Using archetypal analysis to derive a typology of knowledge networks in European bioclusters

Abbasiharofteh, Milad; Hermans, Frans Article — Published Version Using archetypal analysis to derive a typology of knowledge networks in European bioclusters Regional Studies Provided in Cooperation with: Leibniz Institute of Agricultural Development in Transition Economies (IAMO), Halle (Saale) Suggested Citation: Abbasiharofteh, Milad; Hermans, Frans (2025) : Using archetypal analysis to derive a typology of knowledge networks in European bioclusters, Regional Studies, ISSN 1360-0591, Taylor & Francis, London, Vol. 59, Iss. 1, pp. 1-17, https://doi.org/10.1080/00343404.2024.2430354 , https://www.tandfonline.com/doi/full/10.1080/00343404.2024.2430354 This Version is available at: https://hdl.handle.net/10419/315645 Standard-Nutzungsbedingungen: Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Zwecken und zum Privatgebrauch gespeichert und kopiert werden. Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich machen, vertreiben oder anderweitig nutzen. Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, gelten abweichend von diesen Nutzungsbedingungen die in der dort genannten Lizenz gewährten Nutzungsrechte. Terms of use: Documents in EconStor may be saved and copied for your personal and scholarly purposes. You are not to copy documents for public or commercial purposes, to exhibit the documents publicly, to make them publicly available on the internet, or to distribute or otherwise use the documents in public. If the documents have been made available under an Open Content Licence (especially Creative Commons Licences), you may exercise further usage rights as specified in the indicated licence. http://creativecommons.org/licenses/by/4.0/ Using archetypal analysis to derive a typology of knowledge networks in European bioclusters Milad Abbasiharofteh a and Frans Hermans b ABSTRACT The shift to a bioeconomy necessitates radical innovations to move from a fossilto a bio-based economy, with bioclusters playing a crucial role by fostering collaboration among a wide range of co-located actors to find solutions. Through a novel methodological approach in regional studies (i.e., archetypal analysis), this study discusses the specificities of bioclusters and analyses the relational properties within European bioclusters’ knowledge networks. Results suggest four archetypal knowledge networks. It is rare to observe the transition of biocluster networks from one archetype group to another. The findings provide a data-driven bioclusters typology, offering insights into designing informed place-based policies. KEYWORDS bioclusters; knowledge networks; inventive activities; archetypal analysis; unsupervised learning JEL D85, O30, O31, Q57 HISTORY Received 28 September 2022; in revised form 29 October 2024 1. INTRODUCTION The urgency for green innovations, particularly at the regional level, has become increasingly apparent in light of global environmental challenges and the need for sustainable development (Gibbs & O’Neill, 2017). The bioeconomy has come up as a concept promoted by different governments worldwide as a potential pathway for this transition (Stark et al., 2022). The bioeconomy encompasses using renewable biological resources to extract essential components for materials, chemicals, and energy. These raw materials and their associated waste streams are then processed to yield value-added products, including but not limited to food, feed, bioplastics, pharmaceuticals and bioenergy. The overarching objective of the bioeconomy lies in reducing our current reliance on carbon-based sources derived from fossil fuels by replacing them with renewable sources of carbon rooted in photosynthesis. Apart from the shift away from fossil fuels, the bioeconomy also promises to contribute to regional sustainable development, generate employment in high-tech sectors, and promote the creation of green innovations (McCormick & Kautto, 2013). By capitalising on regional assets, such as biodiversity, agricultural lands, and natural resources, regions can harness the potential of the bioeconomy and create new development paths (Refsgaard et al., 2021). Nevertheless, the intricacies of this process remain unclear, underscoring the need for a closer examination of the bioeconomy and how it influences innovation processes at the regional level. In this paper, we focus on the role of bioeconomy clusters, or bioclusters for short, that are essential tools for governments to promote the regional bioeconomy and to establish, promote, and strengthen economic collaboration, learning, and innovation processes within particular regions (Hermans, 2018). The idea is that they possess the potential to integrate local resources, knowledge, and networks into biobased innovations that result in sustainability effects having impacts beyond their local level (Ayrapetyan et al., 2022; Kamath et al., 2022). As such, the growing attention to bioclusters falls into a broader trend in cluster research to take the role of clusters in the generation of sustainable, green or eco-tech innovations more into account (McCauley & Stephens, 2012; Njøs et al., 2017; Sedita & Blasi, 2021). The literature on industrial clusters has paid much attention to the underlying network characteristics of clusters and other territorial innovation systems (Abbasiharofteh, 2020; Abbasiharofteh & Maghssudipour, 2024; Hermans, 2020). This strand of literature discusses that clusters © 2025 The Author(s). Published by Informa UK Limited, trading as Taylor & Francis Group This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. The terms on which this article has been published allow the posting of the Accepted Manuscript in a repository by the author(s) or with their consent. CONTACT Milad Abbasiharofteh [email protected]; [email protected] a Department of Economic Geography, Faculty of Spatial Sciences, University of Groningen, Groningen, the Netherlands b Department of Structural Change, Leibniz Institute of Agricultural Development in Transition Economies, Halle, Germany Supplemental data for this article can be accessed online at https://doi.org/10.1080/00343404.2024.2430354 REGIONAL STUDIES 2025, VOL. 59, NO. 1, 2430354 https://doi.org/10.1080/00343404.2024.2430354 differ based on their level of development and sectoral, institutional, and technological compositions. That said, studies that investigate the knowledge networks within these green-tech, or sustainable clusters, are sparse. Our research question is, therefore: What are the key characteristics of bioclusters’ knowledge networks? With this paper, we aim for two different contributions: a methodological and an empirical one. First, we want to illustrate a relatively new and, we think, promising tool for the further development of network studies within regional studies and economic geography archetypal analysis of networks (D’Esposito et al., 2010). To answer our research question, we utilised a methodological framework based on an unsupervised clustering algorithm to create benchmarks for analysing the structural properties of knowledge networks in biocluster research. Our study marks the first application of this method in regional studies, despite its common use in other scientific disciplines ranging from biology to astronomy (Chan et al., 2003; Gimbernat-Mayol et al., 2022). Our approach has broader applicability