Master's Thesis Shamiram Abdulahad

Collaboration in the Circular Economy in
the Netherlands: A Proximity Approach
MSc thesis Innovation Sciences, Utrecht University (GEO4-2239X)
Shamiram Abdulahad
s.abdulahad2@students.uu.nl
Student number: 6426214
Supervisor: Dr. K. Frenken

Second reader: Dr. I. Wanzenböck
September 7, 2020
1
Table of Contents
Title Page ........................................................................................................................................................ 1

Abstract ........................................................................................................................................................... 3
1. Introduction............................................................................................................................................ 4
2. Theoretical framework .......................................................................................................................... 7
2.1 Research question 1 ............................................................................................................................. 7
2.2 Research question 2 ............................................................................................................................. 9
2.3 Research question 3 ........................................................................................................................... 12
3. Method ................................................................................................................................................. 13
3.1 Data collection.................................................................................................................................... 13
3.2 Data analysis ....................................................................................................................................... 17
4. Results................................................................................................................................................... 19
4.1 Network visualisation ......................................................................................................................... 19
5. Conclusion............................................................................................................................................ 32
6. Discussion............................................................................................................................................. 33
Acknowledgements....................................................................................................................................... 35
References..................................................................................................................................................... 36
Appendix 1 ................................................................................................................................................... 41
Appendix A .............................................................................................................................................. 41
Appendix B .............................................................................................................................................. 44
Appendix C .............................................................................................................................................. 45
Appendix D .............................................................................................................................................. 46
Appendix 2 ................................................................................................................................................... 51
2
Abstract
The extraction of raw material is responsible for the majority of the global GHG emissions. Waste
management is an important aspect of climate change mitigation. The current economic development
model is a linear model, which used too much raw materials. This is unsustainable for the earth. An
important solution to this problem is the implementation of a circular economy. In this paper, the
hindering factors for a Circular Economy in the Netherlands are identified. For this research, the
EUPRO database is used. The proximity dimensions and centrality measures are used to examine the
collaborations in the Netherlands. A social network analysis and count regressions are carried out to
measure the impact of these variables. In the Netherlands, organizations require collaborate with
organizations that have higher geographical proximity, institutional proximity and cognitive proximity
than is beneficial for the success of the collaboration. Thus, this paper provides empirical evidence for
the proximity paradox. Organizations that are well-connected also have a negative influence on the
success of the collaborations in the Netherlands. For these findings, policy recommendations are then
considered.
3
1. Introduction
Climate change is continuously becoming a more pressing issue, as global temperatures continue to rise.
This is reflected in the increasing focus on sustainability (Rajeev et al., 2017). Sustainability is considered
to be the societal issue of this time (TNO, 2019). Climate change poses increasingly severe risks for the
European ecosystems, general health, and economy (European Environment Agency, 2017). Especially
in the Netherlands, the change towards a more sustainable society is slow. In 2017, the Netherlands had
the second lowest share of energy from renewable sources from the EU member states (Eurostat, 2019).
One important aspect of climate change is waste management. Waste management is important for climate
change mitigation for two reasons. First, there are significant non-energy sources of greenhouse gasses
(GHGs), including methane emissions from landfills (Ackerman, 2000). Composting bio-waste, for
example, greatly reduces the amount of GHGs released compared to landfills (Lou & Nair, 2009). Second,
changes in waste management have surprisingly large effects on the way energy is used (Ackerman, 2000).
Industries extracting raw materials are the most energy-intensive branches of the manufacturing process.
Much less energy is used to shape the materials. Thus, recycling of raw materials or using less, reduces
energy use and associated carbon emissions in the most energy-hungry industry branches (Ackerman,
2000).
Currently, the dominant economic development model is a linear economy (Ghisellini et al., 2016). A
linear economy is an economy where the general pathway is the extraction of raw materials, followed by a
transformation into a product which is then used until it has served its purpose, after which it is discarded
as waste, also called “take, make and dispose”. Value is created in this economic system by producing and
selling as many products as possible (Kenniskaarten, 2016). However, research shows that this method is
unsustainable and the negative effects caused by this economic system are threatening the stability of the
economies and the integrity of natural ecosystems that are essential for humanity's survival (Yap, 2005;
Yuan et al., 2006; Feng & Yan, 2007; Ellen McArthur Foundation, 2012; Geng et al., 2012; Preston, 2012;
Stiehl & Hirth, 2012; Su et al., 2013; UNEP, 2013; Waughray, 2013; EC, 2014a; EC, 2014b; Lett, 2014;
Mazzatini, 2014; Park & Chertow, 2014). Thus, it is important to change our economic development
model into something more sustainable.
An increasingly popular solution is the transformation of the economic development model into a circular
economy. Because it is a trending topic, many different definitions have been assigned to it (Kirchherr et
al., 2017). The definition used in this paper is the definition from Kirchherr et al. (2017): “an economic
system that replaces the ‘end-of-life’ concept with reducing, alternatively reusing, recycling and recovering
materials in production/distribution and consumption processes”. Thus, in a circular economy, raw
material extraction and waste generation is highly reduced and ideally even fully diminished. According to
Circle Economy (2019), 62% of global GHGs (excluding those from land use and forestry) come from
the extraction, processing and production of goods to meet society’s needs. Global GHGs may be reduced
even further because the transport of these materials will also no longer be necessary in a circular economy.
Besides environmental benefits, there are also economic benefits. Since 2000, the price of raw materials
has strongly increased (WE Forum, 2014). Switching to a circular economy would mean a decrease in
costs by saving the expenses of the raw materials (Kenniskaarten, 2016). Furthermore, by decoupling
economic growth from the availability of raw materials, a hindering factor for economic growth is removed.
In these ways, the economy may grow into a circular economy.
4
In 2015, The European Commission published its Circular Economy Package, with the stated objective
of "closing the loop" of product lifecycles (Hughes, 2017). According to the Dutch Environmental
Assessment Agency (PBL) (Rood & Hanemaaijer, 2017), the Netherlands has a very promising starting
position for circular economy. The Netherlands has the highest population density and infrastructure
density in Europe and has one of the highest densities of urban area, densities of raw material streams,
percentage of recycled household waste, and amount of patents on waste collection and recycling (Rood
& Hanemaaijer, 2017). Thus, a transition towards a circular economy would be a good measure for climate
change mitigation and waste management in the Netherlands.
Despite the promising starting position, the Netherlands is still far from having a circular economy. The
main reason for this is that, for changes to be implemented towards a circular economy, collaboration is
required between different actors of the society (Ghisellini et al., 2016). Stakeholders need to work
together in new alliances within production chains to achieve a circular economy (Rood & Hanemaaijer,
2017). Stakeholders in a circular economy are also more interdependent than in a linear economy
(Ashton, 2008; Ashton, 2009). The reason for this is that in the transition to a Circular Economy, data
needs to be collected and shared (Preston, 2012). Furthermore, for a successful implementation of a
circular economy, upfront integration of a variety of disciplines in the supply chain, with the co-creation
of an ambitious vision, and responsibilities need to be extended to the actors along the entire supply chain
(Dora, 2019). The reason for this is that the Circular Economy means the cycle of the chain of several
organizations is closed. This means that these organizations are dependent on each other for their raw
materials and if one link in the chain fails the rest of the chain suffers (Kenniskaarten, 2016). The
collaboration between the different actors thus is not yet sufficient. The reason for this most likely comes
from the fact that, the change to a system where the interdependency between the firms is higher comes
with a lot of uncertainty.
Collaboration can provide the benefit of resource sharing, allowing firms to combine knowledge, skills,
and physical assets. Second, collaborative linkages can provide access to knowledge spill overs, serving as
information conduits through which news of technical breakthroughs, new insights to problems, or failed
approaches travels from one firm to another (Ahuja, 2000). In order to stimulate the circular economy,
collaboration between the stakeholders of the circular economy should be improved. To do that a network
analysis should be conducted. From this network analysis, gaps in collaboration between different groups
can be detected.
In terms of circular economy, a lot of focus has been put on SMEs, business models, and studying the
Technological Innovation Systems (TIS) surrounding the circular economy. The collaboration between
the firms, or organizations generally, has not yet been studied in the Netherlands. This research gap could
provide new information because the ties that form could predict the way the circular economy will be
implemented. This could result in new insights and policy recommendations to stimulate the transition.
Because this paper studies the collaboration between companies, this study is carried out on a meso-level.
This leads us to the research questions:

RQ1: What drives collaborations in the circular economy in the Netherlands?
RQ2: What is the impact of the makeup of the collaborations on the chance of a successful outcome of
the collaboration?
5
RQ3: What are the missing links in the Dutch collaboration network, if any?
By focusing on the collaboration within the circular economy, this paper presents a network analysis during
a system transformation. The makeup of the collaborations within the systematic transformation can show
the focus of the organizations and the type of technologies that can result from these collaborations.
Furthermore, the focus is on the policy implications of the results in the network analysis, by trying to
identify which extra link can increase the total connectivity most efficiently and which links are crucial by
finding out which cause most loss of connectivity when they disappear.
By studying the collaborations that will form in a circular economy, this paper can examine the hindering
factors of the organizations within the system. This will help policymakers to understand which efforts are
lacking, and thus which efforts will need to be supported. In this way, the more efficient or sustainable
type of efforts, projects, and organizations can be discovered, and collaborations focused on them can be
supported, which will help the transition towards a circular economy. Furthermore, the methods used in
this paper are largely generalizable to other countries, thus allowing policymakers to apply similar
measures to improve the collaborative efforts in those countries.
6
2. Theoretical framework
To study what collaborations are formed in the circular economy, we look into theories on collaboration.
We will build largely on network theory, and the key notion of ‘proximity’. On this basis, we can perform
a network analysis s performed. In a social network analysis, the ties between parties are studied. This
allows us to uncover patterns in collaborations within the domain of circular economy. Furthermore,
finding places where collaboration is lacking, can show where investment is helpful to further stimulate the
circular economy.
2.1 Research question 1

Nieto and Santamaria (2007) identify four main types of collaborations for firms within an industry: 1)
collaboration with Research Organizations; 2) collaboration with clients; 3) collaboration with suppliers;
and 4) collaboration with competitors. The different types of collaborations tend to be associated with
different types of innovations. Collaborations with ROs and educational facilities are very important for
the creation of new (mostly scientific) knowledge. Governments are often involved in the regulations and
standardisation in upcoming technological systems. Firms who collaborate with suppliers, on the other
hand, are stimulating the transition to a new technological system (Luo et al., 2018). For platform
technologies especially, collaborations with competitors, also called co-opetition, is a prevalent business
model (Ruutu et al., 2017).
In this paper collaboration between all types of organizations is examined. To understand why some
organizations collaborate and others do not, the theory of proximity is helpful. Proximity between
organizations reduces uncertainty in a collaboration, and stimulates learning, knowledge creation and
innovation (Amin & Wilkinson, 1999; Boschma, 2005). Therefore, higher proximities between
organizations will increase the chance of a collaboration between the organizations. Thus, one of the
independent variables for the presence of a collaboration is the proximity between the organizations.
There are five dimensions of proximity (Boschma, 2005; Hardeman et al., 2014): cognitive, organizational,
social, institutional and geographical proximity. Cognitive proximity refers to the ease of knowledge
transfer and mutual learning between actors. This may depend on the similarity of their knowledge bases.
However, regarding cognitive proximity, only a certain level of proximity is needed in order to ensure
mutual understanding. Reversely, a certain level of cognitive distance creates opportunities for new
combinations of knowledge and technologies to occur, leading to innovation. Hence, one expects
organizations to seek partners that are similar, but not too similar. That is, one expects that the association
between cognitive proximity and the probability of collaboration follows an inverted U-shape (Broekel &
Boschma, 2012).
Organizational proximity refers to the extent to which networks occur within the context of an
organizational arrangement, and the extent to which any two actors are under shared hierarchical control.
Social proximity refers to the extent that two actors have established a friendly relation in the past.
Institutional proximity refers to parties that operate under the same set of norms and values. Both formal
and informal institutions structure behaviour by providing particular incentives. Geographical proximity
refers to physical distance, viz. transportation costs (Boschma, 2005). In this research, the social proximity
is not studied due to lack of time. Boschma (2005) explains how the different dimensions of proximity
can promote collaboration. The spreading of tacit knowledge is enhanced through geographical proximity.
This effect can even be seen in codified knowledge, because its application may need further information.
Furthermore, short distances between of similar activities in transparent clusters ensures successful
projects do not go unnoticed. This way, successful projects can be picked up by other organizations without
7
high costs. This way, geographical proximity stimulates collaboration between organizations. Institutions
can be enabling or restricting mechanisms for innovation. Interactive learning and economic coordination
are stimulated by a common language, shared habits, a law system securing ownership and intellectual
property rights, etc. A culture of shared trust is a capability that enhances learning and innovation. The
reason for this is that information can be transmitted more easily when there is cultural proximity and a
common language. Thus, proximity in the institutional rules, institutional proximity, can be an enabling
factor for interactive learning and innovation. Hierarchical organizations or tight relationships between
organizational units can create strong control mechanisms, which can ensure ownership rights and
sufficient rewards for investments in new technology. This can reduce the uncertainty and opportunism
that is present with the creation of new knowledge. Furthermore, strong ties between units can stimulate
the transfer of complex knowledge. Thus, organizational proximity can stimulate new knowledge creation
and the transfer of complex knowledge. Finally, the effective transfer of knowledge requires the absorptive
capacity to identify, interpret and exploit the new knowledge (Cohen & Levinthal, 1990). Organizations
identify new opportunities in close proximity to their existing knowledge base. This also constrains their
improvement in other directions, as knowledge and innovation are cumulative. The outcomes of these
search processes within firms have a high degree of tacit knowledge and localization. This means cognitive
differences will persist as long as the firm-specific competences are difficult to imitate by competitors.
Thus, cognitive proximity stimulates similar searches in opportunities. An inverted U-shape relationship
is suggested between the probability to collaborate and the cognitive proximity of two organizations
(Mowery et al., 1998; Broekel & Boschma, 2012). The reason for this is that organizations are less likely
to collaborate with an organization that does not have any different competences from them, as this will
not result in new knowledge or skills. This will be further discussed below.
In empirical research, the notion of proximity is most used in studies on trade agreements between
countries is through the gravity model (Biggiero & Basevi, 2009; Kepaptsoglou et al., 2010). Trade
agreements can be seen as collaborations between countries. Similarly, one can use this model to analyse
collaboration in the circular economy. In the gravity model, the GDP of a country and the distance
between the capitals are typically used to predict trade agreements. As this social network analysis focuses
on organizations, the four dimensions of proximity are used to indicate distance instead of only
geographical distance. Instead of the GDP, the organization size is used in the social network analysis. The
size of organizations is seen as an important factor for having many collaborations, because having more
resources available to invest in collaborations makes pursuing those collaborations more likely. The size
of organizations is correlated with the use a firm can make of knowledge created by universities (Cohen et
al., 2002; Guena et al., 2003). Furthermore, bigger firms have more resources available for R&D. Higher
R&D employments are correlated with more open innovation (Cohen et al., 2002).
Hypothesis 1: The size of organizations positively influences the chance for a presence of a collaboration.
Hypothesis 2: The proximity between organizations positively influences the chance of collaboration
between the organizations.
Following the proximity framework, Hypothesis 2 will be divided in four sub-hypotheses:

Hypothesis 2a: The Geographical proximity between organizations positively influences the chance of
collaboration between the organizations.
Hypothesis 2b: The Institutional proximity between organizations positively influences the chance of
8
Hypothesis 2c: The Organizational proximity between organizations positively influences the chance of
Hypothesis 2d: Cognitive proximity between organizations positively influences the chance of
collaboration between the organizations, but only up to a certain level (‘inverted-U shape)’.

