causal effect of the policies, we
lag the policy variables by one year. This allows to estimate the effect of these variables on the
position within the network of the following three years9. We do not lag the national network
7The data was collected from http://satellitedebris.net/Database/LaunchHistoryView.php on May 2nd
2015.
8We considered different ways to operationalize public procurement. Using the deployed satellites per year
is another possibility but also using a dummy from the year onwards a country deployed its first satellite or
cumulating the years from the first deployed satellite onwards as a knowledge stock. The results are available on
request.
9As explained in Section 3.2, networks are reconstructed for overlapping 3-year periods.
16
Jena Economic Research Papers 2016 - 016

variables, since we assume that actors decide in the same period about all of their cooperation
partners. To account for heteroscedasticity, we report robust standard errors. Indexing countries
by iand time by t, the generic regression model is the following:
Embeddednessit =β1Network Structureit +β2Policyit−1+β3Controlsit−1+ FEi+ FEt+ε
We use two model specifications due to the lack of policy variables for many countries. In
the restricted models, we omit installed PV capacity,PV R&D expenditures, and conventional
R&D expenditures which we only have for a few, developed countries and allows us to analyze
the factors responsible for international embeddedness of 114 countries. The full models include
all variables but the available data covers only 18 countries10 .
4.3 Results
With four measures for performance and international embeddedness and two specifications, we
end up with eight regression models to analyze the effects of national network structure and
policy intervention (Table 4).11
Publications: In the first two models, we estimate the effect on the number of publications.
The restricted model 1 reveals that mean strength is a strong predictor of the number of publi-
cations while degree centralization has the expected negative effect on publications. The share
in main component has no significant effect. Concerning the policy variables, Kyoto Ratification
has no effect while public procurement proxied by the cumulated number of satellites shows a
positive effect. In the full model 2, the inclusion of additional variables does not change these
results. Fostering PV by means of demand pull (installed PV capacity) as well as technology
push policies (PV R&D expenditures ) has positive effects on the number of publications. The
control variables are insignificant, except for the EU dummy which is negative in model 1.
Degree: The factors influencing international embeddedness in terms of degree are presented
in models 3 and 4. In the restricted model 3, all variables show an effect in the expected direction.
In the full model 4, however, a slightly different picture emerges. Here, degree centralization and
share in main component as well as the Kyoto Ratification and installed PV capacity remain
significant. Since mean strength is highly correlated with installed PV capacity (see Table 8),
it might well be that the effect of the former is at least partly caught by the latter. PV R&D
expenditures show no effect, whereas conventional R&D expenditures seem to negatively affect
degree.
Flow Betweenness: Flow betweenness is analyzed in models 5 and 6. In the first model, all
three structural properties of the national research networks are significant and supportive of
10These countries are: Australia, Austria, Belgium, Canada, Denmark, France, Germany, Italy, Japan, the
Netherlands, Norway, Portugal, South Korea, Spain, Sweden, Switzerland, Turkey, and the USA.
11There is a rather high correlation between mean strength and instal led PV capacity as well as between
R&D expenditures and conventional R&D expenditures. We tested these variables separately to see if there are
changes in the respective coefficients and standard errors. The results reveal hardly any changes, indicating that
multicollinearity is not an issue. The results are available on request.
17
Jena Economic Research Papers 2016 - 016

our hypotheses. Regarding the policy measures, Kyoto Ratification has no significant effects on
embeddedness, while procurement in terms of cumulated number of satellites shows a strong
positive effect. Again, and contrary to our expectation, the coefficient for EU membership is
negative. In the second model, none of the network properties are significant and only the cumu-
lated number of satellites and installed PV capacity exert a positive influence on international
embeddedness and foster access to global knowledge flows.
K-Core: In the case of k-core, model 7 reveals that national collaboration in terms of mean
strength and share in main component are positive predictors of membership in a higher level core
of the global knowledge network. Furthermore, the Kyoto Ratification has a weakly significant
effect. Both control variables have a positive significant effect. It is noteworthy that this is the
only specification where EU Membership shows a positive influence. The full model 8 shows
divergent and almost no significant results. The reason lies in the properties of this measure of
embeddedness. Since the central core of the network is composed of many, highly interrelated
countries (35 countries by the end of our observation period), nearly all 18 countries included in
this model enter the core at some point, so that there is very little variation in the dependent
variable (see Figure 6). As such, this variable does not discriminate between the most central
countries as much as degree and flow betweenness. This is also indicated by the small adj. R2,
which is about an order of magnitude smaller than in most of the other regressions.
Summary: Overall, research performance and international embeddedness in the global re-
search network are strongly influenced by the structure of the national research network as well
as by national policies. As hypothesized for mean strength in H 1intense collaboration within the
national research network increases international embeddedness. This holds true for all models
that include a large set of countries, regardless how embeddedness is measured. However, for the
models which cover only 18 developed countries but include additional explanatory variables,
this relationship does not hold for embeddedness, but for publications. Centralization of the
national research system is detrimental for research output and H 2gains support in the degree
models and the flow betweenness model with the large sample. This indicates that countries
which centralize their PV research activity and focus on ’national champions’ are less embedded
in the international network. Concerning the functioning of the national research system, H 3
assumes that connectedness as measured by share in main component has a positive effect on
embeddedness. This argument finds support in the degree models as well as in the flow be-
tweenness and k-core models for the large sample of countries. In general, the national network
structure seems to be a good predictor of international embeddedness and research performance,
but these results are sensitive to sample size and the concepts used for its measurement.
With respect to the influence of governmental intervention, H 4assumes that direct subsidies
for PV R&D increase embeddedness. However, this is not the case and we can only observe a
positive effect on the number of publications. Apparently, research funds are a valuable input
for research activities without direct effects on international collaboration. In general, demand
side policies have a positive effect as proposed in H 5. If demand side policies are proxied by
installed PV capacity, this holds true except for the k-core. If demand is induced by governments
themselves in the form of public procurement, proxied by the cumulated number of satellites, this
18
Jena Economic Research Papers 2016 - 016