making it suitable for future regional studies that aim to address classification problems to answer theoretically informed research questions. Second, our research makes an empirical contribution by first discussing key characteristics of bioclusters’ knowledge networks and their specificities related to innovation processes in the bioeconomy. Subsequently, we systematically identify and track European bioclusters over time, ‘from the bottom-up’, adding insights to the ongoing debate surrounding this concept. Creating a data-driven typology of bioclusters enabled us to distinguish bioclusters based on their placeand path-dependent attributes. In this paper, we discuss how this typology helps policymakers to design bioeconomy policy in the future that is place-sensitive and address the specific needs of bioclusters. The paper is structured as follows. Section 2 discusses explicitly some of the characteristics of innovation processes within the bioeconomy and bioclusters and draws implications for the resulting knowledge networks and their evolution. Section 3 introduces our methodology with the specific machine learning approach we applied. Section 4 presents and discusses the results and suggests how future studies can take our study as a point of departure to classify other types of clusters. Section 5 concludes the paper and discusses potential policy implications concerning the success of place-based innovation policy measures. 2. INNOVATION AND KNOWLEDGE NETWORKS IN INDUSTRIAL CLUSTERS The interest in the causes and effects of agglomerations and clusters goes a long way back to the work of Marshall (1890), who observed the geographical co-location of firms in different industrial sectors in 19th-century Britain and explained them through the effects of labour market pooling, supplier specialisation and knowledge spillovers. Since the 1990s clusters have become a popular tool for governments to strengthen regional economic competitiveness and innovation (McCauley & Stephens, 2012; Porter, 1998). With innovation being an important goal of many cluster policies, interest has grown into the characteristics of the knowledge networks within agglomerations and clusters and how these change over time (Glückler, 2007; Powell et al., 2005). This type of research investigates the spatial distributions of regional economic processes not in isolation, but instead they consider the dynamic and interconnected nature of economic activities and uses a lens of social relationships, networks and interactions (Bathelt & Glückler, 2003; Boggs & Rantisi, 2003; Oinas & Malecki, 2002). This research investigates how network characteristics of an industrial cluster influence both the performance of the individual organisations within the cluster as well as its overall functioning (Belussi et al., 2010; Breschi & Malerba, 2005; Karlsson et al., 2005). With the ‘relation turn’, the literature on industrial clusters has developed and applied several different methods: analysis tools, models, and procedures to aid in these investigations: social network analysis (SNA) (Giuliani & Pietrobelli, 2011; Ter Wal & Boschma, 2009), (adapted) gravity models, quadratic assignment procedures (QAP) (Broekel et al., 2014; Simensen & Abbasiharofteh, 2022), and statistical network models such as exponential random graph models (ERGMs) and stochastic actororiented models (SAOMs) (Abbasiharofteh, 2020; Hermans, 2020; Maggioni et al., 2007). In this section, we will discuss three topics. First, we discuss some of the current network approaches in regional studies and economic geography, and how they relate to some of the methodological toolboxes in use. We will shortly introduce the archetypal analysis and compare it with existing methods. We will specifically discuss the context of our empirical case study: biocluster and their main overlap and differences with other sectoral industrial clusters. We will end this section with a discussion on how the characteristics of bioeconomy will likely influence also the type of innovation processes and their associated knowledge networks within bioclusters. 2.1. Analysing knowledge networks in industrial clusters: methods and applications Network analysis is a methodological approach used to study the structures of networks, focusing on nodes (actors) and edges (relationships) to understand their interactions and dynamics. Previous research has applied network analysis to industrial and innovation clusters to understand interactions, collaborations, and dynamics within clusters (Broekel et al., 2021). Early network studies focused especially on how the structural position of an actor within the cluster knowledge network would determine its economic and innovative performance. This type of research looks at the influence of structural holes (Ahuja, 2000; Walker et al., 1997) or the position of a firm within the core or the periphery of the cluster (Giuliani, 2013). Later, other models investigate the characteristics of the complete cluster network: how 2 Milad Abbasiharofteh and Frans Hermans REGIONAL STUDIES these networks were formed under the influence of microlevel mechanisms of tie formation between actors, for instance, through reciprocity, homophily or triadic closure. Often, these micro-level processes were then linked to different forms of proximity: geographical, institutional, social, cognitive, and organisational and their interplay (Janssen & Abbasiharofteh, 2022; Kabirigi et al., 2022). Other studies looked into how these networks changed over time in studies of network evolution (Menzel et al., 2017; Nicotra et al., 2013; Ter Wal, 2013), or a combination of these two: how different forms of proximity influence network evolution over time (Balland, 2012; Lazzeretti & Capone, 2016). Although network theory and network tools have become increasingly popular in cluster and agglomeration studies, there is no consensus on the effects of different types of network constellations on a cluster’s innovative performance. What is considered beneficial in some instances can become detrimental in other cases: new and emerging industrial clusters versus old and established industrial clusters, sectoral specifics of clusters (high tech versus low tech), the type of innovation being pursued: incremental or radical and explorative versus exploitative. It is therefore important to analyse the features of a sectoral cluster, in our case bioclusters. This would also align with Hermans (2020), who calls for more reflection on a cluster’s internal and external characteristics in the setup of a network study. In a similar vein, Glückler and Panitz (2021) argue for the use of ‘informed pre-specification’, which they define as ‘actively integrating the already existing substantive understanding of a particular research context into the specification of a formal network model’ (p. 19). In this paper, we utilised archetypal analysis to study knowledge networks in clusters, starting from this ‘informed pre-specification’. The archetypical analysis involves the identification of original patterns or models in multivariate data. In other words, the archetypal model aims to find ‘pure types’ in the data and determine how each datum point is a combination of configurations represented by those pure types (Cutler & Breiman, 1994; Eugster & Leisch, 2009). The flexibility of this method allows for classifying a wide range of data types ranging