While there may be collaboration between organizations, this does not mean that new knowledge is being
created. Collaborations may actually be focused more on exchange rather than the production of new
knowledge (Bozeman & Youtie, 2016). What is more, collaborations may fail, for example, when conflicts
emerge, or unforeseen problems hinder the production of new knowledge. Thus, in order to study
whether the collaborations between organizations will promote knowledge production, the outcomes of
the collaborations are to be assessed. The output of the collaborations that is examined is the success of
the collaborations.
According to Miller (2006), highly diverse partner capabilities may reduce the innovative benefits a
collaboration could provide. The reason for this would be that organizations can only adopt practices and
capabilities that are similar to their own. On the other hand, innovation cannot be achieved from
combining capabilities that are very similar to existing capabilities within the organization. Organizations
can gain more knowledge from collaborating with partners with different capabilities. According to van
Rijnsoever et al. (2015), having a higher group diversity promotes technological diversity. Sampson (2007)
found that R&D alliances with moderate diversity contribute more to firm innovation than alliances with
very low or very high levels of capability diversity. Thus, in case of a network analysis, having a diversity of
the actors in the project is likely to be important to the chance of a successful project. Diversity can be
understood as a lack of one or more of the five dimensions of proximity (van Rijnsoever et al., 2015).
Boschma (2005) explained how each dimension of proximity may have a negative impact on the
performance.
Too much geographical proximity can hinder interactive learning and innovation. This can happen when
regions stop looking outward for opportunities. This way the learning ability of the actors may be thus far
reduced that organizations lose their innovative capacity. This prevents them from being able to respond
to new developments. This is called spatial lock-in. Spatial lock-in is especially likely in highly specialized
regions, because interactive learning will be hindered due to excessive cognitive proximity between local
actors in specialized regions. Institutional proximity may also become a hindering factor to innovative
performance. Institutional environments consist of interdependencies among the actors in the
environment. If all actors adhere to the same institutions, following the same incentives, knowledge
production will be less creative. Instead, by bringing together organizations that adhere to different
institutions, like in university-industry-government partnerships, complementary knowledge can be
brought together creating win-win-win opportunities. Too much organizational proximity can also be
hindering innovation and learning. Hierarchical dependencies between organizational units and a parent
organization may limit creativity, innovation and interactive learning. New ideas are not rewarded in a
bureaucratic system, because hierarchical governance lacks feedback mechanisms that are present in
symmetrical relations. This hinders the occurrence of interactive learning. Strong ties also limit access to
sources of new knowledge and skills outside of the established channels. Thus, a certain level of autonomy
and decentralization is helpful for organizational units. Too much cognitive proximity, finally, can also
hinder innovative performance. One reason for this is that creating new knowledge requires
heterogeneous, complementary knowledge bases. New sources create new ideas and stimulate creativity.
9
Moreover, routines within an organization can hinder the search for new technologies and new market
possibilities. This can create a cognitive lock-in. Here the cumulative nature of knowledge creates a
disadvantage for the organization. This is called the competency trap. Finally, higher cognitive proximity
results in higher chance of involuntary knowledge spill-overs. Due to the higher cognitive proximity, the
partner organizations have the same absorptive capacity, which increases the chance of spill-overs.
According to Boschma (2005), and consistent with Sampson (2007), the relation between cognitive
proximity and collaboration success follows an inverted-U shape, indicating that there is an optimal level
of cognitive proximity. A certain level of cognitive proximity is needed in order to ensure mutual
understanding, but a certain level of cognitive distance is needed to ensure opportunities for new
combinations of knowledge and technologies to occur, leading to innovation.
Hypothesis 3a: The geographical proximity between the organizations in a collaboration negatively
influences the chance of a successful project.
Hypothesis 3b: The institutional proximity between the organizations in a collaboration negatively
Hypothesis 3c: The organizational proximity between the organizations in a collaboration negatively
Hypothesis 3d: The cognitive distance between the companies in a collaboration positively influences the
chance of a successful project, but only up to a certain level (‘inverted-U shape’).
As mentioned in the theory for hypotheses 2 and 3, proximity can have different effects on the outcome
of a collaboration. Boschma and Frenken (2010) introduced the idea of a proximity paradox. The paradox
describes that while proximity provides an important incentive for organizations to collaborate and
exchange knowledge, too much proximity may harm the innovative performance of the organizations. The
high proximity dimensions reduce the uncertainty of the collaboration. However, this reduces the gain
from that collaboration. Broekel and Boschma (2012) argued that the proximity levels at which
organizations tend to collaborate may be higher that the proximity levels that would be optimal for a
successful collaboration. This would reflect biases by organizations to avoid risks by collaborating with
proximate partners, even though more distant partners would yield better results. This would mean
organizations are not gaining the most out of the collaborations they participate in. To test this hypothesis,
one can combine the various hypotheses as formulated under hypothesis 2 and 3. We thus expect that
geographical proximity positively affects the probability of collaboration (hypothesis 2a), but negatively
affects the success of the collaboration (hypothesis 3a). Likewise, we expect that institutional and
organizational proximity positively affects the probability of collaboration (hypothesis 2b and 2c), but
negatively affects the success of the collaboration (hypothesis 3b and 3c). Regarding cognitive proximity,
we expect that the optimum of the inverted-U shape curve lies at a higher level of cognitive proximity when
explaining collaboration than when explaining success. This would indicate that organizations tend to
collaborate with partners with a higher cognitive proximity than what would be optimal for successful
collaboration.
As mentioned above, inter-organizational collaborative linkages provide benefits through resource sharing,
allowing organizations to combine knowledge, skills, physical assets, and access to knowledge spill overs,
serving as information conduits. Three aspects of an organization’s network structure are likely to be
relevant in connection with the above benefits: the number of direct ties maintained by an or, the number
of indirect ties maintained by the organization (the organizations it can reach in the network through its
10
partners and their partners), and the degree to which an organization's partners are linked to each other
(i.e., whether there are structural holes in the organization's ego network) (Ahuja, 2000). One way this can
be measured is by studying the centrality measures. Centrality within a network shows how important that
vertex is within the network. This can be seen as the equivalent of the size of the organization. However,
with network centrality the focus lies on the resources that can be gathered from the network, whereas the
size of the organization focuses on the amount of resources present in the organizations. Of course, size
and centrality are often linked. Degree centrality refers to the count of the other organizations with which
an organization is connected in the network, regardless of the direction. Organizations can benefit from
connecting organizations to other organizations that are unconnected to each other. Brokerage centrality
is defined as the number of organization pairs that are indirectly connected due to the (one-step)
intermediation by the organization in question. Actors that hold a broker position in a network are more
likely to express ideas, are less likely to have those ideas dismissed, and are more likely to have those ideas
evaluated as useful (Burt , 2004). Betweenness centrality refers to the extent to which an organization lies
on indirect knowledge flow paths between other organizations. Betweenness centrality for a given
organization is the number of shortest paths between other organizations which pass through the
organization in question. Finally, Eigenvalue centrality refers to a node’s importance while giving
consideration to the importance of its neighbours. The main principle is that links from important nodes
(as measured by degree centrality) are worth more than links from unimportant nodes (Golbeck, 2013;
Nomaler et al., 2014).
Hypothesis 4: The centrality of the organizations positively influences the chance of a successful project.
Because multiple centrality measures are used, Hypothesis 4 can be divided into four sub hypotheses:
Hypothesis 4a: The degree centrality of the organizations positively influences the chance of a successful
project.
Hypothesis 4b: The betweenness centrality of the organizations positively influences the chance of a
successful project.
Hypothesis 4c: The brokerage centrality of the organizations positively influences the chance of a
successful project.
Hypothesis 4d: The eigenvector centrality of the organizations positively influences the chance of a
successful project.
For this research, the type of collaboration is a government funded project. In order to measure the output,
performance indicators are needed. There are several ways to measure the success of a project (De Wit,
1988). However, these measures are related to input, client/team satisfaction, or time management. These
measures seem to be related to in-firm projects. For this research, these measures are not as relevant to
measure success. Instead the impact of the project is more relevant to whether the findings are relevant.
As the circular economy is still in the exploration phase, it may be more relevant to use a measure similar
to scientific entities. One way that success can be measured, is by counting the amount of documents
(articles) that are produced and the number of citations for each article (Acuna et al., 2012). As good data
on citations rates are only available for scientific articles, and the project the outputs are reports rather
than scientific articles, we do not use citation as a success variable. Instead, we only measure success here
by the amount of documents that are produced.
11
An important factor for the performance of an industry is the structure of the network. One important
aspect of this is the network density. According to one view, densely embedded networks with many
connections are facilitative, and social structures are seen as advantageous to the extent that networks are
"closed" (Coleman, 1988; Walker, Kogut, and Shan, 1997). According to an alternate view, however, social
structural advantages derive from the brokerage opportunities created by an open social structure (Burt,
1992; Ahuja, 2000). Schilling & Phelps (2007) added to this debate that local density and global efficiency
can exist simultaneously, and it is this combination that enhances innovation. Based on this information,
the way to most efficiently stimulate innovation is to stimulate the connections that create local density and
global efficiency. In other words, collaborations that optimally decrease the average path length or increase
the centrality in the network need to be stimulated.
Given the network of collaboration that we obtain from our data collection, we are able to look for “missing
links”. Missing links detection helps to detect highly likely but non-existent links, which can be used as
recommendations to collaborate (Luo et al., 2016). Identifying missing links evaluates existing policies in
terms of their possible inability to connect important organizations. The analysis can give clues for new
policy in terms of what collaboration to target in the future. Such missing links can be identified by
checking, for each possible link that can be added to the network, to what extent the average path length
in the network decreases. For example, as an extreme case, consider a network with two components
(meaning they have an infinite distance). Adding a link between the two components means that the
average path length decreases enormously. A less extreme example is a network with one component but
two different clusters, each with one central node that are not connected. Connecting the central nodes of
each cluster would maximally decrease the average path length. An example of missing link prediction
being done based on path length is the Katz similarity. Katz (1953) describes the similarity between nodes
based on the global path between the nodes. This is done by counting the number of paths between two
nodes and using this to calculate the similarity. The greater the number of paths between to nodes, the
greater the similarity. However, longer paths contribute less to the similarity.
12
3. Method
3.1 Data collection
3.1.1 Dataset composition
For this network analysis, this paper uses projects in which the actors collaborated to establish a network
analysis. This data is collected from the EUPRO dataset, which a database from the Research
infrastructure for research and innovation policy studies (RISIS, www risis.eu). The EUPRO dataset
provides a cleaned dataset on the R&D projects and the participants of these projects. The dataset consists
of 96,674 projects and 526,564 participants. This database was created by the Austrian Institute of
Technology (AIT) in 2005. It was created to aid in the analysis of participation patterns of organizations
in and across different European funding initiatives and the investigation of collaborative network
structures, including their evolution over time (Heller-Schuh et al., 2019). The EUPRO dataset consists
of four components: The CORDIS projects database, the EUREKA funding network, the Joint
Technology Initiatives (JTI) ARTEMIS, ENIAC and ECSEL programmes, and the COST funding
organization database. The CORDIS database contains projects and organizations funded by the
European Union for the Horizon 2020 Framework Programme (FP) from 2014 until 2020. EUREKA is
a public network for international cooperation in R&D and innovation to which the European Union is
subscribed. Eureka is an open platform for international cooperation in innovation with a focus on Small
and Medium Enterprises (SME). Eureka has provided 3,636 million euros in funding to this day (Eureka,
2020) and JTI 2,560 million euros (JTI, 2020). The quality of the raw data extracted from the different
programmes websites is not generally sufficient for policy-relevant analyses. AIT has undertaken
substantial efforts to improve quality and the level of standardisation of the data and to retrieve and add
missing data. Data cleaning and standardisation includes three major steps: identification of unique
organization name, identification of unique organization type, and regionalisation (European NUTS
regions)
The EUPRO dataset is used to gather data on which projects received subsidies/grants for investments
into circular economy in the Netherlands. In this database, the parties involved in the projects and the
results are published. To collect the data about the projects, RISIS was contacted with a set of keywords
(Appendix 1A). These keywords were found through the list of CPC codes from Eurostat (see Appendix
1A). These CPC codes are the patent codes CDC associates with circular economy. The main topics from
these patents were used as keywords. Furthermore, a literature research was done to find the terms most
related to circular economy. For this part of the research theoretical saturation was reached. A link with
clean database with the data about all of the projects was sent. Using this, a network analysis is done on
the collaborations in the circular economy. From these pages, the project, the parties involved, and the
end result are collected. As this paper’s main focus is on the Netherlands, projects with organizations in
the Netherlands are used. After filtering on the projects in the Netherlands, the projects were screened to
ensure that all the projects were related to the circular economy. This was done by checking titles, short
descriptions, and abstracts. From the 1011 projects in the Netherlands, 876 projects remained after the
screening, from which 305 were unique (sometimes records were present several times in a dataset). From
this dataset, projects from the most recent 10 years was used. As the dataset was based data until 2017, the
projects from 2007 to 2017 were used. After, 97 unique projects and 632 unique organizations remained.
As the non-Dutch organizations are not shown for the links with other non-Dutch organizations (which
are most likely most of their links), the network is not complete for the organizations that are not based in
the Netherlands. Thus, the binary logit model of the first research question can only be performed for all
Dutch firms. As the analysis of research question 2 is dependent on the network centrality, the same
problem is present for non-Dutch firms in the second research question. Thus, the non-Dutch
organizations need to be filtered out from the dataset. After filtering for organizations based in the
Netherland, only 139 unique organizations remained. The number of projects remained the same.
13
The dataset consists of 7 documents in total, each containing different variables and information. First, the
“projects_export” file was used. This was the file containing the projects and their information. Here, the
projects were filtered for relevancy, as mentioned above. Then, the “participations_export” file was used.
This file contains information on which participants were present in each project, as well as information
on those participants. For organizations with multiple locations, the city name was added to the
organizations name (stApplicant). A separate column was made containing all the name of the main
organization (MainOrg). For organizations that do not have other locations or overarching structures, the
names of the organization was copied into the new column. These files are combined by the project ID
(RecCtrNr). This means that if a collaboration happens in in multiple RecCtrNrs, it will appear multiple
times. After these are combined, the independent variables need to be created. The dependent variable
(Collab), will be created by combining all duplicates of the same collaboration into a count of the
collaboration. If there is no collaboration for that combination of organizations, Collab is 0. If there are
one or more collaborations for that combination, Collab is 1.
For the independent variables, ten variables from the dataset were used (see table 1): (1) sCity, (2) sCountry
(3) stOrgtyp, (4) MainOrg, (5) emp (size of the organization), (6) StartDate, (7) ProjectEUfunding, (8)
noParticipants. The categories for stOrgtyp are industry organizations (IND), public and private research
organizations (ROR), universities and other educational facilities (EDU), governmental institutions
(GOV), non-commercial/non-profit organizations (NCL), Consultancies (CON, and special interest
groups, like unions, chambers, inter-trade organizations, etc. (OTH). Furthermore, there were 72 rows
with the presence or absence of organizations in the 72 GICS matched. This was used to create the variable
(9) CognProx and (10) sqCogn. For the geographical proximity, the county and city of the first organization
in the collaboration is compared to the county and city of the second organization. If they were in the
same city, Geoprox is TRUE. For the Institutional proximity, if the stOrgtyp of both organizations
matched, Insprox is TRUE. The Organizational proximity (Orgprox) is TRUE, if the MainOrg of both
organizations is matched. The Cognitive Proximity (CognProx) was measured by counting how many of
the 72 rows of GICS industries matched. Then for the second research question, the “documents_export”
file was used to count the number of documents per project to assess the impact of that project.
Furthermore, the centrality measures were created in R and added to the dataset.
14
Table 1: the different variables that are used and their descriptions
Variable Meaning Range
sCity The city in which an organization is located -
sCountry The country in which the organization is -
located
MainOrg The main organization. If an organization has -
multiple locations the main organization is
the same. Each location gets a different name
and is considered a different organization.
stOrgtyp The type of organization according to the Categorical variable:
EUPRO dataset IND, ROR, EDU, GOV, NCL,
CON, OTH
Geoprox Variable for if two organizations are located TRUE/FALSE dummy
in the same city. variable
Orgprox Variable for if two organizations have the TRUE/FALSE dummy
same MainOrg variable
Insprox Variable for if two organizations have the TRUE/FALSE dummy
same stOrgtyp variable
CognProx The amount of similar GICS industries Count variable. 0-72
between two organizations.
sqCogn The square of this CognProx. 0-5184
emp The natural logarithm of the product of the 0-11.778
number of employees of two organizations
StartDate The date that the project started 2007-2017
ProjectEUfunding The amount of funding the EU provided for 0.05-10.811
the project (in million euros)
noParticipants The number of participants in a project 2-37
3.1.2 Dependent Variables