also holds for all cases except for k-core and degree in the small sample. Hypothesis 6assumes
that the Kyoto Ratification induces activities to foster renewable energies which might show in
an increased embeddedness in the global PV research network. However, this is only the case for
the degree models and for k-core in the large sample while there is no influence on research output
and knowledge access in terms of flow betweenness. Overall, governmental interventions have an
effect on international embeddedness, however, the instruments differ in their effect. Creating a
market by means of demand side policies seems more effective for international embeddedness
than the provision of research funds or a general commitment to mitigate climate change.
4.4 Robustness tests
We conduct a robustness test for the econometric analysis to account for possible inconsistencies
in the original data. As mentioned in section 3.1, the way Web of Science stores affiliation data
changed around 1996. Furthermore, with the disbandment of the Soviet Union, several countries
left the sample and new ones emerged. To account for such effects beyond the already present
time fixed effects, we perform regressions with a subsample of the data covering the periods
1997–1999 to 2013–2015. The results as well as the correlations and descriptive statistics are
presented in Tables 7,9, and 10 in the Appendix.
The regression results for this shorter but more reliable period are quite stable and there
are only marginal differences to the results presented above. The significance of mean strength
and degree centralization does not really change, however, the share in main component is no
longer significant for flow betweenness and k-core while the negative effect on publications turns
significant. Concerning the policy variables, Kyoto Ratification is no longer significant in the
regressions with the large sample for degree and k-core. The cumulated number of satellites
is now significant for degree as well. The installed PV capacity loses its significance in the
regression for the publication. Interesting results emerge for the PV R&D expenditures which
have a weak negative effect on degree. Furthermore, conventional R&D expenditures show a
significant negative effect on the number of PV publications.
5 Discussion
The present study analyzed the global research network in photovoltaics based on an original
dataset of scientific publications in the field of PV. We asked two broader research questions:
first, how did the global research network, in which countries are connected via international
co-authorship, evolve over time? And second, what are the country level determinants of in-
ternational embeddedness? To answer the second question, we focused on two types of country
characteristics that seem influential. The first set of factors is comprised of national policies
towards renewable energies and climate change in general and towards PV in particular. Our
results add to the broad literature that analyzes effects of policy on environmentally friendly
innovation (e.g. Popp,2002;Newell,2010;Kemp and Pontoglio,2011;Acemoglu et al.,2012)
and the more recently upcoming literature on the policy mix for innovation (Flanagan et al.,
2011;Rogge and Reichardt,2016;Cantner et al.,2016). With the second set of factors we enter
an emerging research field by relating country level network characteristics – the meso level –
19
Jena Economic Research Papers 2016 - 016

Table 4: OLS Panel regression results for country embeddedness periods 1980–1982 until 2013–2015.
Publications Degree Flow Betwenness K-Core
Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 Model 7 Model 8
National network variables
Mean Strengtht60.791*** 30.512** 1.903*** 0.189 571.828*** 232.811 0.687*** 0.040
(17.171) (12.120) (0.409) (0.200) (202.855) (285.449) (0.147) (0.104)
Degree Centralizationt-334.948*** -280.621*** -16.865*** -15.026*** -5693.235*** -3693.407 -0.986 -1.346
(72.106) (95.600) (3.703) (4.842) (1521.648) (3248.624) (1.242) (1.553)
Share in Main Componentt-36.932 -98.164 11.070*** 10.409*** 1522.974* -1546.532 1.562** 0.490
(41.722) (82.777) (2.373) (2.896) (826.979) (1658.821) (0.677) (0.780)
Policy variables
Kyoto Ratificationt−1-7.117 -37.674 2.479** 3.479* -22.110 80.441 0.665* -0.723*
(31.033) (48.464) (1.183) (1.921) (528.756) (716.144) (0.396) (0.427)
Cum. Number of Satellitest−11.008*** 1.201*** 0.026*** 0.008 15.270*** 16.597*** 0.002 -0.003***
(0.206) (0.121) (0.004) (0.007) (2.629) (3.488) (0.002) (0.001)
Installed PV Capacityt−136.823*** 1.414*** 1049.462*** -0.144
(14.064) (0.518) (307.835) (0.094)
PV R&D Exp.t−10.588* 0.012 11.485 0.003
(0.320) (0.013) (7.112) (0.003)
PV R&D Exp. interp. Dummyt−1-191.626 2.459 -1998.050 -0.101
(143.296) (2.415) (3397.267) (0.252)
Control variables
GDP per Capitat−1-0.002 -0.003 0.000*** 0.000 0.019 -0.061 0.000** 0.000
(0.002) (0.003) (0.000) (0.000) (0.028) (0.059) (0.000) (0.000)
EU Membershipt-30.716** -4.351 0.278 2.103 -585.492* 665.253 1.175** 0.276
(14.584) (21.422) (1.317) (1.406) (306.476) (434.231) (0.545) (0.623)
Conventional R&D Exp.t−10.060 -0.003* 0.312 -0.001
(0.037) (0.002) (0.794) (0.000)
Country fixed effects yes yes yes yes yes yes yes yes
Time fixed effects yes yes yes yes yes yes yes yes
Adj. R2 0.395 0.548 0.396 0.412 0.307 0.403 0.199 0.050
n 114 18 114 18 114 18 114 18
T 34 34 34 34 34 34 34 34
N 1876 447 1876 447 1876 447 1876 447
df 1722 385 1722 385 1722 385 1722 385
Robust standard errors in parentheses. Sig. at *** 0.01, ** 0.5, * 0.1 level.
20
Jena Economic Research Papers 2016 - 016