from texts to cities, to spatial diffusion patterns, and to networks. In this context, the problem is identifying several hypothetical points (archetypes) within a set of multivariate data based on theoretically informed indices (Cutler & Breiman, 1994). Network scientists use this technique to reduce the dimensionality of network data and classify them based on network attributes (Porzio et al., 2008; Ragozini et al., 2017). This technique is relevant for investigating cluster networks for two reasons. First, the model is flexible, enabling us to define network indices that are theoretically relevant for specific types of clusters. In this paper, we will illustrate the approach in the case of bioeconomy clusters, but the method would also be adaptable for use in other industrial clusters. Second, the results show how each data point is a mixture of identified pure types. The latter is particularly important for understanding the evolutionary trajectories of clusters by uncovering the underlying structures and dynamics (see the method section for a more detailed account of the archetypal analysis method). In the following section, we will undertake informed pre-specification to investigate biocluster knowledge networks. 2.2. Green innovations in the context of the bioeconomy and bioclusters In this section, we first analyse some of the particularities of knowledge networks in bioclusters. It is important to note that although the bioeconomy and bioclusters are often associated with positive policy goals as sustainable development, agricultural development, and eco-innovations (Wilde & Hermans, 2021), in reality, many sectors related to the bioeconomy carry a heavy pollution load. Not all bio-based innovations can be called sustainable, and some can even have serious detrimental sustainability effects and trade-offs at other scales and levels (Ayrapetyan & Hermans, 2020; Bröring et al., 2020). Therefore, Kamath et al. (2022) argue that many bioclusters need to make a sustainability transition before contributing to the overarching goals of green innovation. Based on the earlier work of Markusen (1996) and Iammarino and McCann (2006), Hermans (2021) identifies four different types of bioclusters and their potential sustainability effects in their localised production chains from raw materials to production and use (Table 1). Table 1 explores the sustainability of various bioclusters, highlighting their environmental impacts and innovation dynamics. First, agricultural agglomeration and green chemistry bioclusters depend heavily on biomass production, often using imported biomass, which can cause significant environmental pressures. Marshallian bio-districts, which can localise their production processes, may achieve self-sufficiency and thus have a closer link to local natural resources. Life science clusters, on the other hand, rely minimally on biological sources, making biomass sourcing less critical. At the same time, bioclusters servicing global markets can create substantial local environmental pressures due to their concentrated production. Green chemistry clusters can theoretically enhance sustainability through circular ‘cradle-to-cradle’ designs based on industrial ecology principles (Ayrapetyan et al., 2022). The typology of bioclusters shows that bioclusters are a broad topic where innovation processes can vary according to the type of knowledge base: doing, using and interacting (DUI) and science, technology, innovation (STI) (Jensen et al., 2007). Clusters using DUI knowledge, typical of agricultural agglomerations and bio-districts, often lack resources for high-tech innovations. Environmental innovations usually come from government-induced public– private partnerships and scale through local knowledge transfer (Hermans et al., 2019). On the other hand, green chemistry and life science clusters rely on STI knowledge for high-tech innovations. Regional universities play a key role in training specialists, partnering in R&D, and generating spin-offs. These clusters’ transformative potential lies in developing environmentally Using archetypal analysis to derive a typology of knowledge networks in European bioclusters 3 REGIONAL STUDIES friendly production methods and new products, though life sciences’ sustainability impacts are initially more economic and social. 2.3. Pre-specification of biocluster knowledge networks In this paper, we will focus on the types of bioclusters that rely more on codified forms of knowledge: life science clusters and green chemistry clusters. This focus simplifies the complexity associated with the wide variety of bioclusters in theory and practice and aligns with the nature of our patent-based database (see the methodology section). Drawing on Golembiewski et al. (2015) and Van Lancker et al. (2016), we identify four key factors that characterise sustainable bio-based innovation processes and are especially relevant for forming knowledge networks in these clusters. First, a transition to the bioeconomy would require radical changes in the existing business models and supply chains. Even innovations, such as so-called ‘drop-in chemicals’, that focus on the manufacturing of known chemical building blocks, like (bio)methane and (bio)ethanol, will require radical changes in their supply chains as refineries make the transition towards biorefineries and switch from a single source of input (crude oil) that is available all year around to a large number of smaller suppliers and where weather and climate effects have to be taken into account (Birch & Calvert, 2015). The switch towards these new value chains may be challenging and gradual, requiring a focus on different forms of connectivity as this switch involves a move from well-established relationships in the fossil fuel industry to fragmented, shorter connections in new value chains (Wilde & Hermans, 2024). Second, the bioeconomy encompasses a complex knowledge base derived from diverse fields, including life sciences, biotechnology, agronomy, food science, social science, nanotechnology, renewable energy, waste management, information and communication technologies (ICT), and engineering (Golembiewski et al., 2015). The complexity of the knowledge base necessitates integrating diverse knowledge types within our sample, starting from a broad range of patents rather than a narrowly defined knowledge base. In addition, this broad knowledge base also requires specific attention to the issue of multidisciplinarity, as many innovations occur on the interface of specific knowledge bases in different forms of regional diversification (Boschma et al., 2017). Third, policy and institutional effects in the bioeconomy also shape preferred networking types. Established Table 1. Theoretical biocluster types and networks. Biocluster type Examples Links to resource base Production and manufacturing Goal of innovation in sustainability transition Agri-food clusters (agricultural agglomerations) Horticultural clusters, wine clusters, intensive animal husbandry areas Sector specific: .Crops and diary: mainly local .Intensive animal husbandry: global Local negative environmental effects at cluster level Optimising efficiency of production (through NARS) Green chemistry clusters Biorefineries, green chemistry (industrial biotech), paper and pulp clusters Sector specific: .Paper and pulp: mainly local .Biofuel and bioenergy: global Local negative environmental effects at cluster level .Improve efficiency of conversion .Industrial ecology Marshallian biodistricts Design, fashion, leather, wood construction, and building Medium to high dependence on local inputs, often linked to regional branding Risk of outsourcing environmentally polluting activities (textiles, leather) .Demand-driven product design for ecofriendly consumers .Localisation of production chains and agro-ecology Life science clusters Pharmaceuticals and medicine (red biotech), cosmetics Low Low .Innovation as a goal: new products and process incubator .Risk of rebound effect Note: NARS, National Agricultural Research System. Source: Adapted from Hermans (2021), used with permission. 