To analyse the network analysis, the dependent variable for RQ1 is the presence of a collaboration
between two organizations. In this paper, collaboration is measured by the participation in the same
project, which results in the presence of a link between the different organizations. This is measured by
creating a network analysis and seeing which actors work together and which do not. This is done by first
creating a 2-mode network of all projects in the circular economy in the Netherlands and which actors
worked in each project. Then, the 2-mode network is transformed into a 1-mode network of the actors
that worked together. For this, an undirected network analysis is used as collaboration is two-sided. As
very few organizations collaborated multiple times (only 6 out 303 collaborating pairs had more than 1
collaborations). The network was weighted. Furthermore, for this paper, collaborations with multiple
parties are more important than multiple collaborations with the same organizations. presence of a link is
a binary variable. Due to the fact that number of unique collaborations is more important than number of
total collaborations, repeated collaborations are counted as 1. While some organizations may have
multiple collaborations between them, these are counted as a binary variable. For this regression, a Binary
logit model will be used.
The dependent variable for RQ2 is the success of the projects from which the network analysis is created.
One way to measure this is the amount of achievements on the EUPRO page for each project. The data
for this can be found in the EUPRO dataset. This can be done by assessing the quality of the output and
15
grading this output as a continuous variable. However, such data are not available. Therefore, we proxied
the quality of a collaboration by the amount of documents a project delivered, as to indicate the success
of the project. If a project contains many documents discussing this project, the impact of this document
is considered large. These numbers will also be used as a continuous variable. To normalize the number
of documents as outputs for the inputs in a project, we include the amount of funding and the amount of
collaborators as control variables. By doing so, the success variable can be understood as an efficiency
measure.
3.1.3 Independent Variables

The first independent variable of RQ1 is the size of the organizations. This will be determined with the
number of employees. The number of employees are found either on the website of the organizations
itself, its LinkedIn introduction page, or on websites which register and monitor other organizations, such
as dnb, zoominfo, rocketreach, Apollo, etc. For each collaboration, the natural logarithm of the product
of the number of employees of both organizations. The second independent variable of RQ1 is the
proximity between the organizations in the Circular Economy. The four dimensions of proximity will be
measured according to Hardeman et al. (2014): The geographical proximity will be measured by whether
or not they are in the same city/municipality (high proximity), in the same province/county (medium
proximity) or not the same county (low proximity). The province/county area is considered based on
NUTS2 where applicable (for Luxembourg, the provinces were used and for Ukraine the oblasts were
used). The cognitive proximity is measured by the overlap between the amount of industries in which they
work. The organizations were sorted into industries according to GICS. In this case, for example an
electrical manufacturer which sells to other industries is not counted to the industries in which it sells, only
the industry to which the organization itself belongs and has knowledge bases of. This information is also
found on the websites of the organization itself. Furthermore, as government institutions (such as
municipalities) often do not belong to an industry, an extra sector was created specifically for governmental
institutions. Excluded from this are governmental institutions that provide utilities, as these clearly belong
to an industry. There was also a subject variable in the EUPRO dataset. This variable divided the projects
according to 68 keywords, similar the GICS categorization. The reasons these categories were not used
instead were that (1) the keywords were used to describe the project instead of the organizations which
was not useful for this analysis; (2) the variable was not present for each datapoint in the dataset and (3)
the designation of the keywords was not elaborated upon, which made reproducing it unreliable. The
institutional proximity will be a dummy variable, which states whether or not the organizations are in a
similar institutional sphere: academia, industry, government or care. This data is present in the EUPRO
dataset. Finally, organizational proximity is also measured as a dummy variable. This dummy variable
states whether or not two organizations belong to the same overarching hierarchical meta-structure, using
the MainOrg column that was created. For example, Arkema Ritthem and Arkema Rotterdam will be
considered separate organizations with an overarching hierarchical meta-structure and thus
organizationally proximate. The same goes for governmental institutions, as they fall under the federal
government of that country.
For RQ2, the first independent variables are four dimensions of proximity. The four dimensions will be
scored as they were in RQ1. The other independent variables for RQ2 is the network centrality. Network
centrality can be measured by the centrality measures. The centrality indicators from Wasserman and
Faust (1994) are used: degree centrality, brokerage centrality, betweenness centrality, and eigenvector
centrality. These centrality measures are calculated using functions in R in the package “igraph”. The
brokerage is calculated using the functions in the package “statnet”.
16
3.1.4 Control Variables
For RQ2, multiple control variables are used that may well affect the success of a project. The first control
variable is the start date of the project. The reason this variable needs to be controlled for is that a project
which has started earlier has had more time to gain articles, which is the success measure of this research.
For this variable, the “StartDate” variable from EUPRO will be used. The second control variable that is
used is the amount of funding the EU has contributed to the project. The reason for this is that success
may simply be attributed to the amount of money spent on the project. For this variable, the
“ProjectEUfunding” variable from EUPRO will be used. The reason this variable is used instead of
“TotalPojectCosts”, which is the total amount of money spent in total on the project, is that
TotalProjectCosts is sometimes 0, despite the fact that the ProjectEUfunding is higher than 0. This means
that sometimes the variable TotalProjectCosts is not properly measured. Finally, a control variable which
is used is the number of participants in the project. This will be used to ensure that the amount of
participants in an organization is not the reason for improved performance. For this variable, the
“noParticipants” variable from the EUPRO dataset is used.
3.2 Data analysis

To analyze the data, a network graph is created. This will be done using the R language and environment
(R Core Team, 2018), because R provides a wide range of quantitative analyses, which are needed for the
analysis. The network will contain information on the Dutch organizations that participated in EU funded
projects that were coordinated by the Netherlands in 2007-2017. From this network graph, network
measures are calculated. Furthermore, the organizations in the network that are most important are also
extracted. This is done based on degree and brokerage.
For the first two research questions regressions are carried out. The dataset for these regressions contains
every possible collaboration between organizations. This is done by using the R function “combn” to create
a list of all possible combinations. Then, a dataset containing the project information and organization
information is created by merging the “participations_export” and the “projects_export” files. These files
contain information on the organizations and projects, respectively. Here, the data that was collected,
regarding the number of employees and the market areas, was added. Then, the dataset with all
collaborations was joined with the file containing the information. This dataset was used for the
regressions.
For the first research question a regression will be performed to understand the relationship between the
independent variables and the dependent variable. As the dependent variable is binary, a binary logistic
regression is performed. First, a multivariate binary logistic regression will be applied to test the
relationships between the independent variables and binary dependent variable. Second, a correlation
matrix is created for the independent variables to observe any nested correlation that might occur. Third,
a Variance Inflated Factor (VIF) test is conducted to check for multicollinearity. VIF measures the increase
in variance of a regression due to the collinearity between the variables (Baldrich, 2019). For VIF, a cutoff
of 5 is often used (Sheather, 2009). If the VIF is higher than 5, the regression coefficients are poorly
estimated due to multicollinearity.
For the second research question, a regression will be performed to understand the relationship between
the independent variables and the dependent variable. As the dependent variable is a count of the number
17
of documents released, a count regression can be performed. To do this, the documents in the file
“documents_export” were counted for each RecCtrNr. If nothing was counted it was considered a 0. This
data was merged with the main dataset. Then, the dataset was filtered for all collaborations that actually
occurred, as a collaboration between two organizations that did not occur cannot be successful or
unsuccessful. First, a Poisson count model will be applied for each independent variable related to network
and the dependent variable. These network measures are too similar to compare together. This would
result in correlation between the independent variables, which would make the results less accurate. Thus,
the network centrality measures will be ran separately. Second, a Poisson count model will be applied to
test the relationships between the proximity independent variables and the dependent variable. After each
Poisson model, the overdispersion is tested. If the model is overdispersed, a Quasi Poisson model will be
applied. While a negative binomial also works well on overdispersed data, it tends to give more weight to
smaller organizations (Ver hoef and Boveng, 2007). This is not significant for this research. The reason
the VIF is not tested for the Quasi Poisson is that while for binary logistic regression the use of the VIF
has been documented (Midi et al., 2010), this has not been the case for Quasi Poisson regression.
However, as the proximity variables are the same for both research questions and the centrality measures
are tested with univariate regressions, the multicollinearity of the different proximity measures is
considered the same throughout the different regressions.
As the successful links between the organizations are not affected by the zeroes in the network, the non-
Dutch organizations can be included in research question 2. However, the centrality measures depend on
the full network being present, meaning only the proximity measures can be tested for all organizations.
This is done similarly to the regression with the proximity measures for only the Dutch organizations.
Finally, a missing link prediction based on path length is carried out. As this analysis depends on the
network and network measures, only Dutch organizations are examined for this research question. To
perform this analysis, decrease in average path length for newly added links needed to be measured. First,
a function to carry this out for each link in the whole network was attempted. However, due to the fact
that the network was unconnected, this was not possible. While trying to remove the unconnected nodes,
the network was not representable anymore. Thus, the missing link prediction was carried out manually.
This was done by making a list of all the possible combinations between the 25 organizations with the
highest degree (see table 12). This is done by adding the two organizations into the dataset and adding a
fake RecCtrNr (project number) in. This was done for all 300 combinations. Then, a network was created
from the new dataset and the average path length was measured. This is the average of the shortest paths
between each node.
18
4. Results
4.1 Network visualisation
The dataset that was created for all the Dutch organizations that participated in project with a Dutch project
coordinator, contained 139*138/2 = 9591 organization pairs. Table 2 shows the descriptive statistics for
all variables for research question 1, including the mean, minimum value and maximum value. These
rows contained all possible collaborations between the organizations. During data collection, data was not
available for all organizations. From these rows, data was available for 9548 organization pairs. From these
9548 possible collaborations, 303 collaborations occurred. Furthermore, in 850 possible collaborations
were both organizations from the same city (Geoprox = TRUE), organizations were institutionally
proximate in 5329 collaboration (Insprox = TRUE), organizations had organizational proximity in 17 of
the collaborations(Orgprox=TRUE), and the sum of the Cognitive Proximity was 9252 with an average
cognitive proximity of about 0.965. Finally, the average number of employees of the Dutch organizations
was 4171 (rounded down) with a median of 80 employees. The average number of employees of all
organizations was 9308 (rounded down) with a median of 252 employees.
As mentioned before, a network analysis was carried out of the organizations that participated in EU
funded projects. The projects were ranged from 2007 until 2017 and were coordinated by the Netherlands.
The network (see figure 1) contains the collaboration between the Dutch organizations for the projects
that have a Dutch coordinator. The vertices are sized according to their degree. The colours of the nodes
are based on the type of organization.
Table 2 descriptive statistics for the variables for research question 1

Variable Min 1st Quarter Median Mean 3rd Quarter Max
emp 0.000 2.813 4.119 4.165 5.431 10.142
CognProx 0.000 0.000 0.000 0.4151 1.000 19.000
sqCogn 0.000 0.000 0.000 1.223 1.000 361.000
Logical True False
Geoprox 347 9122
Orgprox 17 9452
Insprox 3407 6062
19
Figure 1: Network analysis of the collaboration between all Dutch organizations in EU funded projects
coordinated by the Netherlands. The projects are ranged from 2007 until 2017. The node size is linked to the
degree centrality of that node. The colours of the nodes are related to the type of organization.
The graph of the network of participating organizations in projects contained 139 nodes and 303 edges.
The network contains one hub around the nodes of “Universiteit Twente”, “Eindhoven University of
Technology”, and “Public Private Partnership Institute for Sustainable Process Technology”. This was
confirmed by creating a separate network graph with vertex sizes based on eigenvector (this network is
undirected meaning the hubscore and the eigenvector are equal) (see figure 4 in Appendix 1B). This
confirmed the group of nodes mentioned as a hub. The average degree of each node is 4.36, meaning
each organization collaborated with about four others organizations on average. Meanwhile the median
degree is 3. Furthermore, the average of the betweenness centrality was 133, while the median was 0. This
shows that the betweenness centrality of this network was very skewed. The average brokerage centrality
is 13.94, while the median is 0. This, along with the values in table 4, show that the brokerage centrality in
this network is skewed. The average path length of this network is 4.25. This is the average of the shortest
paths from each node to all the other nodes in the network. The network has a diameter of 9 and a
clustering coefficient of 0.58.
20
In this network, there are 14 articulation points (see table 3). An articulation point in a network is a node
whose removal disconnects the network (Tian et al., 2017). These organizations are important to the
connectivity of the network. Upon closer inspection of the 14 articulation points, a pattern can be observed
in the 14 organizations. The organizations are often either research or educational facilities (high
knowledge production) with over a 1,000 employees, present in many GICS industries or small industry
organizations (firms) with a low amount of employees (micro or small organizations) which are present in
one or two GICS industries (highly specialized). The industries of the organizations present in 1 or two
organizations were examined, but they were not the same nor were they in the same industry group.
Table 3: The organizations which form articulation points in the Network and the type of organization, number
of industries it is present in, the number of employees, and its location.
Organizations Org Number of Number of City County

type GICS employees
industries
Biodetection system BV IND 1 39 Amsterdam North Holland

ECN – Energy Research ROR 3 493 Petten North Holland
Centre of the Netherlands
Avantium Chemicals BV IND 2 220 Amsterdam North Holland
Wageningen UR EDU 5 6500 Wageningen Gelderland
TNO Den Haag ROR 19 2600 Den Haag South Holland
Delft University of EDU 22 3554 Delft South Holland
Technology
Vrije Universiteit Amsterdam EDU 9 4600 Amsterdam North Holland
Gemeente Amsterdam GOV 1 16000 Amsterdam North Holland
Van Gansewinkel IND 1 4125 Eindhoven North Brabant
Stichting Public Private OTH 7 39 Amersfoort Utrecht
Partnership Institute for
sustainable Process
Technology
Eindhoven Univeresity of EDU 13 3239 Eindhoven North Brabant
Technology
Provalor BV IND 2 1 Ursem North Holland
Rood Wit Blauw Water CON 1 70 Almelo Overijssel
Services BV
KWR Water BV ROR 2 170 Nieuwegein Utrecht
Furthermore, there are nodes which are brokers in the network. The nodes with the highest brokerage
centrality are the nodes that increase the number of organization pairs that are indirectly connected due
to presence of the node. The ten nodes with the highest brokerage centrality are shown in table 4. Here
the same pattern as with the articulation points was observed.
21
Table 4: The 10 organizations with the highest brokerage centrality in the Network and the type of organization,
number of industries it is present in, the number of employees, its location, and the brokerage centrality.
Org Number Number of City County Broker