to macro level embeddedness. While there are some studies concerned with the effects on net-
work structure on performance (e.g. Verspagen and Duysters,2004;Uzzi et al.,2007;Fritsch and
Graf,2011), only few studies relate different levels of networks in a research or innovation context
(Paruchuri,2010;Guan et al.,2015b). Our empirical results show that country level network
structures are highly relevant for international embeddedness but also for research performance.
We argue that the structure of national networks should be interpreted as characteristics of the
national research system that are also subject to decisions taken by policy makers.
With respect to the evolution of structural properties of the global PV research
network, we observe in Figures 2and 4that research output and the resulting network of
international research collaboration are constantly growing. This highlights the global awareness
regarding renewable energies and PV in particular as possibilities to mitigate climate change, but
also with respect to existing market opportunity worth exploiting (Oliver and Jackson,1999;
Zheng and Kammen,2014). Especially Asian countries are catching up and in recent years
overtaking European countries in terms of research output but also in actual PV production
(Zheng and Kammen,2014). We also observe an increase in collaboration over time, which
has been found to be a general trend in research and innovation activities (Wuchty et al.,
2007). However, there are some notable differences between countries. While European countries
collaborate quite frequently with international partners, Asian countries conduct most of their
research domestically. This might be related to cultural differences, geographic proximity, or
national strategies. There is not only a vast increase in research output, but also in terms of the
number of actors, which indicates that more and more countries engage in PV research. The
reasons should be found in improved market opportunities or industrial policies (Mazzucato,
2013). Emerging countries (with respect to PV research) are quickly embedded in the global
research network, as documented by the decreasing number of components and share of isolates.
At the same time, mean degree, density as well as transitivity increase, which shows that the
global system functions well and becomes increasingly connected. Nevertheless, there seems to
be a centralization process going on, such that some countries form a highly interconnected core.
Regarding the dynamics within the global network, we see a constant growth process
of the network, however, there are many changes of relative positions of countries. Figure 7
ranks countries according to their degree centrality over time to visualize the dynamics within
the network. While we observe stability among the top five countries, some other top ranked
countries cannot keep up with the pace of the others. Especially Mexico, Russia, and the
Netherlands dropped in the ranking, despite a doubling of their number of connections. Further
interesting changes appear in the middle and lower part of the ranking. Some countries which did
not do research at all in the period 2000–2002 get well connected in the last period. Especially
Malaysia moved among the top countries, which was induced by an overall political commitment
to engage in PV (Muhammad-Sukki et al.,2012). Also countries in the MENA region improved
their position notably due to strategic decisions taken by their governments (Griffiths,2013).
The improvement of some Asian countries, especially China, Taiwan, South Korea, and India,
is rather moderate, given that nowadays they publish most of the research in PV.
We use regression analysis to understand the factors influencing international collaboration
and the embeddedness of countries in the research network. The results are – at least partly
21
Jena Economic Research Papers 2016 - 016

– sensitive to the centrality concept used to measure embeddedness, but as an overall result we
conclude that cohesion and connectedness of the national network positively influence interna-
tional embeddedness. Centralization of the national network, i.e. a focus on ’national champi-
ons,’ seems to be detrimental for performance and embeddedness. This implies that functioning
national research systems in which actors are well connected, diverse, and decentralized are sup-
portive of research output and international embeddedness. However, the establishment of an
institutional systems conducive for such structures is certainly influenced by policy intervention
and strategic decisions of governments.
Policy instruments have a differential effect on international embeddedness. R&D expendi-
tures for PV, which are the most direct way to support research activity, are only beneficial for
publication output, which is in line with previous findings (Adams et al.,2005;Popp,2016).
Contrary to previous findings by Adams et al. (2005), we find no significant influence on in-
ternational embeddedness. Apparently, countries mainly strengthen their internal capabilities
by fostering domestic research activities. However, since such domestic R&D grants have been
found to increase collaboration within the country (Adams et al.,2005;Cantner et al.,2016), we
cannot exclude that there is an indirect effect via the structure of the national research networks.
Demand pull policies are a very robust predictor of research performance and international em-
beddedness across most estimated specifications. Even though they are not necessarily designed
to induce R&D activities and innovation, they apparently provide incentives and create an en-
vironment that strengthens research and international collaboration. In addition to market
oriented demand pull instruments, such as quotas or feed-in-tariffs, we also analyzed the effects
of public procurement. Guerzoni and Raiteri (2015) have shown the relevance of public pro-
curement for innovation. In our case, since we use the cumulative number of satellites to proxy
procurement, this type of policy should be more relevant in the early years of the technology
than during the last decades. However, procurement shows to be a very strong predictor of per-
formance and international embeddedness not only in the long period 1980–2015 but also for the
period 1997–2015. The performance effects hint at long term first-mover advantages and since
spacecraft development is frequently conducted in multinational projects, it might well explain
its effects on international embeddedness (Moloney et al.,2014). Countries that ratified the
Kyoto Protocol have a more diverse set of international research partners even though it seems
irrelevant for publication output. Hence, we do not find that such a general commitment to
reduce emissions triggers research activities as it was found in the case of innovation (Johnstone
et al.,2010;Nesta et al.,2014).
6 Conclusions
We present an attempt to explore the factors that affect the research performance as well as
the embeddedness of countries in the international research network. Overall, we found that
characteristics of national research networks as well as national policies are relevant factors for
the explanation of countries’ research performance and international embeddedness. We could
also show that some policies are more effective in strengthening domestic research output, such
as R&D expenditure, while others, such as a high connectedness of the national research network,
seem more effective in fostering international embeddedness. Even though we did not explicitly
22
Jena Economic Research Papers 2016 - 016