4 Milad Abbasiharofteh and Frans Hermans REGIONAL STUDIES industries, especially those related to fossil fuels and the chemical industry, restrictive policies, and high commercialisation costs lead to significant barriers due to high switching costs and a lack of quality standards (Wilde & Hermans, 2021). Sectors with restrictive policies and high commercialisation barriers are more likely to develop hierarchical knowledge networks. These networks are characterised by a few dominant nodes (key institutions or companies) controlling the flow of information, leading to slower knowledge dissemination but potentially more controlled and high-quality innovation outputs. However, sectors with flexible policies, supportive environments, and efforts to reduce adoption barriers may cultivate smallworld networks. These networks have numerous interconnections among nodes, promoting rapid and wide-reaching knowledge transfer and fostering innovation through frequent interactions and collaborations. These networks can be characterised by their small-worldliness. Finally, it is important to note that we expect a mix of established and emerging industrial clusters, with some clusters emerging around completely net technologies and other bioclusters that are actually in different transition phases, for instance, moving from ‘brown to green’, or ‘path upgrading’ (Trippl et al., 2020). Following Menzel and Fornahl’s (2010) cluster evolution theory, the industrial networks within a cluster only develop in one of the later stages, meaning that some European bioclusters might still be in an early development stage, and one can still expect the connectedness of their networks. This argument also resonates with other empirical findings that show collaborative networks often start as fragmented and ‘fluid’ (Fritsch & Zoellner, 2020). This means for our knowledge networks, we are particularly interested in the issues of (1) connectivity, (2) hierarchy and (3) smallworldliness. We will further elaborate on this in the methodology section. 3. METHODOLOGY 3.1. Construction of bioclusters and their knowledge networks Inspired by the work of Marshall (1890) and Porter (1998), one can describe an industrial cluster as a geographical concentration of firms and affiliated organisations, such as universities, within one or multiple interconnected industries. These entities are linked through input–output relationships, informal interactions, cooperative local networks, and labour mobility. An empirical definition of a cluster is daunting (Martin & Sunley, 2003). More recent conceptual and empirical works suggest that detecting clusters should go beyond finding a geographical concentration of economic activities and consider inter-organisational relations (Janssen & Abbasiharofteh, 2022; Delgado et al., 2016). Following these studies, we have used patents to map the geographical concentration of bioeconomy collaborative relations. In other words, we define bioclusters from a network perspective as regions with a significantly higher number of inventive ties than surrounding regions. The assumption is that this number also correlates with input–output relations, informal interactions, cooperative local networks, and labour mobility. While one should be aware of problems associated with using this type of data (Archibugi & Planta, 1996; Fritsch et al., 2020), the assumption is that joint inventive activities are accompanied by knowledge exchange and mutual learning. We used the Organisation for Economic Cooperation and Development’s (OECD) REGPAT database. 1 This database provides information regarding the geographical location of inventors and applicants at the NUTS3 level, International Patent Classification (IPC) technological classes, and respective filing and grant dates for each patent. Concerning our broad definition of a cluster it is important to note that patents are always filed by one or several inventors, and those individuals can be associated with a business, but also a university, research institute, legal firm or government agency. As such, patents are a source of public and private organisations. Since this study focuses on identifying and analysing bioeconomy clusters, we should first systematically define bioeconomy related patents. Technology codes included in patent documents provide valuable information about utilised technology in inventions. Kriesch and Losacker (2024) employed advanced machine learning techniques to identify bioeconomy patent abstracts. Their research suggests that several agricultural and food-related technology classes are associated with an exceptionally high share of bioeconomy patents. Accordingly, we followed the extensive literature review of Frietsch et al. (2016) that identified the key IPC technology codes for the bioeconomy (see Appendix A in the supplemental data online). Interestingly, the suggested technologies align with the findings of Kriesch and Losacker (2024). For instance, both studies identified bioeconomy technologies such as new plants or processes for obtaining them; plant reproduction by tissue culture techniques (A01H), and organic fertilisers, for instance, fertilisers from waste (C05F). We selected patents that include technology codes associated with biotechnology (biotech) and at least one of 12 following bioeconomy categories: (1) agriculture and forestry, (2) pulp and paper, (3) machines cartons boxes printing, (4) genetic engineering, (5) landscape management, (6) food, (7) proteins, (8) biofuels (including biofuels for transport), (9) biomass, (10) biomaterials, (11) marine, (12) animals livestock management, and (13) food-related household appliances. Using this approach, we filtered 568,883 patents (17% of patents) at the aggregate level. Our approach resonates with our conceptual focus on interdisciplinary inventive activities that underline a set of activities in which biotech contributes to other bioeconomy technologies by providing cuttingedge technology solutions. Next, we projected two-mode (inventor-by-project) collaborative knowledge networks to one-mode (inventor-by-inventor) networks. We used the information on inventors’ geographical position (NUTS2 European regions) to create intraand interregional networks. 