Organizations
type GICS employees centrality
industries
Wolthuis Yvonne CON 1 1 Utrecht Utrecht 586
Barbara
Wageningen UR EDU 5 6500 Wageningen Gelderland 450
Vrije Universiteit EDU 9 4600 Amsterdam North Holland 340
Amsterdam
Vitens N.V. IND 1 1378 Zwolle Overijssel 314
Zwolle
Vitens N.V. IND 1 1378 Utrecht Utrecht 272
Utrecht
Vezet BV IND 4 561 Warmenhuizen North Holland 224
Boerenverstand CON 2 4 Utrecht Utrecht 140
Consultancy
Veolia IND 1 179000 Vroomshoop Overijssel 108
Vroomshoop
Vanderlande IND 1 6500 Veghel North Brabant 104
Industries BV
Van Remmen UV CON 1 11 Wijhe Overijssel 94
Techniek BV

In order to answer the first research question, the binary logit model regression was performed on the
organizations in the Netherlands. The independent variables that were tested were (1) size of the
organization, (2) the geographical proximity at the city level, (3) the organisational proximity, (4) the
Institutional proximity, (5) the cognitive proximity, and (6) the square of the cognitive proximity. The
results of the regression are shown in table 5.
The size of the organizations is significantly correlated with the presence of the link. This means that
Hypothesis 1 is accepted. The geographical proximity at the city level between two organizations is also
significantly associated with a presence of collaboration between these organizations. This means
Hypothesis 2a is accepted. The institutional proximity and the organizational proximity are not
significantly correlated with the presence of a collaboration. This means hypothesis 2b and 2c are not
accepted. The cognitive proximity is significantly correlated with the presence of a collaboration between
two organizations. The square of the cognitive proximity is also significant, but with a negative estimate.
This means that the effect of cognitive proximity on the chance of collaboration has an inverted u shape.
This means that Hypothesis 2d is accepted.
22
Table 5 Binary Logit Model with the presence of collaboration as the dependent variable and the four
dimensions of proximity and the organization size.
From the estimates of the CognProx and the sqCogn, the optimum Cognitive proximity for the presence
of collaboration can be calculated. This is done by calculating the point at which the estimates CognProx
and sqCogn are equal. The formula for this is:
Y= -0.030x2+0.473x
Where z is the variable “emp”. From this formula, the derivative is created. Then, the x is calculated for
where the derivative is 0. The value for is 7.88. This means, the optimal Cognitive proximity score is a
score of 7.88. Being present in 7-8 of the same industries as another organization will optimally increase
the chances of collaborating with that organization.
In order to ensure these results are not caused by multicollinearity between the independent variables, the
Variance Inflation Factor test was performed (see table 6). All variables have a value around 1 except the
CognProx and the sqCogn. This is to be expected as the sqCogn is derived from the CognProx. However,
to ensure that the multicollinearity of CognProx does not extend to the other independent variables,
another binary logit model is carried out, this time without the sqCogn (see Appendix 1C). Here all the
VIF values are around 1 and thus multicollinearity is not present between the different variables.
23
Table 6 Variance Inflation Factor test for the independent variables of the Binary logit model of table 5.
Finally, a correlation matrix was created for the correlation between the independent variables and the
dependent variable (see figure 2). From this figure, the independent variables which were significant in the
regression are also correlated with the dependent variable in the correlation matrix. The variable
CognProx is strongly correlated to the sqCogn. This is to be expected as the sqCogn is derived from the
CognProx. Furthermore, we see a correlation between the company size (emp) and CognProx, sqCogn,
and institutional proximity. As organizations are larger, they can have broader cognitive abilities, increasing
the chance of higher cognitive proximities. Organization size is negatively correlated with institutional
proximity. This means that larger organizations tend to work with a greater variety of types of organizations.
Most likely this is due to fact that large firms have the resources to collaborate with more organizations
and thus more types of organizations. Furthermore, we see a slight correlation between the Geographical
proximity at the city level and cognitive proximity.
24
Figure 2 Correlation matrix between the dependent variable and independent variables of research question 1

The data for this analysis are the 303 present links from the first research question. Here too, some of the
organizations did not have data for some of the measures. Thus, 293 observations are used in the
regression. Table 7 shows the descriptive statistics for all variables for research question 2.
Table 7 Descriptive statistics for all variables for research question 2

Success 0.000 0.000 2.000 2.567 2.000 48.000
CognProx 0.000 0.000 1.000 0.9898 1.000 14.000
sqCogn 0.000 0.000 1.000 4.137 1.000 196.000
Eigenvector 0.000 0.001 0.019 0.216 0.119 0.996
Brokerage 0.000 0.000 12.00 60.85 54.00 463.00
Betweenness 0.00 0.00 104.0 466.6 668.0 3026.3
Degree 1.00 5.00 7.00 8.22 11.00 24.00
ProjectEUfunding 0.595 1.310 3.158 3.651 4.716 10.811
StartDate 2008 2012 2013 2013 2015 2017
noParticipants 3.00 14.00 18.00 17.95 23.00 37.00
Logical True False
Geoprox 24 269
Orgprox 1 292
Insprox 97 196
25
4.3.1 Proximity measures
The proximity measures are taken as the independent variables. The dependent variable is the success of
the project. All models are Quasi Poisson count models. The results of the regression are shown in table
8. Here, the Geoprox has a negative estimate and is significant. This means that the Geographical
proximity between two organizations significantly decreases the chance of a successful project. This
confirms our Hypothesis 3a, less geographical proximity between organizations increases the chance of a
successful project. The Orgprox was not significantly correlated with the dependent variable. Hypothesis
3b cannot be accepted. The Insprox is also significant and has a negative estimate. This means our
hypothesis 3c can be accepted, less institutional proximity between organizations increases the chance of
a successful project. Finally, CognProx and sqCogn are not significant. This means hypothesis 3d cannot
be accepted on a national level.
The estimate of Geoprox for research question 1 was positive and significant. The estimate of Geoprox
in this regression is significantly negative. This asserts the presence of a proximity paradox for the
geographical proximity. The presence of geographical proximity stimulates organizations to collaborate,
while also negatively impacting the success of the project. The estimate of the institutional proximity for
research question 1 was positive. The significant negative estimate of Insprox indicates the presence of a
proximity paradox for the institutional proximity. However, for a proximity paradox to be accepted,
relation between the presence of a collaboration and thee institutional proximity should also have been
significant. As it is, however, the proximity paradox for the institutional proximity has not been proven.
Table 8 A Quasi Poisson count model with the dependent variable as the impact and the independent variable as
the four measures of proximity in column 1 and the four measures of proximity and the square of the cognitive
proximity in column 2.
26
4.3.2 Network Centrality measures
In order to answer the second research question, a count model regression was performed on the
organizations in the Netherlands. First, the network centrality measures are taken as the independent
variables. The centrality measures that are used are the (1) degree centrality, (2) betweenness centrality,
(3) brokerage centrality, and (4) the eigenvector centrality. The dependent variable is the Impact of the
project. The results of the regression are shown in table 9. The Poisson models of these variables were
performed (raw data in Appendix 1D). The dispersion test of all Poisson modes showed that the data was
overdispersed (see Appendix 1D). Thus, all models are Quasi Poisson count models.
Table 9 Quasi Poisson count models with different centrality measures as independent variables and the impact
as the dependent variable
The Quasi Poisson models for the four centrality measures are summarised in table 9. From this table,
we can see that the degree centrality of an organization is significantly correlated with the impact of the
project. This means hypothesis 4a can be accepted. The betweenness centrality of an organization is not
significantly correlated with the impact of the project. This means hypothesis 4b is not accepted. The
brokerage centrality of an organization is significantly correlated with the impact of the project. This means
hypothesis 4c can be accepted. The eigenvector centrality of an organization is significantly correlated with
the impact of the project. However, the estimate is negative. This means hypothesis 3
27
4d has been falsified. In fact, the opposite of the hypothesis seems to be true. Project with organizations
with higher eigenvector centrality are less likely to be successful. The estimates of the brokerage centrality
and degree centrality were also negative and while these results were not significant, this does point towards
a trend of well-connected organizations hindering the success of projects. From these results, it appears
that the projects with participants that are in hubs perform significantly worse than projects without
participants in hubs. This is the contrary to the results from Gloor et al. (2011), where innovators that were
closer to the main hub, were more successful.
Finally, a correlation matrix was created for the correlation between the independent variables for
hypotheses 3 and 4 and the dependent variable (see figure 3). From this figure, instantly a strong
correlation is seen between the different network centrality measures. The brokerage centrality,
betweenness centrality and degree centrality appear to have a positive correlation with CognProx and
sqCogn and a negative correlation with Insprox. This means that organizations that are brokers, have high
betweenness or high degree tend to have high cognitive proximity with organizations. One reason for this
could be that organizations in our network with high brokerage centralities, degree centralities and
betweenness centralities are often Research organizations (ROR) or Educational facilities (EDU), which
are large knowledge producers often with broad industrial presence compared to the rest of the network.
It also shows the same result as in figure 2, where larger organizations often have lower institutional
proximity and higher cognitive proximity than smaller organizations. Furthermore, like in figure 2 a strong
correlation between CognProx and sqCogn is present. Furthermore, the independent variables which were
significant in the regression are also correlated with the dependent variable in the correlation matrix.
Figure 3 Correlation matrix with the dependent variable and the independent variables for hypotheses 3 and 4
compared.
4.3.3 All organizations

Finally, in order to get a better look at the rest of the organizations participating in the projects, the
proximity measures were tested with the impact on all organizations participating in projects with a Dutch
coordinator. For this analysis, 5689 collaborations, of the 199,396 possible collaborations, were present
for projects for the Netherlands. From these 5689, data was available for 5668 collaborations (see table
10).
28
Table 10 descriptive statistics for the variables for the international dataset of research question 2
Success 0.000 0.000 2.000 3.718 2.000 48.000
CognProx 0.000 0.000 1.000 1.024 1.000 21.000
sqCogn 0.000 0.000 1.000 3.956 1.000 441.000
ProjectEUfunding 0.050 2.762 3.780 4.639 6.994 10.811
StartDate 2008 2010 2013 2013 2015 2017
noParticipants 2.00 14.00 18.00 19.81 26.00 37.00
Logical True False NA
Geoprox1 694 4995
Geoprox2 104 5585
Orgprox 13 5655 21
Insprox 1712 3977
Quasi Poisson regressions were run as they were in the previous analyses (see table 11). However, as this
data contains international organizations, another geographical proximity measure was used. Geoprox1 is
a TRUE/FALSE variable on whether or not two organizations in a collaboration are from the same
country. The variable previously known as Geoprox is now Geoprox2. In this regression, collaborations
between organizations that have Geographical proximity significantly decrease the chance of a project
being successful. This means hypotheses 3a can be accepted. The institutional proximity and the
organizational proximity are not significantly correlated with the dependent variable. This means
Hypothesis 3b and 3c cannot be accepted. The institutional proximity was significant with a negative
estimate for the organizations from the Netherlands, however this effect is not significant for the
organizations from all over Europe. Even if the Geographical proximity per country was present, the effect
was not significant. One reason this might be the case, is that not all collaborations are taken account for
non-Dutch organizations. Most of their collaborations will be in the country in which they are present.
Thus, this effect may be better observed if this effect is observed per country.
The sqCogn is significantly correlated with success, but has a negative estimate. This is in line with theory
of optimal proximity (hypothesis 3d). From the estimates of the CognProx and the sqCogn, the optimum
Cognitive proximity can be calculated. This is done by calculating the point at which the estimates
CognProx and sqCogn are equal. The formula for this case would be:
Y = -0.012x2+0.163x
Where x is the cognitive proximity, and a, b, and c are the control variables StartDate, ProjectEUfunding,
and noParticipants, respectively. From this formula, the derivative is created. Then, the x is calculated for
where the derivative is 0. The value for is 6.79. This means, the optimal Cognitive proximity score is a
score of 6.79. Having a collaboration with organizations that are present 6-7 of the same industries will
optimally increase the chances of the project being successful.
29
Table 11 A Quasi Poisson regression with the dependent variable, the independent variables, and the control
variables of hypothesis 4 for all firms participating in projects with a Dutch coordinator.
Regarding the proximity paradox, we can observe that the paradox holds for geographical proximity, both
in this regression and the regression in chapter 4.3.2. Organizations prefer to collaborate within the same
city, but the more successful projects tend to be the ones that collaborate over distance. Regarding
organizational and institutional proximity, we did not find this pattern in this regression. Finally, regarding
cognitive proximity, we computed to optimal cognitive proximity to collaborate (7-8) and to have the
highest probability for success (6-7). The difference in the values of the coefficients suggest that the
proximity paradox is also present for the cognitive proximity. A larger cognitive proximity is necessary to
stimulate collaboration, than to stimulate success. However, while there is evidence for the proximity
paradox in the cognitive dimension when comparing the levels of optimal proximity for the probability of
collaboration and for performance, it should be noted that the differences between the optimal levels are
quite small.

The 25 organizations that collaborated the most based on degree are shown in table 12. There is clearly
overlap between these firms and the articulation points (Table 3). These 25 organizations were combined
to create all possible combinations between the organizations. This resulted in 300 unique collaborations
30
(See Appendix 2). The best collaboration to add to the network based on the improvement of the average
path length is the collaboration between TNO Den Haag and Public Private Partnership Institute.
Through this collaboration, the average path length changed from 4.25 to 4.20. After that, any other
collaboration with TNO Den Haag only lowers the average path length to 4.239731. All the other
collaborations do not improve the average path length. Thus, generally TNO Den Haag can create
important links that improve the average path length.
Table 12 The 25 organizations with the most collaborations (based on degree)