test for its effect, we argue that embeddedness in the global research network is important for
access to global knowledge flows.
As with any research, our study is not without limitations and some of them might affect
the interpretation of our results more than others. Publication data is far from perfect to
measure collaboration: the intensity of collaboration is not accounted for, collaboration might
not be properly reflected in co-authorship, or affiliation information is incomplete (for further
issues with publication data, see Katz and Martin,1997;Laudel,2002). Unfortunately, our
analysis suffers from incomplete data, especially concerning R&D expenditures and demand pull
instruments. These policy indicators are only available for a small – and certainly not random –
subset of countries. Increasing the reliability and scope of the data would increase the reliability
of our results and related studies. Finally, since we focus on a highly specific technology in which
policy plays an important role (Cantner et al.,2016), we expect that especially our estimates
on national policies are sensitive to the technology which limits generalizability.
In future research it would be important to understand how the different policies interact
within the institutional framework to affect the network structures. The linkages between the
national – meso – network and the global – macro – network remain another challenge for future
inquiry. Furthermore, we would like to point out that not a single policy, but the combination of
policies as well as the consistency and stringency of policies and governmental strategy influence
each other and form a policy mix (Cantner et al.,2016;Rogge and Reichardt,2016).
Acknowledgements: This paper was written as part of the research project GRETCHEN
(The impact of the German policy mix on technological and structural change in renewable
power generation technologies, www.project-gretchen.de), which is funded by the German
Ministry of Education and Research (BMBF) within its funding priority ”Economics of Climate
Change” under the funding label Econ-C-026. We gratefully acknowledge this support. We
would like to thank the GRETCHEN team members and especially Karoline Rogge for valuable
discussions. Previous versions of the paper were presented at the Doctoral CGDE-Workshop
2015 in Halle, the 2015 European Meeting on Applied Evolutionary Economics in Maastricht,
the XII. Buchenbach-Workshop 2015 in Buchenbach, the Jahrestagung des Evolutorischen Auss-
chusses des Vereins f¨ur Socialpolitik 2015 in Bremen, the 5th Governance of a Complex World
conference 2016 in Valencia and as a poster at the 16th International Joseph A. Schumpeter
Society Conference 2016 in Montreal. We are grateful for discussions by and with Muhammad
Ali, Uwe Cantner, Robin Cowan, Dirk Fornahl, Johannes Herrmann, Frieder Kropfh¨außer, and
Friedrich Thießen.
23
Jena Economic Research Papers 2016 - 016

A Appendix
A.1 Ranking of countries
Table 5: Rank of the Degree of the top 15 countries.
Rank
2003-05
Degree
2003-05
Degree
2008-10
Degree
2013-15
∆Rank
03-05–
08-10
∆Rank
03-05–
13-15
Rank
2013-15
Germany 1 45 65 76 0 -2 3
France 2 43 54 73 -1 -3 5
USA 3 42 63 87 1 2 1
United Kingdom 4 36 53 78 0 2 2
Italy 5 34 44 68 -1 -1 6
Japan 7 30 42 64 0 -1 8
The Netherlands 7 30 34 54 -7 -10 17
Spain 9 26 47 73 4 4 5
Sweden 9 26 36 50 -1 -11 20
Switzerland 10 24 39 55 1 -4 14
Russia 11 22 25 49 -7 -11 22
Belgium 12 21 31 56 -4 0 12
Australia 13 19 27 54 -4 -4 17
China 15 18 35 65 3 8 7
Austria 15 18 23 54 -7 -2 17
Table 6: Rank of the Degree of the 15 most increasing countries.
Rank
2003-
05
Degree
2003-
05
Degree
2008-
10
Degree
2013-
15
∆Rank
03-05–
08-10
∆Rank
03-05–
13-15
Rank
2013-
15
Qatar 133 na 1 28 28 89 44
United Arab Emirates 133 na 3 27 51 87 46
Serbia 133 na 10 19 85 73 60
Malaysia 91 0 18 50 60 71 20
Kazakhstan 133 na 1 15 28 65 68
Philippines 133 na 1 11 28 58 75
Luxembourg 133 na 8 10 76 56 77
Norway 91 0 15 32 52 51 40
Costa Rica 133 na 1 5 28 45 88
Ghana 133 na 1 5 28 45 88
Croatia 91 0 7 25 31 39 52
Saudi Arabia 49 5 18 61 18 39 10
Iraq 91 0 2 25 2 39 52
Burkina Faso 133 na 1 4 28 38 95
Nepal 133 na 1 4 28 38 95
24
Jena Economic Research Papers 2016 - 016

A.2 Descriptives small dataset
Table 7: Descriptive statistics of the 1997-1999 until 2013-2015 periods.
Min. Median Mean Max. SD Obs.
Dependent variables
Publicationst0.000 1.000 33.101 3371.000 155.350 2482
Degreet0.000 6.000 12.245 87.000 15.148 1760
Flow Betweennesst0.000 140.500 1064.318 45521.000 3115.896 1760
K-Coret0.000 5.000 7.444 27.000 6.993 1760
National network variables
Mean Strengtht0.000 0.953 1.538 19.589 2.180 1760
Degree Centralizationt0.000 0.143 0.144 0.667 0.117 1465
Share in Main Componentt0.073 0.520 0.583 1.000 0.277 1760
National policy variables
Kyoto Ratificationt−10.000 1.000 0.551 1.000 0.497 2482
Cum. Number of Satellitest−10.000 0.000 37.909 3412.000 302.727 2482
Installed PV Capacityt−10.000 1.099 2.127 9.138 2.374 374
PV R&D Exp. t-1 0.000 3.435 17.249 395.660 37.759 408
PV R&D Exp. interp. Dummyt−10.000 0.000 0.114 1.000 0.318 493
Controls
GDP per Capitat−1428.150 8915.322 15706.152 164136.454 17703.133 2346
EU Membershipt0.000 0.000 0.149 1.000 0.356 2482
Conventional R&D Exp.t−10.000 43.189 267.541 4552.800 636.316 393
25
Jena Economic Research Papers 2016 - 016