2 Following a common approach, the data were disaggregated Using archetypal analysis to derive a typology of knowledge networks in European bioclusters 5 REGIONAL STUDIES into six five-year time-windows: (1) 1986–90, (2) 1991– 95, (3) 1996–2000, (4) 2001–05, (5) 2006–10 and (6) 2011–15 (Abbasiharofteh et al., 2023a; Li et al., 2014; Menzel et al., 2017; Ter Wal, 2014). For the sake of simplicity, we address each time-window with the last year of each time-window. As a cut-off point for the cluster region, we defined the 90th percentile of the density for each time-window (see Appendix B in the supplemental data online). However, it is essential to note that defining cluster based on administrative boundary such as the NUTS classification system, is criticised due to problems associated with the so-called modifiable area unit problem (Brenner, 2017; Scholl & Brenner, 2014). To tackle this problem, we included collaborators outside bioclusters as a part of bioclusters’ networks if there is at least one collaborative tie between them. Inspired by the relational turn in economic geography, this implies that knowledge transfer relations are not limited to the administrative boundaries of a given cluster. Thus, we allocated an interregional collaborative connection to more than one NUTS2 region if they are home to inventors who developed a given patent (Bathelt & Glückler, 2003). 3.2. Stylised characterisation of cluster knowledge networks Based on our review of innovation processes within the bioeconomy we identified three elements as particularly salient: hierarchy (as displayed in scale-free networks), small word networks and connectivity. The scale-free networks of Barabási and Albert (1999) and the small-world networks of Watts and Strogatz (1998) belong to the most well-known network structures in science. These two networks can have both positive and negative effects on innovation outcomes. The scale-free, centralised networks of Barabási and Albert (1999) possess many ‘structural holes’. Bridging a structural hole provides individuals with novel and non-redundant information (Burt, 1992, 2005). The skewed degree distribution of scale-free networks, with only a few nodes that possess the vast majority of connections, results in a star-like network configuration that increases the likelihood of actors receiving new information. If we define innovation as the novel combination of existing knowledge, then such a scale-free network could potentially be associated with more innovations due to the efficiency of a centralised network in knowledge circulation (Albert et al., 2000). Another significant network pattern discussed in the literature that holds potential for fostering innovation is the ‘small-world’ structure, as introduced by Watts and Strogatz (1998). This type of network exhibits a high level of clustering (transitivity) and short average path length between its nodes. Scholars in the field of social sciences have emphasised the importance of clustering as a form of social embeddedness resulting from shared past experiences. Clustering, in turn, facilitates trust formation, limits opportunistic behaviour, and reduces transaction costs (Coleman, 1988; Granovetter, 1985; Uzzi, 1997). Besides clustering, the second critical element is connectivity within the network. Connectivity plays a critical role in how knowledge is transferred among individuals and organisations because highly fragmented knowledge networks do not provide the required knowledge transfer channels (Fleming et al., 2007). Connectivity can work in two ways. As clusters develop, it is expected that their knowledge networks will become more connected over time (Menzel & Fornahl, 2010), however as we have argued in the theoretical section in the switching process from established industries towards more biobased production, the connectivity of a cluster might at the first time show a decrease of connectivity over time. Figure 1 brings the two connectivity and degree distribution elements into a single framework. We can identify four quadrants based on the network hierarchy (scale-free versus small-world) and connectivity (low versus high density). It is plausible that complex knowledge networks entail structural properties of both small-world and scale-free networks at the same time. For instance, an ill-structured core–periphery knowledge network could show smallworldness at the core (i.e., a high degree of clustering and short average path length) and scale-freeness in the periphery where poorly connected nodes are connected to nodes at the core (Vicente, 2017). 3.3. Using archetypal analysis to create a knowledge network typology Investigating the structural properties of knowledge networks requires using multiple network indices as proxies for various aspects of networks. Although network indices might theoretically capture different structural properties of a network, they are often correlated. The archetypal analysis method suggested by Cutler and Breiman (1994) is an unsupervised machine learning technique that reduces the dimensionality of data by grouping observations into a set of archetypes using a data-driven approach. Instead of looking for ‘typical’ observations (cluster centres), it seeks extremal points in multidimensional data. We used this method to benchmark bioclusters knowledge networks (Porzio et al., 2008; Ragozini et al., 2017). We describe this method below in two steps: (1) selecting a set of parameters that describe the structure of networks; and (2) defining a set of archetypes that maximise within-archetypes commonalities and between-archetypes differences. 3.3.1. Selecting a set of network parameters We have operationalised the four network types of Figure 1by several indicators. Table 2 presents a list of network measures and corresponding methods that reflect the extent to which the observed networks are scale-free, small-world and connected. We refrained from including distance-based indices (e.g., the average path length) in the analysis because the observed bioclusters collaboration networks are often fragmented. To control for the clustering problem associated with the projection of bipartite networks, we applied the so6 Milad Abbasiharofteh and Frans Hermans REGIONAL STUDIES called ‘network of places’ (Abbasiharofteh et al., 2023a; Lucena-Piquero & Vicente, 2019; Pizarro, 2007). This clustering problem occurs when working with the projection of bipartite networks. This clustering becomes problematic because team size increases over time (Van der Wouden, 2020) and varies across different technologies (Broekel, 2019), causing a different degree of clustering and biases cluster-based measures for structural properties. To address the projection bias and reduce the impact of clustering caused by network projection, we utilise the concept of ‘structural equivalence’, which has its roots in social network theory. This concept was developed by Lorrain and White (1971) and Burt (1987). In this approach, nodes within a network are considered structurally equivalent if they share identical relationships. This equivalence enables them to access similar resources within the network (Gnyawali & Madhavan, 2001; Stuart & Podolny, 1996). Figure 2 illustrates three networks. The visualisation on the left-hand side shows a bipartite network in which inventors A, B and C collaborate on developing a patent, and inventors C and D are involved in developing another patent. This network can be projected to a onemode network shown in the middle, which provides a high degree of clustering due to projecting a two-mode network. The method of networks of places (Figure 2, right) enables us to project a two-mode network without increasing the degree of clustering in the network. Earlier research shows that the new network (i.e., a network of places) is not biased towards different team sizes over time and across technologies. It represents the structural properties of original networks (Abbasiharofteh et al., 2023a). Our analysis is based on the empirical investigation of networks of places (for the sake of consistency, we use the term knowledge networks or collaborative networks). The second problem is associated with the fact that some network indices are size variant. This implies that estimated indices of collaborative networks with different sizes need to be normalised and then used as inputs for the archetypal analysis method. To normalise, we plotted each network index (corresponding to a biocluster in a given time-window) on the y-axis and the size of the network (node number) on the x-axis. We calculated the residuals of the non-parametric regression fit (the distance between each dot and the blue spline in Figure 3, left panel). In this way, the network size does not strongly affect the normalised coefficients (Figure 3, right panel). 