Organizations
degree
Delft University of Technology 27
Wageningen UR 24
TNO - Netherlands Organization for Applied 21
Scientific Research Den Haag
Vrije Universiteit Amsterdam/VU University 12
Amsterdam
Universiteit Twente 12
Stichting Public Private Partnership Institute for 12
Sustainable Process Technology
Eindhoven University of Technology 12
Van Houtum Papier BV 10
Universiteit Leiden/ Leiden University 10
Smurfit Kappa Group 10
SAPPI Netherlands Services 10
Parenco BV 10
Mayr-Melnhof Eerbeek BV 10
KWR Water BV 10
Eska Graphic Board BV 10
ECN – Energy Research Centre of the Netherlands 10
DS SMITH Packaging Netherlands 10
Crown van Gelder Papierfabrieken 10
Rijksuniversiteit Groningen 8
Provalor BV 8
Nederlands Normalisatie-Instituut 8
Gemeente Amsterdam 7
BTG Biomass Technology Group BV 7
Avantium Chemicals BV 7
Vitens N.V. Zwolle 6
31
5. Conclusion
The aim of this paper was to study the collaboration in the circular economy in the Netherlands. To do
that, this paper focuses on the presence of collaborations and the impact of the collaboration. In this
paper, a social network was created in order to study the collaboration in the circular economy in the
Netherlands. The network was created to gather information about the collaborations within the circular
economy in the Netherlands. The resulting network had 139 participants and 303 links. This paper
examines the factors that drive collaboration in circular economy in the Netherlands. A binary logit model
regression was performed to analyse the factors that increase collaboration. The results show that high
Geographical and Cognitive proximity are driving factors to collaboration in the circular economy within
the Netherlands. The size of an organization is also a driver of collaboration between two organizations.
Moreover, this paper provides empirical evidence for the presence of the inverted U shape of probability
between the cognitive proximity and the likelihood to cooperate. The optimal score for the highest
likelihood of collaboration is a similarity in 7-8 industries.
The second research question examines the factors that drive the success of projects in the circular
economy. Quasi Poisson regressions were run to examine the effect of proximity and centrality measures
on the success of the project. The results show that collaborations with lower proximity drive the success
of projects, while organizations that are closer to hubs hinder project success. The negative impact of the
eigenvector centrality on success indicates that we may need to rethink the way we think about factors such
as network position and firm size. The results also show empirical evidence of the inverted U shape of
probability between cognitive proximity between two organizations and the success of the project. The
optimum for this function is a score of 6-7. Furthermore, within a country the institutional proximity
between organizations also negatively impacts the success of the projects. When looking at the
international level, this effect seems to disappear. Finally, the results from the first and second research
question show that a higher proximity is desired for a higher chance of collaborating, but undesirable for
the success of the project. This effect is also shown in the optimal scores for cognitive proximity in both
research questions, as the optimal score for collaboration is between 10 and 11 and the optimal score for
success is between 6 and 7. A higher cognitive proximity is required for collaboration than for the success
of the project. This provides empirical evidence for the proximity paradox. This paradox is also present
for the geographical proximity. Here a positive estimate is shown for collaboration between geographically
proximate organizations while a negative estimate is shown for the success of geographically proximate
organizations.
The final research question examines the missing links and important nodes in the circular economy in
the Netherlands. Here, the improvement of the average path length for 300 important collaborations were
tested. TNO - Netherlands Organization for Applied Scientific Research Den Haag is an important node
in the network. Collaborations between with this organization will result in an improvement of the average
path length and thus of the knowledge flow in the network. The most fruitful collaboration would be a
collaboration between TNO Den Haag and Stichting Public Private Partnership Institute for Sustainable
Process-Technology.
32
6. Discussion
The conclusions from this research include multiple implications for improving the performance of the
circular economy in the Netherlands. In the theoretical framework of this paper, it is theorized that while
organizations tend to collaborate with organization with high proximity measures, a lower proximity is
beneficial to the result of the collaboration. The theoretical implications of this paper is to confirm the
theory for geographical proximity and cognitive proximity. In the first regression (see table 5), it is shown
that a higher geographical proximity results in a higher chance of a collaboration, while in the regression
for the second research question (Tables 9 and 11), a higher geographical proximity has a negative effect
on the success of the project. For cognitive proximity, in research question 1 it is shown that the optimum
score for collaboration lies around 7-8, while for the success it lies around the 6-7.
Another theoretical implication of this research, is the rejection of the idea that for nodes with higher
resources, such as organizations with a higher eigenvector centrality, the performance of the collaboration
improves. The eigenvector significantly decreases the chance of a successful project. When looking at
existing theory, articles seem to imply that well-connected organizations are a positive factor for the success
of projects (Gloor et al., 2011). A possible reason for the different results in this research is that the success
was controlled for the amount of money spent on the project and the amount of participants in this paper.
This is not the case in the Gloor et al. (2011) article. This makes the measurement of success also a
measure of how efficiently success can be achieved in a project. Thus, the organizations with a great
network position may be less efficient at achieving success than organizations that are less well-positioned.
A possible reason for is are that the well-connected may not feel need to perform well in projects compared
to organizations that are only poorly connected. Another reason might be that they are often asked as a
collaboration partners regardless of their performance, but due to their reputation, size and/or resources.
Thus, this paper proposes that firms that have a higher centrality are less efficient for the success.
From the results, some policy recommendations can be derived. From the results of the second research
question, a negative correlation between the organizations closest to the hub and the success of the project.
If the government of a country wanted to stimulate the circular economy, it could stimulate collaborations
between organizations that are not close to the hubs in the network. Collaborations between organizations
that are not at the centre of the network stimulate the success of a project. From tables 3 and 4, it is shown
that mostly enterprises and micro firms are brokers and highly connected organizations. Thus, choosing
non-central organizations could be done by stimulating SMEs to participates in projects more than
enterprises or micro organizations. This way the success of projects can be achieved more efficiently. This
is especially the case for SMEs with a broad presence in different industries. Furthermore, according to
the results, organizations from different cities are less likely to collaborate. However, when they collaborate
the success of the project is more likely. In an international setting, collaboration between different
countries seems to improve the project performance best. Thus, increasing collaborations between
organizations from different countries is an important stimulant of project success and would improve the
performance of the circular economy in that country. Finally, cognitive proximity is shown to be subject
to the proximity paradox, where collaboration requires a higher proximity than success. Reducing this
effect could increase performance of the actors in the Circular Economy. One way to do this is to stimulate
collaborations between organizations from different sectors in upcoming government-funded projects.
This way, cognitive lock-in is less likely to occur.
For the Netherlands and Dutch organizations specifically, there are additional recommendations. First,
given the fact that geographical proximity negatively affects the chance for success but positively affects the
chance of collaboration, it would be beneficial for the success of the project for those collaborations to be
33
between organizations in different cities for the collaborations that happen between the Dutch
organizations. Furthermore, institutional proximity negatively affects the chance of success for
organizations within the Netherlands. Thus, it is recommended to stimulate collaborations between
different types of organizations, so with a high institutional distance. Having a higher diversity will stimulate
the success of the projects and according to (van Rijnsoever et al., 2015) will stimulate the technological
diversity, which is important for a technological system. Finally, in order to drive the circular economy in
the Netherlands, it is recommended for the flow of information to stimulate TNO Den Haag to
collaborate with different organizations. This organization is both an articulation point and the only
organization with which the average path length seems to decrease when other large nodes collaborate
with it. A good collaboration would be with Stichting Public Private Partnership Institute, as this showed
a great network improvement.
As this study is a deductive one, the research method uses a positivist approach. A positivist approach
holds that society operates according to general laws and that these can be detected by the researcher.
Thus, to check the validity of the research in this paper, the validity measures for positivist research will
be used (Herrmann & Vaskelainen, 2018). The first validity measure is the construct validity. This measure
is to check that the correct operational measures are used to study the concepts in this paper. The way of
measuring the concepts in this paper was taken from other peer-reviewed articles. This way the validity of
the operationalization was ensured. The second validity measure is the internal validity is the internal
validity. Internal validity is gained when a causal relationship is established between the variables. The
causal relationship between the independent variables and the dependent variable have been backed up
by other papers in the theory. Third is the external validity. External validity is used to define the domain
to which this paper can be generalized. The external validity of this research is quite high because the
variables used in this paper are not specific to a country. Furthermore, in the second research question
the proximity analysis was performed over international organizations, meaning these results are indicative
of the circular economy in other countries as well. This means that while some of the recommendations
(especially from research question 3) were specific to the Netherlands, the rest were general to other
countries as well. Finally, the reliability of the paper, which refers to the fact that the operations of a study—
such as the data collection procedures—can be repeated, with the same results. The method has been
clearly explained, all new data that has been added has been retrieved from a public source, and the source
for each datapoint is present in the dataset. Thus, the replicability of this paper is high as well.
The main limitation of this research is the impact measure from the second research question. This is
based on the number of articles of each project in the project database. As it is unclear how representative
these data are of the impact of the project, adding another performance indicator would have made the
research more reliable. However, in the EUPRO dataset there did not seem to be another performance
indicator that could be used on all or even most projects. This, due to the difference in the project output.
Some projects had scientific article(s) as output, while others had multiple documents, and others again
had pilots, etc. This made it difficult to create another performance indicator. Another limitation is the
fact that the link prediction could not be carried out according to e.g. the Katz algorithm, due to the
disconnectedness of the network. If more time had been present, the unconnected nodes could have been
removed manually and a link prediction could have been carried out. This could have made the results
from the link prediction, more reliable and complete. Finally, a limitation of this paper is that the data for
all of Europe could not be examined instead of only in the Netherlands. This would have given a more
reliable insight of the effects of proximity on an international level. For this paper, this insight remains
limited.
34
In the results, it shows that working with international organizations improves performance in the circular
economy for the Netherlands. This is interesting due to the fact that circular economy is mostly looked at
as a local system. The influence of organizations in a circular economy interacting on an international level
would be an interesting direction for future research. Another interesting direction to study would be the
effect of network position and size on the efficiency to produce impactful outcomes. Current research
correlates networks position and size of an organization to higher success. Thus, more research into why
the network position of the organizations can lead to lower performance would be interesting. In the light
of proximity theory, it would also be interesting to extent this research by distinguishing between different
innovation stages (Research, Development and Marketing). It has been argued that different stages in the
innovation process require different types of proximity (Davids & Frenken, 2018). Based on the type of
proximity that are present in a case study, the innovation stage can be identified and a transition to the
next stage can be stimulated. Therefore, it would be interesting for future research if the proximities can
be used to identify the stage or change of innovation stage of the circular economy (or any system).
Acknowledgements
I would like to thank first and foremost my supervisor, Dr. Frenken, for his continuous support and
guidance. Furthermore, I would like to thank Fatjeta Salihu for providing aid with coding difficulties.
Finally, I would like to thank Dr. Thomas Scherngell (Austrian Institute of Technology) for supporting
me in getting access to the EUPRO database via the RISIS infrastructure.
35
References
Achterberg, E., Hinfelaar, J., & Bocken, N. (2016). Master Circular Business with the Value Hill, 1–16.
Ackerman, F. (2000). Waste Management and Climate Change. Local Environment, 5(2), 223-229.
Acuna, D. E., Allesina, S., & Kording, K. P. (2012). Predicting scientific success. Nature, 489(7415),
201-202.
Ahuja, G. (2000). Collaboration networks, structural holes, and innovation: A longitudinal study.
Administrative science quarterly, 45(3), 425-455.
Amin, A., & Wilkinson, F. (1999). Learning, proximity and industrial performance: an introduction .
Cambridge Journal of Economics, 121-125.
Ashton, W. (2008). Understanding the organization of industrial ecosystems. Journal of Industrial
Ecology, 12(1), 34-51.
Ashton, W. (2009). The structure, function, and evolution of a regional industrial ecosystem. Journal of
Industrial Ecology, 13(2), 228-246.
Baldrich, J. F. (2019). Regression Diagnostics.
Biggiero, L., & Basevi, M. (2009). Testing the gravity model through network analysis. In European
Trade Study Group Conference’, Rome , 10-12.
Boeing, G. (2016). Visual Analysis of Nonlinear Dynamical Systems: Chaos, Fractals, Self-Similarity and
the limits of Predicition. systems, 4(4), 37.
Boschma, R. (2005). Proximity and Innovation: A Critical Assessment. Regional Studies, 39(1), 61-74.
Boschma, R. A., Frenken, K. (2010) The spatial evolution of innovation networks. A proximity
perspective. In R. A. Boschma, R. Martin (eds), The Handbook of Evolutionary Economic
Geography, pp. 120–135. Cheltenham: Edward Elgar
Bozeman, B., & Youtie, J. (2016). Trouble in paradise: Problems in academic research co-authoring. .
Science and engineering ethics, 22(6), , 1717-1743.
Broekel, T., & Boschma, R. (2012). Knowledge networks in the Dutch aviation industry: the proximity
paradox. Journal of Economic Geography, 12(2), 409-433.
Burt, R. (1992). Structural Holes: The Social Structure of Competition. Cambridge, MA: Harvard
University Press.
Burt, S. (2004). Structural Holes and Good Ideas. American Journal of Sociology, 110(2), 349-399.
Camisón-Zornoza, C., Lapiedra-Alcamí, R., Segarra-Ciprés, M., & Boronat-Navarro, M. (2004). A meta-
analysis of innovation and organizational size. Organization studies, 25(3), 331-361.
Circle Economy. (2019). The Circularity Gap report. Retrieved from Circle Economy: https://bfc732f7-
80e9-4ba1-b429-
7f76cf51627b.filesusr.com/ugd/ad6e59_ba1e4d16c64f44fa94fbd8708eae8e34.pdf
Cohen, W. M., & Levinthal, D. A. (1990). Absorptive capacity: A new perspective on learning and
innovation. Administrative science quarterly, 128-152.
Cohen, W.M., R.R. Nelson and J. Walsh. 2002. Links and Impacts: The Influence of Public Research
on Industrial R&D. Management Science. Vol. 48: 1-23.
36
Coleman, J. (1988). "Social capital in the creation of human capital.". American Journal of Sociology, 94,
95-120.
Davids, M., & Frenken, K. (2018). Proximity, knowledge base and the innovation process: towards an
integrated framework. Regional Studies, 52(1), 23-34.
De Wit, A. (1988). Measurement of project success. International journal of project management, 6(3),
164-170.
Dhanaraj, C., & Beamish, P. (2003). A resource-based approach to the study of export performance.
Journal of Small Business Management, 41, 242-261.
Don, R. (2019). Game theory. Retrieved from The Stanford Encyclopedia of Philosophy (Winter 2019
Edition), Edward N. Zalta (ed.): https://plato.stanford.edu/archives/win2019/entries/game-
theory/
Dora, M. (2019). Collaboration in a circular economy: Learning from the farmers to reduce food waste.
EC. (2014). MEMO, Questions and Answers on the Commission Communication “Towards a Circular
Economy” and the Waste Targets Review. European Commission.
EC. (2014). Towards a Circular Economy: a Zero Waste Programme for Europe. Communication from
the Commission to the European Parliament, the Council, the European Economic and Social
Committee and the Committee of the Regions. (p. 398). European Commission.
Ellen McArthur Foundation. (2012). Towards the Circular Economy. Retrieved from Ellen McArthur
Foundation: http://www.ellenmacarthurfoundation.org/business/reports
European Environment Agency. (2017, Jan 25). Climate change poses increasingly severe risks for
ecosystems, human health and the economy in Europe. Retrieved from European Environment
Agency: https://www.eea.europa.eu/highlights/climate-change-poses-increasingly-severe
Eurostat. (2018). Patents related to recycling and secondary raw materials (cei_cie020). Retrieved from
Eurostat: https://ec.europa.eu/eurostat/cache/metadata/en/cei_cie020_esmsip2.htm
Eurostat. (2019). Renewable energy statistics. Retrieved from Eurostat statistics explained:
https://ec.europa.eu/eurostat/statistics-explained/index.php/Renewable_energy_statistics
Faber, J., & Frenken, K. (2009). Models in evolutionary economics and environmental policy: towards
an evolutionary environmental economics. Technological Forecasting and Social Change, 76,
462-470.
Feng, Z., & Yan, N. (2007). Putting a circular economy into practice in China. Sustainable Science, 2,
95-101.
Fleming, L., King, C. III, Juda, A. I. (2007) Small worlds and regional innovation. Organization Science,
18: 938–954.
Frenken, K., Hardeman, S., & Hoekman, J. (2009). Spatial scientometrics: Towards a cumulative
research program. Jouranl of Infometrics, 3, 222-232.
Geissdoerfer, M., Savaget, P., Bocken, N. M., & Hultink, E. J. (2017). The Circular Economy–A new
sustainability paradigm?. Journal of cleaner production, 143, 757-768.
Geng, Y., Fu, J., Sarkis, J., & Xue, B. (2012). Towards a circular economy indicator system in China: an
evaluation and critical analysis. Journal Cleaner Production, 23, 216-224.
Geuna, A., Fontana, R., & Matt, M. (2003). Firm size and openness: the driving forces of university-
industry collaboration.
37
Ghisellini, P., Cialani, C., & Ulgiati, S. (2016). A review on circular economy: the expected transition to
a balanced interplay of environmental and economic systems. Journal of Cleaner Production,
114, 11-32.
Gloor, P., Dorsaz, P., & Fuehres, H. (2011). Analyzing success of startup entrepreneurs by measuring
their social network distance to a business networking hub. In Proceedings 3rd international
conference on collaborative innovation networks coins, Basel. Sept (pp. 8-10).
Golbeck, J. (2013). Network structures and measures. Newnes.
Hardeman, S., Frenken, K., Nomaler, Z., & ter Wal, A. (2014). Characterizing and comparing
innovation systems by different 'modes' of knowledge production : a proximity approach.
Science and Public Policy, 42(4), 530-548.
Herrmann, A., & Vaskelainen, T. (2018). Session 1: Introduction and Fundamentals of Research
design. Retrieved from Blackboard:
https://uu.blackboard.com/webapps/blackboard/execute/content/file?cmd=view&content_id=_3
063240_1&course_id=_116610_1
Heller-Schuh, B., Barber, M., Züger, M., & Scherngell, T. (2019) Report on the content and technical
structure of the EUPRO infrastructure. Vienna: Australian Institute of Technology. Retrieved
from: RISIS
Homrich, A. S., Galvao, G., Abadia, L. G., & Carvalho, M. M. (2018). The circular economy umbrella:
Trends and gaps on integrating pathways. Journal of Cleaner Production, 175, 525-543.
Hughes, R. (2017). The EU Circular Economy package – life cycle thinking to life cycle law? Procedia
CIRP, 61, 10-16.
Ji, L., Liu, C., Huang, L., & Huang, G. (2018). The evolution of resources conservation and recycling
over the past 30 years: a bibliometric overview. Resources, Conservation and Recycling, 134, 34-
43.
Katz, L. (1953). A new status index derived from sociometric analysis. Psychometrika, 18(1), 39-43.
Kenniskaarten. (2016). Circular economy. Retrieved from Het groene brein:
https://kenniskaarten.hetgroenebrein.nl/en/knowledge-map-circular-economy/how-is-a-circular-
economy-different-from-a-linear-economy/
Kepaptsoglou, K., Karlaftis, M., & Tsamboulas, D. (2010). The Gravity Model Specification for
Modeling International Trade Flows and Free Trade Agreement Effects: A 10-Year Review of
Empirical Studies. The Open Economics Journal, 3, 1-13.
Kirchherr, J., Reike, D., & Hekkert, M. (2017). Conceptualizing the circular economy: An analysis of
114 definitions. Resources, Conservation and Recycling, 221-232.
Lett, L. (2014). Las amenazas globales, el reciclaje de residuos y el concepto de economia circular. Riv.
Argent. Microbiol., 46(1), 1-2.
Lou, X., & Nair, J. (2009). The impact of landfilling and composting on greenhouse gas emissions – A
review. Bioresource Technology, 100(16), 3792-3798.
Luo, P., Li, Y., Wu, C., & Chen, K. (2016). Detecting the missing links in social networks based on
utility analysis. Journal of computational science, 16, 51-58.
Mazzatini, U. (2014). Green report. Retrieved from Rivoluzione a Davos, il big business mondiale vuole
l'economia circolare: http://www.greenreport.it/news/consumi/rivoluzione-a-davos-il-big-
business-mondiale-vuole-leconomia-circolare/
38
Mesa, J. A., Esparragoza, I., & Maury, H. (2019). Trends and Perspectives of Sustainable Product
Design for Open Architecture Products: Facing the Circular Economy Model. International
Journal of Precision Engineering and Manufacturing-Green Technology, 6(2), 377-391.
Midi, H., Sarkar, S. K., & Rana, S. (2010). Collinearity diagnostics of binary logistic regression
model. Journal of Interdisciplinary Mathematics, 13(3), 253-267.
Mowery, D. C., Oxley, J. E., Silverman, B. S. (1998) Technological overlap and interfirm cooperation:
implications for the resource-based view of the firm. Research Policy, 27: 507–523
Nieto, M. J., & Santamaría, L. (2007). The importance of diverse collaborative networks for the
novelty of product innovation. Technovation, 27(6-7), 367-377.
Nomaler, Ö., Frenken, K., Giuliani, E., & Arora, S. (2014, September 23). “In The Loop”: On cycles in
knowledge networks.
Park, J., & Chertow, M. (2014). Establishing and testing the “reuse potential” indicator for managing
waste as resources. J. Environ. Manag., 137, 45-53.
Pla-Barber, J., & Alegre, J. (2007). Analysing the link between export intensity, innovation and firm size
in a science-based industry. International Business Review, 16(3), 275-293.
Preston, F. (2012). A Global Redesign? Shaping the Circular Economy. Retrieved from Chatham
House:
https://www.chathamhouse.org/sites/default/files/public/Research/Energy%2C%20Environment
%20and%20Development/bp0312_preston.pdf
R Core Team (2018). R: A language and environment for statistical computing. R Foundation for
Statistical Computing, Vienna, Austria. URL https://www.R-project.org/.
Rajeev, A., Rupesh, P. K., Padhia, S. S., & Govindan, K. (2017). Evolution of sustainability in supply
chain management: A literature review. Journal of Cleaner Production, 162, 299-314.
Ripanti, E. F., & Tjahjono, B. (2019). Unveiling the potentials of circular economy values in logistics
and supply chain management. The International Journal of Logistics Management.
Rood, T., & Hanemaaijer, A. (2017). Waarom een circulaire economie? Retrieved from Planbureau
voor de leefomgeving: https://themasites.pbl.nl/circulaire-economie/
Ruutu, S., Casey, T., & Kotovirta, V. (2017). Development and competition of digital service platforms:
A system dynamics approach. Technological Forecasting and Social Change, 117, 119-130.
Schilling, M., & Phelps, C. (2007). Interfirm Collaboration Networks: The Impact of Large-Scale
Network Structure on Firm Innovation. Management Science, 53(7), 1113–1126.
Sheather, S. (2009). A modern approach to regression with R. Springer Science & Business Media.
Silva, A., Stocker, L., Mercieca, P., & Rosano, M. (2016). The role of policy labels, keywords and
framing in transitioning waste policy. Journal of cleaner production, 115, 224-237.
Stiehl, C., & Hirth, T. (2012). Vom additiven Umweltschutz zur nachhaltigen Produktion. Chemie
Ingenieur Technik, 84, 963-968.
Su, B., Heshmati, A., Geng, Y., & Yu, X. (2013). A review of the circular economy in China: moving
from rethoric to implementation. Journal of Cleaner Production, 42, 215-227.
Tian, L., Bashan, A., Shi, D. N., & Liu, Y. Y. (2017). Articulation points in complex networks. Nature
communications, 8(1), 1-9.
39
TNO. (2019, December 13). THE CIRCULAR ECONOMY AND THE ENVIRONMENT:
'DIRECTING AND ACCELERATING SUSTAINABILITY'. Retrieved from TNO:
https://www.tno.nl/en/focus-areas/circular-economy-environment/
UNEP. (2013). Resource Efficiency: Economics and Outlook for China. Retrieved from United Nations
Environment Programme:
http://www.unep.org/pdf/China_Resource_Efficiency_in_English_2013.pdf
van Rijnsoever, F. J., van den Berg, J., Koch, J., & Hekkert, M. P. (2015). Smart innovation policy: How
network position and project composition affect the diversity of an emerging technology.
Research Policy, 44(5), 1094-1107.
Ver Hoef, J. M., & Boveng, P. L. (2007). Quasi‐Poisson vs. negative binomial regression: how should
we model overdispersed count data?. Ecology, 88(11), 2766-2772.
Walker, G., Kogut, B., & Shan, W. (1997). "Social capital, structural holes and the formation of an
industry network." . Organization Science, 8, 109-125.
Waughray, D. (2013). Davos 2013: Circular Economy Offers Opportunities for Latin America.
Retrieved from The Guardian: http://www.theguardian.com/sustainable-business/davos-2013-
circular-economy-opportunities-latin-america
WE Forum. (2014). Towards the Circular Economy: Accelerating the scale-up across global supply
chains. Geneva, Switzerland: World Economic Forum.
Yap, N. (2005). Towards a circular economy: progress and challenges. Greener Managment
International, 50, 11-24.
Yuan, Z., Bi, J., & Moriguichi, X. (2006). The circular economy; a new development strategy in China.
Journal of Industrial Ecology, 10, 4-8.
40
Appendix 1
Appendix A
The keywords used to select the projects related to circular economy and their source:
patent codes cpc, Eurostat 2018
circular economy
recycling
waste collection
waste transportation
waste separation
waste processing
waste management
recovery of materials
reuse technologies
waste reduction
fuel reprocessing
wastewater treatment
Achterberg, E., Hinfelaar, J., & Bocken, N. (2016). Master Circular Business with the Value Hill, 1–16.
“circular product design”