A.3 Correlation tables
Table 8: Correlation table for the periods 1980-1982 until 2013-2015.
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
1 Publicationst1.000
2 Degreet0.601 1.000
3 Flow Betweennesst0.831 0.778 1.000
4 K-Coret0.440 0.896 0.588 1.000
5 Mean Strengtht0.617 0.750 0.592 0.724 1.000
6 Degree Centralizationt0.147 0.396 0.185 0.511 0.539 1.000
7 Share in Main Componentt0.080 -0.066 0.059 -0.134 0.084 0.391 1.000
8 Kyoto Ratificationt−10.147 0.381 0.210 0.505 0.292 0.297 0.036 1.000
9 Cum. Number of Satellitest−10.191 0.191 0.177 0.105 0.095 0.104 -0.071 -0.010 1.000
10 Installed PV Capacityt−10.653 0.843 0.778 0.752 0.712 0.277 0.350 0.583 0.173 1.000
11 PV R&D Exp.t−10.537 0.282 0.395 0.077 0.243 -0.111 -0.056 -0.060 0.652 0.315 1.000
12 PV R&D Exp. interp. Dummyt−10.042 0.077 0.048 0.077 0.058 0.035 0.068 0.096 0.093 0.067 -0.058 1.000
13 GDP per Capitat−10.164 0.420 0.266 0.392 0.320 0.163 -0.116 0.132 0.081 0.396 0.178 0.144 1.000
14 EU Membershipt0.085 0.355 0.179 0.317 0.206 0.170 -0.137 0.175 -0.033 0.096 -0.172 0.047 0.310 1.000
15 Conventional R&D Exp.t−10.247 0.011 0.106 -0.140 0.045 -0.232 -0.232 -0.148 0.446 0.141 0.761 -0.027 0.051 -0.250 1.000
Table 9: Correlation table for the periods 1997-1999 until 2013-2015.
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
1 Publicationst1.000
2 Degreet0.602 1.000
3 Flow Betweennesst0.828 0.782 1.000
4 K-Coret0.433 0.884 0.580 1.000
5 Mean Strengtht0.616 0.728 0.579 0.698 1.000
6 Degree Centralizationt0.120 0.337 0.145 0.455 0.499 1.000
7 Share in Main Componentt0.116 -0.039 0.083 -0.138 0.148 0.403 1.000
8 Kyoto Ratificationt−10.074 0.245 0.136 0.354 0.162 0.171 0.064 1.000
9 Cum. Number of Satellitest−10.225 0.246 0.225 0.154 0.122 0.142 -0.024 -0.040 1.000
10 Installed PV Capacityt−10.613 0.814 0.743 0.678 0.609 0.029 0.498 0.384 0.194 1.000
11 PV R&D Exp.t−10.741 0.509 0.585 0.216 0.408 -0.103 0.261 -0.109 0.677 0.471 1.000
12 PV R&D Exp. interp. Dummyt−10.007 0.011 0.012 -0.010 0.000 -0.017 0.047 0.019 0.141 0.057 -0.036 1.000
13 GDP per Capitat−10.187 0.416 0.264 0.393 0.308 0.123 -0.073 0.078 0.073 0.226 0.204 0.125 1.000
14 EU Membershipt0.085 0.432 0.215 0.407 0.250 0.205 -0.057 0.172 -0.042 0.061 -0.258 0.084 0.358 1.000
15 Conventional R&D Exp.t−10.496 0.334 0.330 0.098 0.297 -0.166 0.202 -0.087 0.406 0.409 0.746 0.031 0.134 -0.345 1.000
26
Jena Economic Research Papers 2016 - 016

A.4 Regression results period 1997-1999 until 2013-2015
Table 10: OLS Panel regression results for country embeddedness periods 1997-1999 until 2013-2015.
Publications Degree Flow Betwenness K-Core
Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 Model 7 Model 8
National network variables
Mean Strengtht60.772*** 29.786*** 1.947*** 0.135 602.299*** 312.425 0.658*** -0.009
(15.923) (9.814) (0.297) (0.208) (189.412) (274.486) (0.136) (0.137)
Degree Centralizationt-292.796*** -273.114* -11.071*** -12.213* -4652.019*** -6442.561 -0.995 -0.392
(85.106) (157.808) (2.832) (6.566) (1261.829) (4937.240) (1.371) (1.655)
Share in Main Componentt-84.859* -72.020 6.204*** 7.556** 747.921 -743.784 1.076 0.388
(43.604) (59.082) (1.866) (3.456) (634.060) (1522.624) (0.765) (1.153)
Policy variables
Kyoto Ratificationt−1-5.920 15.729 1.143 3.301** -19.300 1187.732 0.309 -1.020*
(20.202) (25.504) (0.860) (1.591) (338.253) (801.680) (0.402) (0.521)
Cum. Number of Satellitest−12.984*** 2.994*** 0.034*** 0.011** 45.680*** 43.204*** 0.003 -0.003**
(0.948) (0.159) (0.011) (0.005) (12.337) (3.898) (0.002) (0.001)
Installed PV Capacityt−110.312 0.622** 686.463*** -0.260**
(8.020) (0.276) (248.741) (0.115)
PV R&D Exp.t−11.024*** -0.018* 20.052 0.000
(0.326) (0.010) (14.594) (0.002)
PV R&D Exp. interp. Dummyt−1-191.416 1.792 -2059.710 -0.221
(138.991) (2.134) (3566.510) (0.267)
Control variables
GDP per Capitat−1-0.003* -0.004* 0.000 0.000 0.000 -0.070 0.000* 0.000*
(0.001) (0.002) (0.000) (0.000) (0.029) (0.058) (0.000) (0.000)
EU Membershipt-16.459* -1.073 -682.242*** 0.685
(9.861) (1.580) (257.069) (0.721)
Conventional R&D Exp.t−1-0.059** 0.002 -1.212 0.000
(0.023) (0.001) (1.391) (0.000)
Country fixed effects yes yes yes yes yes yes yes yes
Time fixed effects yes yes yes yes yes yes yes yes
Adj. R2 0.436 0.582 0.307 0.202 0.335 0.410 0.143 0.174
n 114 18 114 18 114 18 114 18
T 17 17 17 17 17 17 17 17
N 1416 275 1416 275 1416 275 1416 275
df 1279 231 1279 231 1279 231 1279 231
Robust standard errors in parentheses. Sig. at *** 0.01, ** 0.5, * 0.1 level.
27
Jena Economic Research Papers 2016 - 016