3.3.2. Defining a set of archetypes Based on the previous step, each network can be presented as a set of numerical vectors (i.e., normalised network indices). The archetypal analysis method is an unsupervised machine learning technique to define distinct categories by minimising the within-group and maximising the between-group squared errors in each mixture of identified archetypes in the multivariate Euclidean space (Cutler & Breiman, 1994). Formally, Ragozini et al. (2017) take a set of n networks V={xi} i ¼1, … , n, x i ∊ Rp and a division C ¼(C 1 , … , C k ) of Ω in K groups, and define an internal similarity as R(x i , C h ), x i ∊ C h , and external dissimilarity as D(x i , Ch), x i ∉ C h . A mixing function Φ(.) combines two internal similarity and external dissimilarity Figure 1. Stylised examples of knowledge network types with a similar number of nodes. Note: Randomly removing half of the ties leads to a highly fragmented network with a scale-free structure, whereas the smallworld network is not split into multiple components. Using archetypal analysis to derive a typology of knowledge networks in European bioclusters 7 REGIONAL STUDIES functions: T(xi,Ch)=F(R(xi,Ch); D(xi,Ch)) (1) A set of archetypes A ¼(a 1 , … , a k ) is then defined as: A={ah[Rp|ahargmax T(xi,Ch), h =1, ...,K} (2) This equation provides multiple solutions suggesting a different number of archetypes with corresponding residual sum of squares (RSS) coefficients that show the ability of defined archetypes to categorise networks into distinct groups (for technical details, see Ragozini et al., 2017). Using the elbow plot, one can select the minimum number of archetypes that capture the optimal degree of variance in the observed data. Finally, one can use a fuzzy membership function and estimate a ‘pureness parameter’ to maximise the internal similarity of archetypes. Then, networks are assigned to a specific archetype when its distance in the archetypal space is closer to one archetype and far from the other. They are not assigned a particular archetype if they have a similar distance to two or several defined archetypes. To provide an example of the above-discussed method and to test the reliability of the archetypal analysis method in grouping networks based on their structural properties, we simulated 60 networks with three distinct structural properties. We used the archetypal analysis method to group them based on network indices described in Table 1. More specifically, we simulated 20 small-world networks, 20 scale-free networks and 20 random networks and randomly removed between 10% and 50% of their ties to create different degrees of connectivity. Figure 4 shows that this technique provides a reliable method of grouping networks based on structural properties. A1 represents an archetype for small-world structural properties, A2 for scale-free ones, and A3 and A4 for random networks. Four random, five smallworld and one scale-free networks are assigned to none of the four archetype groups, perhaps due to their low connectivity. In contrast, the method is highly accurate in assigning networks to each group. That is, the number of false-positive cases is zero. Table 2. Overview of network indices for describing the structure of bioclusters knowledge networks. Structural property Measure Method Variable Small-worldness Community number Number of identified communities using the multilevel modularity optimisation algorithm COMMUNITY Share of intercommunity ties Number of ties connecting communities divided by the total number of ties using the multilevel modularity optimisation algorithm INTERCOM Modularity Modularity scores MODULARITY Transitivity Clustering coefficient (Wasserman & Faust, 1994)TRANSITIVITY Scale-freeness a Network centrality GINI coefficient for actors’ degree centrality (Giuliani, 2013)CENTRALITY Connectivity Share of the largest component Number of nodes in the largest component divided by the total number of nodes COMPONENT Share of isolates Number of isolated nodes divided by the total number of nodes ISOLATES Density Number of ties divided by the number of possible ties (Wasserman & Faust, 1994) DENSITY Note: a We used other network centrality indices and coreness (K-core decomposition) as a proxy for scale-freeness. We refrain from including these indices in the analysis as these indices are highly correlated with the GINI coefficient for actors’ degree centrality. Figure 2. The method of networks of places. 8 Milad Abbasiharofteh and Frans Hermans REGIONAL STUDIES Bathelt, H., & Glückler, J. (2003). Toward a relational economic geography. Journal of Economic Geography, 3(2), 117–144. https://doi.org/10.1093/jeg/3.2.117 Beise, M., & Gemünden, H. G. (2004). Lead markets: A new framework for the international diffusion of innovation. In K. Macharzina, M. Oesterle & J. Wolf (Eds.), Management international review (pp. 83–98). Gabler. Belussi, F., Sammarra, A., & Sedita, S.R. (2010). Learning at the boundaries in an ‘open regional innovation system’: A focus on firms’ innovation strategies in the Emilia Romagna life science industry. Research Policy, 39(6), 710–721. Birch, K., & Calvert, K. (2015). Rethinking ‘drop-in’ biofuels. Science & Technology Studies, 28(1), 52–72. Boggs, J. S., & Rantisi, N. M. (2003). The ‘relational turn’ in economic geography. Journal of Economic Geography, 3(2), 109–116. Boschma, R., Coenen, L., Frenken, K., & Truffer, B. (2017). Towards a theory of regional diversification: Combining insights from evolutionary economic geography and transition studies. Regional Studies, 51(1), 31–45. Brenner, T. (2017). Identification of clusters – An actor-based approach (Working Papers on Innovation and Space No. 2.17). PhilippsUniversität Marburg. Breschi, S., & Malerba, F. (2005). Clusters, networks, and innovation. Oxford University Press. Broekel, T. (2019). Using structural diversity to measure the complexity of technologies. PloS One, 14(5), e0216856. Broekel, T., Balland, P.