“circular materials”
“classic long life”
“product as a service”
“product life extension”
“sharing platforms”
“sharing economy”
“sell and buy back”
“repair and maintenance”
“recaptured material”
Homrich, A. S., Galvao, G., Abadia, L. G., & Carvalho, M. M. (2018). The circular economy umbrella:
Trends and gaps on integrating pathways. Journal of Cleaner Production, 175, 525-543.
“cradle-to-cradle”
“industrial ecology”
“biomimicry”
“performance economy”
“regenerative design”
Ripanti, E. F., & Tjahjono, B. (2019). Unveiling the potentials of circular economy values in logistics
and supply chain management. The International Journal of Logistics Management.
“waste elimination”
“leakage minimisation”
41
Ji, L., Liu, C., Huang, L., & Huang, G. (2018). The evolution of resources conservation and recycling
over the past 30 years: a bibliometric overview. Resources, Conservation and Recycling, 134, 34-43.
“life cycle assessment”

“material flow analysis”
“municipal solid waste”
“waste of electric and electronic equipment”
“wastewater”
Mesa, J. A., Esparragoza, I., & Maury, H. (2019). Trends and Perspectives of Sustainable Product
Design for Open Architecture Products: Facing the Circular Economy Model. International Journal of
Precision Engineering and Manufacturing-Green Technology, 6(2), 377-391.
“optimization of end-of-life”
“resource efficiency”
“multiple lifecycle generation”
“reduction of unwanted emissions”
“extension of operational life”
“use of low impact materials”
Silva, A., Stocker, L., Mercieca, P., & Rosano, M. (2016). The role of policy labels, keywords and
framing in transitioning waste policy. Journal of cleaner production, 115, 224-237.
“zero waste”
“sustainable material management”
Geissdoerfer, M., Savaget, P., Bocken, N. M., & Hultink, E. J. (2017). The Circular Economy–A new
sustainability paradigm?. Journal of cleaner production, 143, 757-768.
“Industrial symbiosis”
“industrial ecology”
“eco-industrial park”
42
Table 13: CPC codes related to Circular Economy (according to Eurostat, 2018)
Y02W 10/00 - Technologies for wastewater treatment
• Y02W 10/10 - Biological treatment of water, wastewater, or sewage
o Y02W 10/12 - Anaerobic processes with biogas recycling, capture or flaring
o Y02W 10/15 - Aerobic processes
o Y02W 10/18 - Constructed wetlands
• Y02W 10/20 - Sludge processing
o Y02W 10/23 - Anaerobic processes with biogas recycling, capture or flaring
o Y02W 10/27 - Aerobic processes
• Y02W 10/40 - Valorisation of by-products of wastewater, sewage or sludge processing
o Y02W 10/45 - Obtention of biopolymers
Y02W 30/00 - Technologies for solid waste management

• Y02W 30/10 - Related to waste collection, transportation, transferor storage, e.g. segregated
refuse collecting, electric or hybrid propulsion
• Y02W 30/20 - Related to waste processing or separation
• Y02W 30/40 - Bio-organic fraction processing; Production of fertilisers from the organic
fraction of waste or refuse
o Y02W 30/43 - Aerobic fermentation, e.g. composting
o Y02W 30/47 - Anaerobic fermentation, e.g. methanation combined with capture,
recycling or flaring
• Y02W 30/50 - Reuse, recycling or recovery technologies
o Y02W 30/52 - Dismantling or mechanical processing of waste for the recovery of
materials during separation, disassembly, pre-processing or upgrading
o Y02W 30/54 - Metal recycling
o Y02W 30/56 - Disassembly of vehicles for recovery of salvageable parts
o Y02W 30/58 - Construction or demolition waste
o Y02W 30/60 - Glass recycling
o Y02W 30/62 - Plastics recycling
o Y02W 30/64 - Paper recycling
o Y02W 30/66 - Disintegrating fibre-containing textile articles to obtain fibres for re-
use
o Y02W 30/68 - Rubber waste recycling
o Y02W 30/70 - Recovery of polymers other than plastics or rubbers
o Y02W 30/72 - Recovery of luminescent materials
o Y02W 30/74 - Recovery of fats, fatty oils, fatty acids or other fatty substances, e.g.
lanolin or waxes
o Y02W 30/76 - Recovery of tanning agents from leather
o Y02W 30/78 - Recycling of wood or furniture waste
o Y02W 30/80 - Packaging reuse or recycling
o Y02W 30/82 - Recycling of waste of electrical or electronic equipment
o Y02W 30/84 - Recycling of batteries
o Y02W 30/86 - Recycling of fuel cells
o Y02W 30/88 - Nuclear fuel reprocessing
o Y02W 30/90 - Reuse, recycling or recovery technologies crosscutting to different
types of waste
Y02W 90/00 - Enabling technologies or technologies with a potential or indirect contribution to

greenhouse gas emissions mitigation
• Y02W 90/20 - Computer systems or methods specially adapted for waste reduction or
recycling of materials or goods
43
Appendix B
Figure 4: Network analysis of the collaboration between all Dutch organisations in EU funded projects
coordinated by the Netherlands. The projects are ranged from 2007 until 2017. The node size is linked to the
hubscore of that node. The colors of the nodes are related to the type of organisation.
44
Appendix C
Table 14: Binary logit model with the dimensions of proximity and the size of the organisations as independent
variable and the presence of a link as the dependent variable
Table 15: The Variance Inflation Factor (VIF) of the binary logit model of table 1
45
Appendix D
All poisson regressions for the success of a project.
Control Variables:
Start date of the project (StartDate)
The amount of funding the EU provided (ProjectEUfunding)
The number of participants (noParticipants)
Independent variable: Eigenvector centrality
ptest1<- glm(n.x~eigenvec+ StartDate.x + noParticipants.x + ProjectEUfundin

g.x, family = "poisson", data = df_Succes)
> summary(ptest1)
Call:
glm(formula = n.x ~ eigenvec + StartDate.x + noParticipants.x +
ProjectEUfunding.x, family = "poisson", data = df_Succes)
Deviance Residuals:
Min 1Q Median 3Q Max
-4.2630 -1.4132 -0.8786 -0.3495 10.2463
Coefficients:
Estimate Std. Error z value Pr(>|z|)
(Intercept) 9.075e+00 6.836e-01 13.275 < 2e-16 ***
eigenvec -1.691e+00 2.713e-01 -6.231 4.64e-10 ***
StartDate.x -6.712e-09 5.600e-10 -11.987 < 2e-16 ***
noParticipants.x 4.054e-02 5.255e-03 7.714 1.22e-14 ***
ProjectEUfunding.x 9.066e-08 1.908e-08 4.752 2.01e-06 ***
---
Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1
(Dispersion parameter for poisson family taken to be 1)

Null deviance: 1879.8 on 292 degrees of freedom
Residual deviance: 1414.6 on 288 degrees of freedom
AIC: 2029.6
Number of Fisher Scoring iterations: 7

> dispersiontest(ptest1)
Overdispersion test
data: ptest1
z = 3.4688, p-value = 0.0002614
alternative hypothesis: true dispersion is greater than 1
sample estimates:
dispersion
9.38744
Independent variable: betweenness centrality
ptesty2<- glm(n.x~between+ StartDate.x + noParticipants.x + ProjectEUfundin

g.x, family = "poisson", data = df_Succes)
> summary(ptesty2)
Call:
glm(formula = n.x ~ between + StartDate.x + noParticipants.x +
Deviance Residuals:
-4.4833 -1.3650 -1.2684 0.0179 10.5672
46
Coefficients:
(Intercept) 1.228e+01 6.590e-01 18.640 < 2e-16 ***
between -2.080e-04 6.038e-05 -3.445 0.000571 ***
StartDate.x -9.274e-09 5.189e-10 -17.873 < 2e-16 ***
ProjectEUfunding.x 1.597e-07 1.647e-08 9.695 < 2e-16 ***
---
Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1
AIC: 2068.3