References
Acemoglu, D., Aghion, P., Bursztyn, L., and Hemous, D. (2012). The environment and directed
technical change. American Economic Review, 102(1):131–66.
Adams, J. (2013). Collaborations: The fourth age of research. Nature, 497(7451):557–560.
Adams, J. D., Black, G. C., Clemmons, J. R., and Stephan, P. E. (2005). Scientific teams and
institutional collaborations: Evidence from U.S. universities, 1981-1999. Research Policy,
34(3):259–285.
Ahuja, G. (2000). Collaboration networks, structural holes, and innovation: A longitudinal
study. Administrative Science Quarterly, 45(3):425–455.
Arrow, K. J. (1962). Economic welfare and the allocation of resources for invention. In Nelson,
R., editor, The Rate and Direction of Innovative Activity: Economic and Social Factors,
pages 609–625. Princeton University Press, Princeton.
Aschhoff, B. and Sofka, W. (2009). Innovation on demand–can public procurement drive market
success of innovations? Research Policy, 38(8):1235–1247.
Barabasi, A., Jeong, H., Neda, Z., Ravasz, E., Schubert, A., and Vicsek, T. (2002). Evolution
of the social network of scientific collaborations. Physica, A 311(3–4):590–614.
Bathelt, H., Malmberg, A., and Maskell, P. (2004). Clusters and knowledge: Local buzz, global
pipelines and the process of knowledge creation. Progress in Human Geography, 28(1):31–56.
Boschma, R. (2005). Proximity and innovation: A critical assessment. Regional Studies,
39(1):61–74.
Bozeman, B. and Corley, E. (2004). Scientists’ collaboration strategies: implications for scientific
and technical human capital. Research Policy, 33(4):599–616.
Breschi, S. and Lissoni, F. (2004). Knowledge networks from patent data: methodological issues
and research targets. In Moed, H., Gl¨anzel, W., and Schmoch, U., editors, Handbook of
quantitative science and technology research: the use of publication and patent statistics in
studies on S&T systems. Springer, Berlin/Heidelberg/New York.
Cantner, U. and Graf, H. (2006). The network of innovators in Jena: An application of social
network analysis. Research Policy, 35(4):463–480.
Cantner, U. and Graf, H. (2011). Innovation networks: formation, performance and dynamics.
In Antonelli, C., editor, Handbook on the Economic Complexity of Technological Change,
chapter 15, pages 366–394. Edward Elgar, Cheltenham, UK.
Cantner, U., Graf, H., Herrmann, J., and Kalthaus, M. (2016). Inventor networks in renewable
energies: The influence of the policy mix in germany. Research Policy, 45(6):1165–1184.
Cantner, U. and Rake, B. (2014). International research networks in pharmaceuticals: Structure
and dynamics. Research Policy, 43(2):333–348.
Carlsson, B. and Stankiewicz, R. (1991). On the nature, function and composition of techno-
logical systems. Journal of Evolutionary Economics, 1(2):93–118.
Cowan, R. and Jonard, N. (2004). Network structure and the diffusion of knowledge. Journal
of Economic Dynamics and Control, 28(8):1557–1575.
28
Jena Economic Research Papers 2016 - 016

Crespi, G. A. and Geuna, A. (2008). An empirical study of scientific production: A cross country
analysis, 1981–2002. Research Policy, 37(4):565–579.
Dechezleprˆetre, A., Glachant, M., and M´eni`ere, Y. (2008). The clean development mecha-
nism and the international diffusion of technologies: An empirical study. Energy Policy,
36(4):1273–1283.
Defazio, D., Lockett, A., and Wright, M. (2009). Funding incentives, collaborative dynamics
and scientific productivity: Evidence from the eu framework program. Research Policy,
38(2):293–305.
Dong, B., Xu, G., Luo, X., Cai, Y., and Gao, W. (2012). A bibliometric analysis of solar power
research from 1991 to 2010. Scientometrics, 93(3):1101–1117.
Dopfer, K., Foster, J., and Potts, J. (2004). Micro-meso-macro. Journal of Evolutionary Eco-
nomics, 14(3):263–279.
Dosi, G. (1988). The nature of the innovative process. In Dosi, G., Freeman, C., Nelson,
R., Silverberg, G., and Soete, L., editors, Technical Change and Economic Theory, pages
221–238. Pinter, London.
Du, H., Li, N., Brown, M. A., Peng, Y., and Shuai, Y. (2014). A bibliographic analysis of recent
solar energy literatures: The expansion and evolution of a research field. Renewable Energy,
66(0):696 – 706.
Ebadi, A. and Schiffauerova, A. (2013). Impact of funding on scientific output and collaboration:
A survey of literature. Journal of Information & Knowledge Management, 12(04):1350037.
Edler, J. and Georghiou, L. (2007). Public procurement and innovation – resurrecting the
demand side. Research Policy, 36:949–963.
Ergas, H. (1987). Does technology policy matter? In Guile, B. and Brooks, H., editors,
Technology and Global Industry: Companies and Nations in the World Economy, pages
191–245. National Academy Press, Washington DC.
Feenstra, R. C., Inklaar, R., and Timmer, M. P. (2015). The next generation of the penn world
table. American Economic Review, 105(5):3150–3182.
Flanagan, K., Uyarra, E., and Laranja, M. (2011). Reconceptualising the ’policy mix’ for
innovation. Research Policy, 40(5):702 – 713.
Fleming, L., King, Charles, I., and Juda, A. I. (2007). Small worlds and regional innovation.
Organization Science, 18(6):938–954.
Freeman, L. C. (1979). Centrality in social networks conceptual clarification. Social Networks,
1(3):215–239.
Freeman, L. C., Borgatti, S. P., and White, D. R. (1991). Centrality in valued graphs: A measure
of betweenness based on network flow. Social Networks, 13(2):141–154.
Fritsch, M. and Graf, H. (2011). How sub-national conditions affect regional innovation systems:
The case of the two Germanys. Papers in Regional Science, 90(2):331–353.
Geroski, P. (1990). Procurement policy as a tool of industrial policy. International Review of
Applied Economics, 4(2):182–198.
Graf, H. (2011). Gatekeepers in regional networks of innovators. Cambridge Journal of Eco-
nomics, 35(1):173–198.
29
Jena Economic Research Papers 2016 - 016