-A., Burger, M., & van Oort, F. (2014). Modeling knowledge networks in economic geography: A discussion of four methods. Annals of Regional Science, 53(2), 423–452. Broekel, T., Lazzeretti, L., Capone, F., & Hassink, R. (2021). Rethinking the role of local knowledge networks in territorial innovation models. Industry & Innovation, 28(7), 805–814. Bröring, S., Laibach, N., & Wustmans, M. (2020). Innovation types in the bioeconomy. Journal of Cleaner Production, 266(6), 121939. Burt, R. S. (1987). Social contagion and innovation: Cohesion versus structural equivalence. American Journal of Sociology, 92(6), 1287–1335. https://doi.org/10.1086/228667 Burt, R. S. (1992). Structural holes: The social structure of competition. Harvard University Press. Burt, R. S. (2005). Brokerage and closure: An introduction to social capital. Oxford University Press. Chan, B. H. P., Mitchell, D. A., & Cram, L. E. (2003). Archetypal analysis of galaxy spectra. Monthly Notices of the Royal Astronomical Society, 338(3), 790–795. https://doi.org/10.1046/ j.1365-8711.2003.06099.x Coleman, J. S. (1988). Social capital in the creation of human capital. American Journal of Sociology, 94, 95–120. https://doi.org/10. 1086/228943 Csardi, G., & Nepusz, T. (2006). The igraph software package for complex network research. InterJournal, Complex Systems, 1695. Cutler, A., & Breiman, L. (1994). Archetypal analysis. Technometrics, 36(4), 338–347. https://doi.org/10.1080/ 00401706.1994.10485840 D’Esposito, M. R., Ragozini, G., & Vistocco, D. (2010). Exploring data through archetypes. In H. Locarek-Junge & C. Weihs (Eds.), Classification as a tool for research (pp. 287–298). Springer. Delgado, M., Porter, M. E., & Stern, S. (2016). Defining clusters of related industries. Journal of Economic Geography, 16(1), 1–38. Eugster, M. J. A., & Leisch, F. (2009). From Spider-Man to hero – Archetypal analysis in R. Journal of Statistical Software, 30(8), 1–23. Eugster, M. J. A., & Leisch, F. (2011). Weighted and robust archetypal analysis. Computational Statistics & Data Analysis, 55(3), 1215–1225. Fleming, L., King, C., & Juda, A. I. (2007). Small worlds and regional innovation. Organization Science, 18(6), 938–954. https://doi.org/10.1287/orsc.1070.0289 Frietsch, R., Neuhäusler, P., Rothengatter, O., & Jonkers, K. (2016). Societal grand challenges from a technological perspective – Methods and identification of classes of the International Patent Classification IPC (Discussion Papers in Innovation Systems and Policy Analysis No. 53). Fraunhofer ISI. Fritsch, M., Titze, M., & Piontek, M. (2020). Identifying cooperation for innovation – A comparison of data sources. Industry & Innovation, 27(6), 630–659. https://doi.org/10. 1080/13662716.2019.1650253 Fritsch, M., & Zoellner, M. (2020). The fluidity of inventor networks. Journal of Technology Transfer, 45(4), 1063–1087. Gibbs, D., & O’Neill, K. (2017). Future green economies and regional development: A research agenda. Regional Studies, 51(1), 161–173. Gimbernat-Mayol, J., Dominguez Mantes, A., Bustamante, C. D., Mas Montserrat, D., & Ioannidis, A. G. (2022). Archetypal analysis for population genetics. PLoS Computational Biology, 18(8), e1010301. https://doi.org/10.1371/journal.pcbi.1010301 Giuliani, E. (2013). Network dynamics in regional clusters: Evidence from Chile. Research Policy, 42(8), 1406–1419. https://doi.org/ 10.1016/j.respol.2013.04.002 Giuliani, E., & Petrobelli, C. (2011). Social network analysis methodologies for the evaluation of cluster development programs. InterAmerican Development Bank IDB Publ. Glückler, J. (2007). Economic geography and the evolution of networks Johannes. Journal of Economic Geography, 7(5), 619–634. https://doi.org/10.1093/jeg/lbm023 Glückler, J., & Panitz, R. (2021). Unleashing the potential of relational research: A meta-analysis of network studies in human geography. Progress in Human Geography, 45(6), 1531–1557. Gnyawali, D. R., & Madhavan, R. (2001). Cooperative networks and competitive dynamics: A structural embeddedness perspective. The Academy of Management Review, 26(3), 431. https:// doi.org/10.2307/259186 Golembiewski, B., Sick, N., & Bröring, S. (2015). The emerging research landscape on bioeconomy: What has been done so far and what is essential from a technology and innovation management perspective? Innovative Food Science & Emerging Technologies, 29(1), 308–317. Granovetter, M. (1985). Economic action and social structure: The problem of embeddedness. American Journal of Sociology, 91(3), 481–510. https://doi.org/10.1086/228311 Hermans, F. (2018). The potential contribution of transition theory to the analysis of bioclusters and their role in the transition to a bioeconomy. Biofuels, Bioproducts and Biorefining, 12(2), 265– 276. https://doi.org/10.1002/bbb.1861 Hermans, F. (2020). The contribution of statistical network models to the study of clusters and their evolution. Papers in Regional Science, 100(2), 379–404. https://doi.org/10.1111/pirs. 12579 Hermans, F. L. P. (2021). Bioclusters and sustainable development. In S. R. Sedita, & S. Blasi (Eds.), Rethinking clusters: Place-based value creation in sustainability transitions. Springer. Hermans, F., Geerling-Eiff, F., Potters, J., & Klerkx, L. (2019). Public–private partnerships as systemic agricultural innovation policy instruments – Assessing their contribution to innovation system function dynamics. NJAS: Wageningen Journal of Life Sciences, 88(1), 76–95. Iammarino, S., & McCann, P. (2006). The structure and evolution of industrial clusters: Transactions, technology and knowledge spillovers. Research Policy, 35(7), 1018–1036. https://doi.org/ 10.1016/j.respol.2006.05.004 Janssen, M. J., & Abbasiharofteh, M. (2022). Boundary spanning R&D collaboration: Key enabling technologies and missions as alleviators of proximity effects? Technological Forecasting and Social Change, 180(7), 121689. https://doi.org/10.1016/j. techfore.2022.121689 Using archetypal analysis to derive a typology of knowledge networks in European bioclusters 15 REGIONAL STUDIES Jensen, M. B., Johnson, B., Lorenz, E., & Lundvall, B. Å. (2007). Forms of knowledge and modes of innovation. Research Policy, 36(5), 680–693. https://doi.org/10.1016/j.respol.2007. 01.006 Kabirigi, M., Abbasiharofteh, M., Sun, Z., & Hermans, F. (2022). The importance of proximity dimensions in agricultural knowledge and innovation systems: The case of banana disease management in Rwanda. Agricultural Systems, 202(3), 103465. Kamath, R., Sun, Z., & Hermans, F. (2022). Policy instruments for green-growth of clusters: Implications from an agent-based model. Environmental Innovation and Societal Transitions, 43, 257–269. https://doi.org/10.1016/j.eist.2022.04.003 Karlsson, C., Johansson, B., & Stough, R. (2005). New horizons in regional science. Industrial clusters and inter-firm networks. Edward Elgar. Kriesch, L., & Losacker, S. (2024). A global patent dataset for the bioeconomy. Papers in Innovation Studies, 2024(/08), 1–15. Lazzeretti, L., & Capone, F. (2016). How proximity matters in innovation networks dynamics along the cluster evolution. A study of the high technology applied to cultural goods. Journal of Business Research, 69(12), 5855–5865. Li, G.