> dispersiontest(ptesty2)
Overdispersion test
data: ptesty2
z = 3.6509, p-value = 0.0001306
sample estimates:
dispersion
9.105404
Independent variable: brokerage centrality
ptest3<- glm(n.x~brok+ StartDate.x + noParticipants.x + ProjectEUfunding.x,

family = "poisson", data = df_Succes)
> summary(ptest3)
Call:
glm(formula = n.x ~ brok + StartDate.x + noParticipants.x + ProjectEUfundin
g.x,
family = "poisson", data = df_Succes)
Deviance Residuals:
-4.6098 -1.3448 -1.0749 0.0607 10.8714
Coefficients:
(Intercept) 1.365e+01 6.666e-01 20.472 < 2e-16 ***
brok -3.048e-03 4.019e-04 -7.584 3.34e-14 ***
StartDate.x -1.022e-08 5.267e-10 -19.397 < 2e-16 ***
---
Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1

AIC: 2016

Overdispersion test
47
data: ptest3
z = 3.5781, p-value = 0.000173
sample estimates:
dispersion
8.507991
Independent variable: degree centrality
ptesty5<- glm(n.x~degree+ StartDate.x + noParticipants.x + ProjectEUfunding

.x, family = "poisson", data = df_Succes)
> summary(ptesty5)
Call:
glm(formula = n.x ~ degree + StartDate.x + noParticipants.x +
Deviance Residuals:
-4.5897 -1.3426 -1.0396 -0.0512 11.1700
Coefficients:
(Intercept) 1.315e+01 6.679e-01 19.686 < 2e-16 ***
degree -6.119e-02 8.998e-03 -6.801 1.04e-11 ***
StartDate.x -9.593e-09 5.187e-10 -18.494 < 2e-16 ***
---
Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1

AIC: 2031.6
Overdispersion test
data: ptesty5
z = 3.4426, p-value = 0.0002881
sample estimates:
dispersion
8.997904
Independent variables:
Square of the cognitive proximity
Cognitive proximity
Institutional proximity
Organisational proximity
Geographical proximity
ptesty7<- glm(n.x~sqCogn+ CognProx+ Insprox +Orgprox+ Gprox3+ StartDate.x+

ProjectEUfunding.x+ noParticipants.x, family = "poisson", data = df_Succes)
> summary(ptesty7)
Call:
glm(formula = n.x ~ sqCogn + CognProx + Insprox + Orgprox + Gprox3 +
48
StartDate.x + ProjectEUfunding.x + noParticipants.x, family = "poisson"
,
data = df_Succes)
Deviance Residuals:
-4.8786 -1.4778 -1.0783 0.1493 9.6929
Coefficients:
(Intercept) 1.105e+01 6.476e-01 17.064 < 2e-16 ***
sqCogn -6.826e-03 3.702e-03 -1.844 0.065227 .
CognProx 1.248e-01 3.921e-02 3.183 0.001456 **
InsproxTRUE -2.818e-01 8.345e-02 -3.377 0.000732 ***
OrgproxTRUE -5.632e-01 1.003e+00 -0.561 0.574576
Gprox3TRUE -1.050e+00 1.738e-01 -6.041 1.53e-09 ***
StartDate.x -8.497e-09 5.201e-10 -16.338 < 2e-16 ***
---
Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1

AIC: 2013.1

Overdispersion test
data: ptesty7
z = 3.6844, p-value = 0.0001146
sample estimates:
dispersion
8.123143
Independent variables (international organisations included):

Square of the cognitive proximity
Cognitive proximity
Institutional proximity
Organisational proximity
Geographical proximity
ptest7<- glm(n.x~sqCogn+ CognProx+ Insprox +Orgprox+ Gprox3+ StartDate.x+ P

rojectEUfunding.x+ noParticipants.x, family = "poisson", data = df_Succes)
> summary(ptest7)
Call:
glm(formula = n.x ~ sqCogn + CognProx + Insprox + Orgprox + Gprox3 +
StartDate.x + ProjectEUfunding.x + noParticipants.x, family = "poisson"
,
data = df_Succes)
Deviance Residuals:
-8.6588 -1.3889 -1.0556 0.2139 9.3545
Coefficients:
49
(Intercept) 2.112e+01 1.470e-01 143.743 < 2e-16 ***
sqCogn -8.454e-03 8.582e-04 -9.851 < 2e-16 ***
CognProx 1.301e-01 8.536e-03 15.247 < 2e-16 ***
InsproxTRUE -6.751e-02 1.779e-02 -3.794 0.000148 ***
OrgproxTRUE -2.547e-01 2.455e-01 -1.038 0.299465
Gprox3TRUE -5.352e-01 7.120e-02 -7.517 5.6e-14 ***
StartDate.x -1.477e-08 1.135e-10 -130.183 < 2e-16 ***
noParticipants.x -5.660e-02 1.044e-03 -54.191 < 2e-16 ***
---
Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1
Null deviance: 49187 on 5332 degrees of freedom
Residual deviance: 27521 on 5324 degrees of freedom
(20 observations deleted due to missingness)
AIC: 38824