Griffiths, S. (2013). Strategic considerations for deployment of solar photovoltaics in the middle
east and north africa. Energy Strategy Reviews, 2(1):125–131.
Groba, F. and Breitschopf, B. (2013). Impact of renewable energy policy and use on innovation.
Technical report, Deutsches Institut f¨ur Wirtschaftsforschung 1318.
Guan, J., Yan, Y., and Zhang, J. (2015a). How do collaborative features affect scientific output?
evidences from wind power field. Scientometrics, 102(1):333–355.
Guan, J., Zhang, J., and Yan, Y. (2015b). The impact of multilevel networks on innovation.
Research Policy, 44(3):545 – 559.
Guerzoni, M. and Raiteri, E. (2015). Demand-side vs. supply-side technology policies: Hidden
treatment and new empirical evidence on the policy mix. Research Policy, 44(3):726–747.
Gupta, A. K., Tesluk, P. E., and Taylor, M. S. (2007). Innovation at and across multiple levels
of analysis. Organization Science, 18(6):885–897.
Herstad, S. J., Aslesen, H. W., and Ebersberger, B. (2014). On industrial knowledge bases, com-
mercial opportunities and global innovation network linkages. Research Policy, 43(3):495–
504.
Hidalgo, C. A. (2015). Disconnected! the parallel streams of network literature in the natural
and social sciences. ArXiv e-prints.
Huang, M.-H., Dong, H.-R., and Chen, D.-Z. (2013). The unbalanced performance and regional
differences in scientific and technological collaboration in the field of solar cells. Sciento-
metrics, 94(1):423–438.
IEA (2016). International energy agency data service.
Jaffe, A. B., Newell, R. G., and Stavins, R. N. (2002). Environmental policy and technological
change. Environmental and Resource Economics, 22(1):41–70.
Jaffe, A. B., Newell, R. G., and Stavins, R. N. (2005). A tale of two market failures: Technology
and environmental policy. Ecological Economics, 54(2-3):164–174.
Johnstone, N., Haˇsˇciˇc, I., and Popp, D. (2010). Renewable energy policies and technological
innovation: Evidence based on patent counts. Environmental and Resource Economics,
45(1):133–155.
Kang, M. J. and Park, J. (2013). Analysis of the partnership network in the clean development
mechanism. Energy Policy, 52(0):543–553.
Katz, J. and Martin, B. R. (1997). What is research collaboration? Research Policy, 26(1):1–18.
Kemp, R. and Pontoglio, S. (2011). The innovation effects of environmental policy instruments
– a typical case of the blind men and the elephant? Ecological Economics, 72(0):28–36.
Laudel, G. (2002). What do we measure by co-authorships? Research Evaluation, 11(1):3–15.
Lee, S. and Bozeman, B. (2005). The impact of research collaboration on scientific productivity.
Social Studies of Science, 35(5):673–702.
Lundvall, B.-A. (1992). National Systems of Innovation: Towards a Theory of Innovation and
Interactive Learning. Pinter Publishers, London.
Malerba, F. (2002). Sectoral systems of innovation and production. Research Policy, 31:247–264.
30
Jena Economic Research Papers 2016 - 016

Mazzucato, M. (2013). The Entrepreneurial State: Debunking Public vs. Private Sector Myths.
Anthem Press, London.
Moloney, M., Smith, D. H., and Graham, S. (2014). A summary of nrc findings and recommen-
dations on international collaboration in space exploration. Technical report, 40th COSPAR
Scientific Assembly. Held 2-10 August 2014, in Moscow, Russia, Abstract PEX.1-8-14.
Mowery, D. and Rosenberg, N. (1979). The influence of market demand upon innovation: A
critical review of some recent empirical studies. Research Policy, 8(2):103–153.
Muhammad-Sukki, F., Munir, A. B., Ramirez-Iniguez, R., Abu-Bakar, S. H., Yasin, S. H. M.,
McMeekin, S. G., and Stewart, B. G. (2012). Solar photovoltaic in malaysia: The way
forward. Renewable and Sustainable Energy Reviews, 16(7):5232–5244.
Nelson, R. R., editor (1993). National Innovation Systems: A Comparative Analysis. Oxford
University Press, New York.
Nesta, L., Vona, F., and Nicolli, F. (2014). Environmental policies, competition and innovation
in renewable energy. Journal of Environmental Economics and Management, 67(3):396–411.
Newell, R. G. (2010). The role of markets and policies in delivering innovation for climate change
mitigation. Oxford Review of Economic Policy, 26(2):253–269.
Newman, M. (2001). The structure of scientific collaboration networks. Proceedings of the
National Academy of Sciences of the United States of America, 98(2):404–409.
Oliver, M. and Jackson, T. (1999). The market for solar photovoltaics. Energy Policy, 27(7):371–
385.
Owen-Smith, J., Riccaboni, M., Pammolli, F., and Powell, W. W. (2002). A comparison of
U.S. and European university-industry relations in the life sciences. Management Science,
48(1):24–43.
Ozman, M. (2009). Inter-firm networks and innovation: a survey of literature. Economics of
Innovation and New Technology, 18(1):39–67.
Paruchuri, S. (2010). Intraorganizational networks, interorganizational networks, and the impact
of central inventors: A longitudinal study of pharmaceutical firms. Organization Science,
21(1):63–80.
Peters, M., Schneider, M., Griesshaber, T., and Hoffmann, V. H. (2012). The impact of
technology-push and demand-pull policies on technical change – does the locus of policies
matter? Research Policy, 41(8):1296–1308.
Petroni, G., Venturini, K., and Santini, S. (2010). Space technology transfer policies: Learning
from scientific satellite case studies. Space Policy, 26(1):39–52.
Phelps, C., Heidl, R., and Wadhwa, A. (2012). Knowledge, networks, and knowledge networks:
A review and research agenda. Journal of Management, 38:1115–1166.
Poirier, J., Johnstone, N., Haˇsˇciˇc, I., and Silva, J. (2015). The benefits of international co-
authorship in scientific papers. the case of wind energy technologies. Technical report,
OECD Environment Working Papers, No. 81, OECD Publishing.
Polzin, F., Migendt, M., T¨aube, F. A., and von Flotow, P. (2015). Public policy influence on
renewable energy investments—a panel data study across OECD countries. Energy Policy,
80:98–111.
31
Jena Economic Research Papers 2016 - 016