-C., Lai, R., D’Amour, A., Doolin, D. M., Sun, Y., Torvik, V. I., Yu, A. Z., & Fleming, L. (2014). Disambiguation and coauthorship networks of the US patent inventor database (1975-2010). Research Policy, 43(6), 941–955. https://doi.org/ 10.1016/j.respol.2014.01.012 Lorrain, F., & White, H. C. (1971). Structural equivalence of individuals in social networks. The Journal of Mathematical Sociology, 1(1), 49–80. https://doi.org/10.1080/0022250X.1971.9989788 Lucena, D. (2017). Places: Structural equivalence analysis for two-mode networks R package v0.2.0. Lucena-Piquero, D., & Vicente, J. (2019). The visible hand of cluster policy makers: An analysis of aerospace valley (2006–2015) using a place-based network methodology. Research Policy, 48(3), 830–842. https://doi.org/10.1016/j.respol.2019.01.001 Maggioni, M. A., Nosvelli, M., & Uberti, T. E. (2007). Space versus networks in the geography of innovation: A European analysis. Papers in Regional Science, 86(3), 471–493. Markusen, A. (1996). Sticky places in slippery space: A typology of industrial districts. Economic Geography, 72(3), 293. Marshall, A. (1890). Principles of economics. Macmillan. Martin, R., & Sunley, P. (2003). Deconstructing clusters: Chaotic concept or policy panacea? Journal of Economic Geography, 3(1), 5–35. McCauley, S. M., & Stephens, J. C. (2012). Green energy clusters and socio-technical transitions: Analysis of a sustainable energy cluster for regional economic development in central Massachusetts, USA. Sustainability Science, 7(2), 213–225. McCormick, K., & Kautto, N. (2013). The bioeconomy in Europe: An overview. Sustainability, 5(6), 2589–2608. https://doi.org/ 10.3390/su5062589 Menzel, M.-P., Feldman, M. P., & Broekel, T. (2017). Institutional change and network evolution: Explorative and exploitative tie formations of co-inventors during the dot-com bubble in the research triangle region. Regional Studies, 51(8), 1179–1191. https://doi.org/10.1080/00343404.2016.1278300 Menzel, M.-P., & Fornahl, D. (2010). Cluster life cycles – Dimensions and rationales of cluster evolution. Industrial and Corporate Change, 19(1), 205–238. Nathan, M., & Rosso, A. (2022). Innovative events: Product launches, innovation and firm performance. Research Policy, 51(1), 104373. https://doi.org/10.1016/j.respol.2021.104373 Nicotra, M., Romano, M., & Del Giudice, M. (2013). The evolution dynamic of a cluster knowledge network: The role of firms’ absorptive capacity. Journal of the Knowledge Economy, 45(6), 425. Njøs, R., Jakobsen, S.-E., Wiig Aslesen, H., & Fløysand, A. (2017). Encounters between cluster theory, policy and practice in Norway: Hubbing, blending and conceptual stretching. European Urban and Regional Studies, 24(3), 274–289. https:// doi.org/10.1177/0969776416655860 Oinas, P., & Malecki, E. J. (2002). The evolution of technologies in time and space: From national and regional to spatial innovation systems. International Regional Science Review, 25(1), 102–131. Organisation for Economic Co-operation and Development (OECD). (2008). The OECD REGPAT Database (OECD Science, Technology and Industry Working Papers No. 2008/ 02). https://www.oecd-ilibrary.org/science-and-technology/ the-oecd-regpat-database_241437144144 Pizarro, N. (2007). Structural identity and equivalence of individuals in social networks: Beyond duality. International Sociology, 22(6), 767–792. https://doi.org/10.1177/0268580907082260 Porter, M. E. (1998). Clusters and competition: New agendas for companies, governments, and institutions. In M. E. Porter (Ed.), On competition (pp. 197–299). Harvard Business School Press. Porzio, G. C., Ragozini, G., & Vistocco, D. (2008). On the use of archetypes as benchmarks. Applied Stochastic Models in Business and Industry, 24(5), 419–437. https://doi.org/10.1002/asmb.727 Powell, W. W., White, D. R., Koput, K. W., & Owen-Smith, J. (2005). Network dynamics and field evolution: The growth of interorganizational collaboration in the life sciences. American Journal of Sociology, 110(4), 1132–1205. Ragozini, G., Palumbo, F., & D’Esposito, M. R. (2017). Archetypal analysis for data-driven prototype identification. Statistical Analysis and Data Mining: The ASA Data Science Journal, 10(1), 6–20. https://doi.org/10.1002/sam.11325 Refsgaard, K., Kull, M., Slätmo, E., & Meijer, M. W. (2021). Bioeconomy – A driver for regional development in the Nordic countries. New Biotechnology, 60, 130–137. https://doi. org/10.1016/j.nbt.2020.10.001 Scholl, T., & Brenner, T. (2014). Detecting spatial clustering using a firm-level cluster index. Regional Studies, 50(6), 1054–1068. https://doi.org/10.1080/00343404.2014.958456 Sedita, S. R., & Blasi, S. (2021). Rethinking clusters: Place-based value creation in sustainability transitions. Spinger. Simensen, E. O., & Abbasiharofteh, M. (2022). Sectoral patterns of collaborative tie formation: Investigating geographic, cognitive, and technological dimensions. Industrial and Corporate Change, 31(5), 1–36. https://doi.org/10.1093/icc/dtac021 Stark, S., Biber-Freudenberger, L., Dietz, T., Escobar, N., Förster, J. J., Henderson, J., Laibach, N., & Börner, J. (2022). Sustainability implications of transformation pathways for the bioeconomy. Sustainable Production and Consumption, 29, 215– 227. https://doi.org/10.1016/j.spc.2021.10.011 Stuart, T. E., & Podolny, J. M. (1996). Local search and the evolution of technological capabilities. Strategic Management Journal, 17(S1), 21–38. https://doi.org/10.1002/smj.4250171004 Ter Wal, A. (2014). The dynamics of the inventor network in German biotechnology: Geographic proximity versus triadic closure. Journal of Economic Geography, 14(3), 589–620. https://doi. org/10.1093/jeg/lbs063 Ter Wal, A. L. J. (2013). Cluster emergence and network evolution: A longitudinal analysis of the inventor network in SophiaAntipolis. Regional Studies, 47(5), 651–668. Ter Wal, A., & Boschma, R. A. (2009). Applying social network analysis in economic geography: Framing some key analytic issues. Annals of Regional Science, 43(3), 739–756. https://doi. org/10.1007/s00168-008-0258-3 Trippl, M., Baumgartinger-Seiringer, S., Frangenheim, A., Isaksen, A., & Rypestøl, J. O. (2020). Unravelling green regional industrial path development: Regional preconditions, asset modification and agency. Geoforum, 111(1), 189–197. 16 Milad Abbasiharofteh and Frans Hermans REGIONAL STUDIES Uzzi, B. (1997). Social structure and competition in interfirm networks: The paradox of embeddedness. Administrative Science Quarterly, 42(1), 35–67. https://doi.org/10.2307/2393808 Van der Wouden, F. (2020). A history of collaboration in US invention: Changing patterns of co-invention, complexity and geography. Industrial and Corporate Change, 29(3), 599–619. https:// doi.org/10.1093/icc/dtz058 Van Lancker, J., Wauters, E., & Van Huylenbroeck, G. (2016). Managing innovation in the bioeconomy: An open innovation perspective. Biomass and Bioenergy, 90, 60–69. https://doi.org/ 10.1016/j.biombioe.2016.03.017 Vicente, J. (2017). Network failures and policy challenges along the life cycle of cluster. In D. Fornahl, & R. Hassink (Eds.), The life cycle of clusters: A policy perspective (pp. 56–75). Edward Elgar. Walker, G., Kogut, B., & Shan, W. (1997). Social capital, structural holes and the formation of an industry network. Organization Science, 8(2), 109–125. Wasserman, S., & Faust, K. (1994). Social network analysis: Methods and applications. Cambridge University Press. Watts, D. J., & Strogatz, S. H. (1998). Collective dynamics of ‘small-world’ networks. Nature, 393(6684), 440–442. https:// doi.org/10.1038/30918 Wilde, K., & Hermans, F. (2021). Innovation in the bioeconomy: Perspectives of entrepreneurs on relevant framework conditions. Journal of Cleaner Production, 314, 127979. https://doi.org/10. 1016/j.jclepro.2021.127979 Wilde, K., & Hermans, F. (2024). Transition towards a bioeconomy: Comparison of conditions and institutional work in selected industries. Environmental Innovation and Societal Transitions, 50, 100814. Using archetypal analysis to derive a typology of knowledge networks in European bioclusters 17 REGIONAL STUDIES