Overdispersion test
data: ptest7
z = 25, p-value < 2.2e-16
sample estimates:
dispersion
6.700497
50
Appendix 2
Table 16 The table with the combinations that are added for the missing link prediction.
index V1 V2
1 AVANTIUM CHEMICALS BV BTG Biomass Technology Group BV
CROWN VAN GELDER
2 AVANTIUM CHEMICALS BV PAPIERFABRIEKEN N.V.
3 AVANTIUM CHEMICALS BV Delft University of Technology
DS SMITH PACKAGING NETHERLANDS
4 AVANTIUM CHEMICALS BV BV
ECN - Energy Research Centre of the
5 AVANTIUM CHEMICALS BV Netherlands
6 AVANTIUM CHEMICALS BV Eindhoven University of Technology
7 AVANTIUM CHEMICALS BV ESKA GRAPHIC BOARD BV
8 AVANTIUM CHEMICALS BV GEMEENTE AMSTERDAM
9 AVANTIUM CHEMICALS BV KWR WATER BV
10 AVANTIUM CHEMICALS BV MAYR-MELNHOF EERBEEK BV
11 AVANTIUM CHEMICALS BV Nederlands Normalisatie-Instituut
12 AVANTIUM CHEMICALS BV Parenco BV
13 AVANTIUM CHEMICALS BV PROVALOR B.V.
14 AVANTIUM CHEMICALS BV Rijksuniversiteit Groningen
15 AVANTIUM CHEMICALS BV SAPPI NETHERLANDS SERVICES BV
16 AVANTIUM CHEMICALS BV Smurfit Kappa Group
STICHTING PUBLIC PRIVATE
PARTNERSHIP INSTITUTE FOR
17 AVANTIUM CHEMICALS BV SUSTAINABLE PROCESSTECHNOLOGY
TNO - Netherlands Organisation for Applied
18 AVANTIUM CHEMICALS BV Scientific Research Den Haag
19 AVANTIUM CHEMICALS BV Universiteit Leiden /Leiden University
20 AVANTIUM CHEMICALS BV Universiteit Twente
21 AVANTIUM CHEMICALS BV VAN HOUTUM PAPIER BV
22 AVANTIUM CHEMICALS BV VITENS N.V. Zwolle
Vrije Universiteit Amsterdam/VU University
23 AVANTIUM CHEMICALS BV Amsterdam
24 AVANTIUM CHEMICALS BV Wageningen UR
CROWN VAN GELDER
25 BTG Biomass Technology Group BV PAPIERFABRIEKEN N.V.
26 BTG Biomass Technology Group BV Delft University of Technology
27 BTG Biomass Technology Group BV BV
28 BTG Biomass Technology Group BV Netherlands
29 BTG Biomass Technology Group BV Eindhoven University of Technology
30 BTG Biomass Technology Group BV ESKA GRAPHIC BOARD BV
31 BTG Biomass Technology Group BV GEMEENTE AMSTERDAM
32 BTG Biomass Technology Group BV KWR WATER BV
51
33 BTG Biomass Technology Group BV MAYR-MELNHOF EERBEEK BV
34 BTG Biomass Technology Group BV Nederlands Normalisatie-Instituut
35 BTG Biomass Technology Group BV Parenco BV
36 BTG Biomass Technology Group BV PROVALOR B.V.
37 BTG Biomass Technology Group BV Rijksuniversiteit Groningen
38 BTG Biomass Technology Group BV SAPPI NETHERLANDS SERVICES BV
39 BTG Biomass Technology Group BV Smurfit Kappa Group
40 BTG Biomass Technology Group BV SUSTAINABLE PROCESSTECHNOLOGY
41 BTG Biomass Technology Group BV Scientific Research Den Haag
42 BTG Biomass Technology Group BV Universiteit Leiden /Leiden University
43 BTG Biomass Technology Group BV Universiteit Twente
44 BTG Biomass Technology Group BV VAN HOUTUM PAPIER BV
45 BTG Biomass Technology Group BV VITENS N.V. Zwolle
46 BTG Biomass Technology Group BV Amsterdam
47 BTG Biomass Technology Group BV Wageningen UR
CROWN VAN GELDER
48 PAPIERFABRIEKEN N.V. Delft University of Technology
CROWN VAN GELDER DS SMITH PACKAGING NETHERLANDS
49 PAPIERFABRIEKEN N.V. BV
CROWN VAN GELDER ECN - Energy Research Centre of the
50 PAPIERFABRIEKEN N.V. Netherlands
CROWN VAN GELDER
51 PAPIERFABRIEKEN N.V. Eindhoven University of Technology
CROWN VAN GELDER
52 PAPIERFABRIEKEN N.V. ESKA GRAPHIC BOARD BV
CROWN VAN GELDER
53 PAPIERFABRIEKEN N.V. GEMEENTE AMSTERDAM
CROWN VAN GELDER
54 PAPIERFABRIEKEN N.V. KWR WATER BV
CROWN VAN GELDER
55 PAPIERFABRIEKEN N.V. MAYR-MELNHOF EERBEEK BV
CROWN VAN GELDER
56 PAPIERFABRIEKEN N.V. Nederlands Normalisatie-Instituut
CROWN VAN GELDER
57 PAPIERFABRIEKEN N.V. Parenco BV
CROWN VAN GELDER
58 PAPIERFABRIEKEN N.V. PROVALOR B.V.
CROWN VAN GELDER
59 PAPIERFABRIEKEN N.V. Rijksuniversiteit Groningen
CROWN VAN GELDER
60 PAPIERFABRIEKEN N.V. SAPPI NETHERLANDS SERVICES BV
CROWN VAN GELDER
61 PAPIERFABRIEKEN N.V. Smurfit Kappa Group
52
CROWN VAN GELDER PARTNERSHIP INSTITUTE FOR
62 PAPIERFABRIEKEN N.V. SUSTAINABLE PROCESSTECHNOLOGY
CROWN VAN GELDER TNO - Netherlands Organisation for Applied
63 PAPIERFABRIEKEN N.V. Scientific Research Den Haag
CROWN VAN GELDER
64 PAPIERFABRIEKEN N.V. Universiteit Leiden /Leiden University
CROWN VAN GELDER
65 PAPIERFABRIEKEN N.V. Universiteit Twente
CROWN VAN GELDER
66 PAPIERFABRIEKEN N.V. VAN HOUTUM PAPIER BV
CROWN VAN GELDER
67 PAPIERFABRIEKEN N.V. VITENS N.V. Zwolle
CROWN VAN GELDER Vrije Universiteit Amsterdam/VU University
68 PAPIERFABRIEKEN N.V. Amsterdam
CROWN VAN GELDER
69 PAPIERFABRIEKEN N.V. Wageningen UR
70 Delft University of Technology BV
71 Delft University of Technology Netherlands
72 Delft University of Technology Eindhoven University of Technology
73 Delft University of Technology ESKA GRAPHIC BOARD BV
74 Delft University of Technology GEMEENTE AMSTERDAM
75 Delft University of Technology KWR WATER BV
76 Delft University of Technology MAYR-MELNHOF EERBEEK BV
77 Delft University of Technology Nederlands Normalisatie-Instituut
78 Delft University of Technology Parenco BV
79 Delft University of Technology PROVALOR B.V.
80 Delft University of Technology Rijksuniversiteit Groningen
81 Delft University of Technology SAPPI NETHERLANDS SERVICES BV
82 Delft University of Technology Smurfit Kappa Group
83 Delft University of Technology SUSTAINABLE PROCESSTECHNOLOGY
84 Delft University of Technology Scientific Research Den Haag
85 Delft University of Technology Universiteit Leiden /Leiden University
86 Delft University of Technology Universiteit Twente
87 Delft University of Technology VAN HOUTUM PAPIER BV
88 Delft University of Technology VITENS N.V. Zwolle
89 Delft University of Technology Amsterdam
90 Delft University of Technology Wageningen UR
DS SMITH PACKAGING ECN - Energy Research Centre of the
91 NETHERLANDS BV Netherlands
53
DS SMITH PACKAGING
92 NETHERLANDS BV Eindhoven University of Technology
DS SMITH PACKAGING
93 NETHERLANDS BV ESKA GRAPHIC BOARD BV
DS SMITH PACKAGING
94 NETHERLANDS BV GEMEENTE AMSTERDAM
DS SMITH PACKAGING
95 NETHERLANDS BV KWR WATER BV
DS SMITH PACKAGING
96 NETHERLANDS BV MAYR-MELNHOF EERBEEK BV
DS SMITH PACKAGING
97 NETHERLANDS BV Nederlands Normalisatie-Instituut
DS SMITH PACKAGING
98 NETHERLANDS BV Parenco BV
DS SMITH PACKAGING
99 NETHERLANDS BV PROVALOR B.V.
DS SMITH PACKAGING
100 NETHERLANDS BV Rijksuniversiteit Groningen
DS SMITH PACKAGING
101 NETHERLANDS BV SAPPI NETHERLANDS SERVICES BV
DS SMITH PACKAGING
102 NETHERLANDS BV Smurfit Kappa Group
DS SMITH PACKAGING PARTNERSHIP INSTITUTE FOR
103 NETHERLANDS BV SUSTAINABLE PROCESSTECHNOLOGY
DS SMITH PACKAGING TNO - Netherlands Organisation for Applied
104 NETHERLANDS BV Scientific Research Den Haag
DS SMITH PACKAGING
105 NETHERLANDS BV Universiteit Leiden /Leiden University
DS SMITH PACKAGING
106 NETHERLANDS BV Universiteit Twente
DS SMITH PACKAGING
107 NETHERLANDS BV VAN HOUTUM PAPIER BV
DS SMITH PACKAGING
108 NETHERLANDS BV VITENS N.V. Zwolle
DS SMITH PACKAGING Vrije Universiteit Amsterdam/VU University
109 NETHERLANDS BV Amsterdam
DS SMITH PACKAGING
110 NETHERLANDS BV Wageningen UR
111 Netherlands Eindhoven University of Technology
112 Netherlands ESKA GRAPHIC BOARD BV
113 Netherlands GEMEENTE AMSTERDAM
114 Netherlands KWR WATER BV
54
115 Netherlands MAYR-MELNHOF EERBEEK BV
116 Netherlands Nederlands Normalisatie-Instituut
117 Netherlands Parenco BV
118 Netherlands PROVALOR B.V.
119 Netherlands Rijksuniversiteit Groningen
120 Netherlands SAPPI NETHERLANDS SERVICES BV
121 Netherlands Smurfit Kappa Group
ECN - Energy Research Centre of the PARTNERSHIP INSTITUTE FOR
122 Netherlands SUSTAINABLE PROCESSTECHNOLOGY
ECN - Energy Research Centre of the TNO - Netherlands Organisation for Applied
123 Netherlands Scientific Research Den Haag
124 Netherlands Universiteit Leiden /Leiden University
125 Netherlands Universiteit Twente
126 Netherlands VAN HOUTUM PAPIER BV
127 Netherlands VITENS N.V. Zwolle
ECN - Energy Research Centre of the Vrije Universiteit Amsterdam/VU University
128 Netherlands Amsterdam
129 Netherlands Wageningen UR
130 Eindhoven University of Technology ESKA GRAPHIC BOARD BV
131 Eindhoven University of Technology GEMEENTE AMSTERDAM
132 Eindhoven University of Technology KWR WATER BV
133 Eindhoven University of Technology MAYR-MELNHOF EERBEEK BV
134 Eindhoven University of Technology Nederlands Normalisatie-Instituut
135 Eindhoven University of Technology Parenco BV
136 Eindhoven University of Technology PROVALOR B.V.
137 Eindhoven University of Technology Rijksuniversiteit Groningen
138 Eindhoven University of Technology SAPPI NETHERLANDS SERVICES BV
139 Eindhoven University of Technology Smurfit Kappa Group
140 Eindhoven University of Technology SUSTAINABLE PROCESSTECHNOLOGY
141 Eindhoven University of Technology Scientific Research Den Haag
142 Eindhoven University of Technology Universiteit Leiden /Leiden University
55
143 Eindhoven University of Technology Universiteit Twente
144 Eindhoven University of Technology VAN HOUTUM PAPIER BV
145 Eindhoven University of Technology VITENS N.V. Zwolle
146 Eindhoven University of Technology Amsterdam
147 Eindhoven University of Technology Wageningen UR
148 ESKA GRAPHIC BOARD BV GEMEENTE AMSTERDAM
149 ESKA GRAPHIC BOARD BV KWR WATER BV
150 ESKA GRAPHIC BOARD BV MAYR-MELNHOF EERBEEK BV
151 ESKA GRAPHIC BOARD BV Nederlands Normalisatie-Instituut
152 ESKA GRAPHIC BOARD BV Parenco BV
153 ESKA GRAPHIC BOARD BV PROVALOR B.V.
154 ESKA GRAPHIC BOARD BV Rijksuniversiteit Groningen
155 ESKA GRAPHIC BOARD BV SAPPI NETHERLANDS SERVICES BV
156 ESKA GRAPHIC BOARD BV Smurfit Kappa Group
157 ESKA GRAPHIC BOARD BV SUSTAINABLE PROCESSTECHNOLOGY
158 ESKA GRAPHIC BOARD BV Scientific Research Den Haag
159 ESKA GRAPHIC BOARD BV Universiteit Leiden /Leiden University
160 ESKA GRAPHIC BOARD BV Universiteit Twente
161 ESKA GRAPHIC BOARD BV VAN HOUTUM PAPIER BV
162 ESKA GRAPHIC BOARD BV VITENS N.V. Zwolle
163 ESKA GRAPHIC BOARD BV Amsterdam
164 ESKA GRAPHIC BOARD BV Wageningen UR
165 GEMEENTE AMSTERDAM KWR WATER BV
166 GEMEENTE AMSTERDAM MAYR-MELNHOF EERBEEK BV
167 GEMEENTE AMSTERDAM Nederlands Normalisatie-Instituut
168 GEMEENTE AMSTERDAM Parenco BV
169 GEMEENTE AMSTERDAM PROVALOR B.V.
170 GEMEENTE AMSTERDAM Rijksuniversiteit Groningen
171 GEMEENTE AMSTERDAM SAPPI NETHERLANDS SERVICES BV
172 GEMEENTE AMSTERDAM Smurfit Kappa Group
173 GEMEENTE AMSTERDAM SUSTAINABLE PROCESSTECHNOLOGY
174 GEMEENTE AMSTERDAM Scientific Research Den Haag
175 GEMEENTE AMSTERDAM Universiteit Leiden /Leiden University
176 GEMEENTE AMSTERDAM Universiteit Twente
177 GEMEENTE AMSTERDAM VAN HOUTUM PAPIER BV
178 GEMEENTE AMSTERDAM VITENS N.V. Zwolle
179 GEMEENTE AMSTERDAM Amsterdam
56
180 GEMEENTE AMSTERDAM Wageningen UR
181 KWR WATER BV MAYR-MELNHOF EERBEEK BV
182 KWR WATER BV Nederlands Normalisatie-Instituut
183 KWR WATER BV Parenco BV
184 KWR WATER BV PROVALOR B.V.
185 KWR WATER BV Rijksuniversiteit Groningen
186 KWR WATER BV SAPPI NETHERLANDS SERVICES BV
187 KWR WATER BV Smurfit Kappa Group
188 KWR WATER BV SUSTAINABLE PROCESSTECHNOLOGY
189 KWR WATER BV Scientific Research Den Haag
190 KWR WATER BV Universiteit Leiden /Leiden University
191 KWR WATER BV Universiteit Twente
192 KWR WATER BV VAN HOUTUM PAPIER BV
193 KWR WATER BV VITENS N.V. Zwolle
194 KWR WATER BV Amsterdam
195 KWR WATER BV Wageningen UR
196 MAYR-MELNHOF EERBEEK BV Nederlands Normalisatie-Instituut
197 MAYR-MELNHOF EERBEEK BV Parenco BV
198 MAYR-MELNHOF EERBEEK BV PROVALOR B.V.
199 MAYR-MELNHOF EERBEEK BV Rijksuniversiteit Groningen
200 MAYR-MELNHOF EERBEEK BV SAPPI NETHERLANDS SERVICES BV
201 MAYR-MELNHOF EERBEEK BV Smurfit Kappa Group
202 MAYR-MELNHOF EERBEEK BV SUSTAINABLE PROCESSTECHNOLOGY
203 MAYR-MELNHOF EERBEEK BV Scientific Research Den Haag
204 MAYR-MELNHOF EERBEEK BV Universiteit Leiden /Leiden University
205 MAYR-MELNHOF EERBEEK BV Universiteit Twente
206 MAYR-MELNHOF EERBEEK BV VAN HOUTUM PAPIER BV
207 MAYR-MELNHOF EERBEEK BV VITENS N.V. Zwolle
208 MAYR-MELNHOF EERBEEK BV Amsterdam
209 MAYR-MELNHOF EERBEEK BV Wageningen UR
210 Nederlands Normalisatie-Instituut Parenco BV
211 Nederlands Normalisatie-Instituut PROVALOR B.V.
212 Nederlands Normalisatie-Instituut Rijksuniversiteit Groningen
213 Nederlands Normalisatie-Instituut SAPPI NETHERLANDS SERVICES BV
214 Nederlands Normalisatie-Instituut Smurfit Kappa Group
215 Nederlands Normalisatie-Instituut SUSTAINABLE PROCESSTECHNOLOGY
57
216 Nederlands Normalisatie-Instituut Scientific Research Den Haag
217 Nederlands Normalisatie-Instituut Universiteit Leiden /Leiden University
218 Nederlands Normalisatie-Instituut Universiteit Twente
219 Nederlands Normalisatie-Instituut VAN HOUTUM PAPIER BV
220 Nederlands Normalisatie-Instituut VITENS N.V. Zwolle
221 Nederlands Normalisatie-Instituut Amsterdam
222 Nederlands Normalisatie-Instituut Wageningen UR
223 Parenco BV PROVALOR B.V.
224 Parenco BV Rijksuniversiteit Groningen
225 Parenco BV SAPPI NETHERLANDS SERVICES BV
226 Parenco BV Smurfit Kappa Group
227 Parenco BV SUSTAINABLE PROCESSTECHNOLOGY
228 Parenco BV Scientific Research Den Haag
229 Parenco BV Universiteit Leiden /Leiden University
230 Parenco BV Universiteit Twente
231 Parenco BV VAN HOUTUM PAPIER BV
232 Parenco BV VITENS N.V. Zwolle
233 Parenco BV Amsterdam
234 Parenco BV Wageningen UR
235 PROVALOR B.V. Rijksuniversiteit Groningen
236 PROVALOR B.V. SAPPI NETHERLANDS SERVICES BV
237 PROVALOR B.V. Smurfit Kappa Group
238 PROVALOR B.V. SUSTAINABLE PROCESSTECHNOLOGY
239 PROVALOR B.V. Scientific Research Den Haag
240 PROVALOR B.V. Universiteit Leiden /Leiden University
241 PROVALOR B.V. Universiteit Twente
242 PROVALOR B.V. VAN HOUTUM PAPIER BV
243 PROVALOR B.V. VITENS N.V. Zwolle
244 PROVALOR B.V. Amsterdam
245 PROVALOR B.V. Wageningen UR
246 Rijksuniversiteit Groningen SAPPI NETHERLANDS SERVICES BV
247 Rijksuniversiteit Groningen Smurfit Kappa Group
248 Rijksuniversiteit Groningen SUSTAINABLE PROCESSTECHNOLOGY
58
249 Rijksuniversiteit Groningen Scientific Research Den Haag
250 Rijksuniversiteit Groningen Universiteit Leiden /Leiden University
251 Rijksuniversiteit Groningen Universiteit Twente
252 Rijksuniversiteit Groningen VAN HOUTUM PAPIER BV
253 Rijksuniversiteit Groningen VITENS N.V. Zwolle
254 Rijksuniversiteit Groningen Amsterdam
255 Rijksuniversiteit Groningen Wageningen UR
256 SAPPI NETHERLANDS SERVICES BV Smurfit Kappa Group
257 SAPPI NETHERLANDS SERVICES BV SUSTAINABLE PROCESSTECHNOLOGY
258 SAPPI NETHERLANDS SERVICES BV Scientific Research Den Haag
259 SAPPI NETHERLANDS SERVICES BV Universiteit Leiden /Leiden University
260 SAPPI NETHERLANDS SERVICES BV Universiteit Twente
261 SAPPI NETHERLANDS SERVICES BV VAN HOUTUM PAPIER BV
262 SAPPI NETHERLANDS SERVICES BV VITENS N.V. Zwolle
263 SAPPI NETHERLANDS SERVICES BV Amsterdam
264 SAPPI NETHERLANDS SERVICES BV Wageningen UR
265 Smurfit Kappa Group SUSTAINABLE PROCESSTECHNOLOGY
266 Smurfit Kappa Group Scientific Research Den Haag
267 Smurfit Kappa Group Universiteit Leiden /Leiden University
268 Smurfit Kappa Group Universiteit Twente
269 Smurfit Kappa Group VAN HOUTUM PAPIER BV
270 Smurfit Kappa Group VITENS N.V. Zwolle
271 Smurfit Kappa Group Amsterdam
272 Smurfit Kappa Group Wageningen UR
SUSTAINABLE TNO - Netherlands Organisation for Applied
273 PROCESSTECHNOLOGY Scientific Research Den Haag
SUSTAINABLE
274 PROCESSTECHNOLOGY Universiteit Leiden /Leiden University
SUSTAINABLE
275 PROCESSTECHNOLOGY Universiteit Twente
59
SUSTAINABLE
276 PROCESSTECHNOLOGY VAN HOUTUM PAPIER BV
SUSTAINABLE
277 PROCESSTECHNOLOGY VITENS N.V. Zwolle
SUSTAINABLE Vrije Universiteit Amsterdam/VU University
278 PROCESSTECHNOLOGY Amsterdam
SUSTAINABLE
279 PROCESSTECHNOLOGY Wageningen UR
TNO - Netherlands Organisation for
280 Applied Scientific Research Den Haag Universiteit Leiden /Leiden University
281 Applied Scientific Research Den Haag Universiteit Twente
282 Applied Scientific Research Den Haag VAN HOUTUM PAPIER BV
283 Applied Scientific Research Den Haag VITENS N.V. Zwolle
TNO - Netherlands Organisation for Vrije Universiteit Amsterdam/VU University
284 Applied Scientific Research Den Haag Amsterdam
285 Applied Scientific Research Den Haag Wageningen UR
286 Universiteit Leiden /Leiden University Universiteit Twente
287 Universiteit Leiden /Leiden University VAN HOUTUM PAPIER BV
288 Universiteit Leiden /Leiden University VITENS N.V. Zwolle
289 Universiteit Leiden /Leiden University Amsterdam
290 Universiteit Leiden /Leiden University Wageningen UR
291 Universiteit Twente VAN HOUTUM PAPIER BV
292 Universiteit Twente VITENS N.V. Zwolle
293 Universiteit Twente Amsterdam
294 Universiteit Twente Wageningen UR
295 VAN HOUTUM PAPIER BV VITENS N.V. Zwolle
296 VAN HOUTUM PAPIER BV Amsterdam
297 VAN HOUTUM PAPIER BV Wageningen UR
298 VITENS N.V. Zwolle Amsterdam
299 VITENS N.V. Zwolle Wageningen UR
60
Vrije Universiteit Amsterdam/VU
300 University Amsterdam Wageningen UR
61

Master's Thesis Shamiram Abdulahad

Uploaded by

Copyright:

Available Formats

Master's Thesis Shamiram Abdulahad

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Master's Thesis Shamiram Abdulahad

Uploaded by

Copyright:

Available Formats

Collaboration in the Circular Economy in

the Netherlands: A Proximity Approach

MSc thesis Innovation Sciences, Utrecht University (GEO4-2239X)

Supervisor: Dr. K. Frenken

Title Page ........................................................................................................................................................ 1

This leads us to the research questions:

2.1 Research question 1

Following the proximity framework, Hypothesis 2 will be divided in four sub-hypotheses:

2.2 Research question 2

3.1.2 Dependent Variables

3.1.3 Independent Variables

3.2 Data analysis

Table 2 descriptive statistics for the variables for research question 1

Organizations Org Number of Number of City County

Biodetection system BV IND 1 39 Amsterdam North Holland

Org Number Number of City County Broker

4.2 Research question 1

4.3 Research question 2

Table 7 Descriptive statistics for all variables for research question 2

4.3.3 All organizations

4.4 Research question 3

Table 12 The 25 organizations with the most collaborations (based on degree)

patent codes cpc, Eurostat 2018

“circular product design”

“life cycle assessment”

Y02W 30/00 - Technologies for solid waste management

Y02W 90/00 - Enabling technologies or technologies with a potential or indirect contribution to

Independent variable: Eigenvector centrality

ptest1<- glm(n.x~eigenvec+ StartDate.x + noParticipants.x + ProjectEUfundin

(Dispersion parameter for poisson family taken to be 1)

Number of Fisher Scoring iterations: 7

Independent variable: betweenness centrality

ptesty2<- glm(n.x~between+ StartDate.x + noParticipants.x + ProjectEUfundin

Number of Fisher Scoring iterations: 7

Independent variable: brokerage centrality

ptest3<- glm(n.x~brok+ StartDate.x + noParticipants.x + ProjectEUfunding.x,

(Dispersion parameter for poisson family taken to be 1)

Number of Fisher Scoring iterations: 7

Independent variable: degree centrality

ptesty5<- glm(n.x~degree+ StartDate.x + noParticipants.x + ProjectEUfunding

Null deviance: 1879.8 on 292 degrees of freedom

ptesty7<- glm(n.x~sqCogn+ CognProx+ Insprox +Orgprox+ Gprox3+ StartDate.x+

(Dispersion parameter for poisson family taken to be 1)

Number of Fisher Scoring iterations: 7

Independent variables (international organisations included):

ptest7<- glm(n.x~sqCogn+ CognProx+ Insprox +Orgprox+ Gprox3+ StartDate.x+ P

Number of Fisher Scoring iterations: 6

You might also like