Popp, D. (2002). Induced innovation and energy prices. American Economic Review, 92(1):160–
180.
Popp, D. (2016). Economic analysis of scientific publications and implications for energy research
and development. Nature Energy, 1(4):16020.
Powell, W. W., Koput, K. W., and Smith-Doerr, L. (1996). Interorganizational collaboration
and the locus of innovation: Networks of learning in biotechnology. Administrative Science
Quarterly, 41(1):116–145.
Powell, W. W., Koput, K. W., Smith-Doerr, L., and Owen-Smith, J. (1999). Network position
and firm performance: Organizational returns to collaboration in the biotechnology indus-
try. In Andrews, S. B. and Knoke, D., editors, Research in the Sociology of Organizations,
pages 129–159. JAI Press, Greenwich, CT.
Rennings, K. (2000). Redefining innovation - eco-innovation research and the contribution from
ecological economics. Ecological Economics, 32(2):319 – 332.
Rogge, K. S. and Reichardt, K. (2016). Policy mixes for sustainability transitions: An extended
concept and framework for analysis. Research Policy, 45(8):1620–1635.
Schilling, M. A. and Phelps, C. C. (2007). Interfirm collaboration networks: The impact of
large-scale network structure on firm innovation. Management Science, 53(7):1113–1126.
Seidman, S. B. (1983). Network structure and minimum degree. Social Networks, 5(3):269–287.
Stek, P. E. and van Geenhuizen, M. (2015). Mapping innovation in the global photovoltaic
industry: A bibliometric approach to cluster identification and analysis. Technical report,
ERSA 55th Congress, World Renaissance: Changing roles for people and places, Lisbon,
Portugal, 25-28 August 2015.
Suominen, A. (2014). Phases of growth in a green tech research network: a bibliometric evalu-
ation of fuel cell technology from 1991 to 2010. Scientometrics, 100(1):51–72.
Ubfal, D. and Maffioli, A. (2011). The impact of funding on research collaboration: Evidence
from a developing country. Research Policy, 40(9):1269–1279.
UNFCC (1997). Kyoto Protocol To The United Nations Framework Convention On Climate
Change. FCCC/CP/1997/L7/Add1, Kyoto.
Uzzi, B., Amaral, L. A. N., and Reed-Tsochas, F. (2007). Small-world networks and management
science research: a review. European Management Review, 4(2):77–91.
Verspagen, B. and Duysters, G. (2004). The small worlds of strategic technology alliances.
Technovation, 24:563–571.
Wagner, C. S. and Leydesdorff, L. (2005). Mapping the network of global science: comparing
international co-authorships from 1990 to 2000. International Journal of Technology and
Globalisation, 1(2):185–208.
Wangler, L. U. (2013). Renewables and innovation: did policy induced structural change in the
energy sector effect innovation in green technologies? Journal of Environmental Planning
and Management, 56(2):211–237.
Wanzenb¨ock, I., Scherngell, T., and Brenner, T. (2014). Embeddedness of regions in European
knowledge networks: a comparative analysis of inter-regional r&d collaborations, co-patents
and co-publications. Annals of Regional Science, 53(2):337–368.
32
Jena Economic Research Papers 2016 - 016

Wanzenb¨ock, I., Scherngell, T., and Lata, R. (2015). Embeddedness of european regions in
european union-funded research and development (R&D) networks: A spatial econometric
perspective. Regional Studies, 49(10):1685–1705.
Wassermann, S. and Faust, K. (1994). Social Network Analysis: Methods and Applications.
Cambridge University Press, Cambridge.
Watanabe, C., Wakabayashi, K., and Miyazawa, T. (2000). Industrial dynamism and the cre-
ation of a ”virtuous cycle” between r&d, market growth and price reduction: The case of
photovoltaic power generation (pv) development in japan. Technovation, 20(6):299–312.
West, J. (2014). Too little, too early: California’s transient advantage in the photovoltaic solar
industry. The Journal of Technology Transfer, 39(3):487–501.
Wuchty, S., Jones, B. F., and Uzzi, B. (2007). The increasing dominance of teams in production
of knowledge. Science, 316(5827):1036–1039.
Zheng, C. and Kammen, D. M. (2014). An innovation-focused roadmap for a sustainable global
photovoltaic industry. Energy Policy, 67(0):159–169.
33
Jena Economic Research Papers 2016 - 016