The dynamism of clustering: Interweaving material and discursive processes

11
The dynamism of clustering: Interweaving material and discursive processes Arnt Fløysand a , Stig-Erik Jakobsen b,, Ove Bjarnar c a Department of Geography, University of Bergen, Fosswinckelsgt. 6, N-5007 Bergen, Norway b Bergen University College, P.O. Box 7030, N-5020 Bergen, Norway c Molde University College, P.O. Box 2110, N-6402 Molde, Norway article info Article history: Received 26 January 2011 Received in revised form 2 April 2012 Keywords: Clusters Actant Feedback loops Møre maritime cluster Hordaland subsea cluster abstract The aim of this paper is to contribute to an understanding of clusters, including both the material and discursive dynamism of cluster construction, and shed light on how clusters—once established—affect the actors, institutions and processes that constitute them. It does this by viewing clusters as an actant, i.e. something that acts or to which activity is granted by others. The empirical analysis examines two clusters in the public cluster programme Norwegian Centre of Expertise (NCE): the Møre maritime cluster and the Hordaland subsea cluster. It focuses on the type of development paths they are following and how the material and discursive processes are interweaved in these paths. The clusters are related to the concept of cluster construction, which is triggered by ideas, representations, policy and industry prac- tice. The Møre maritime cluster is characterized by bottom-up clustering processes and illustrates how the material practices of firms can trigger clustering processes such as the establishment of a cluster and the identification of a prototype of best cluster practice. On the other hand, the Hordaland subsea cluster expresses a top-down process and how the ideal world of academics and policy-making can encourage processes of clustering among co-located firms. Based on these observations of material and discursive interweaved clustering processes and how they affect both those who are practicing and those who are promoting them, we find it reasonable to argue for a stronger awareness of such feedback loops in cluster studies. Ó 2012 Elsevier Ltd. All rights reserved. 1. Introduction Clusters and clustering processes have caught the imagination of academics, policymakers and business people during the last cou- ple of decades (Asheim et al., 2006; Porter, 2000). Clusters come in many types, sizes and origins, and there is a wide array of cluster definitions. Concepts tend to develop a life of their own and this certainly applies to the concept of clusters (Martin and Sunley, 2003). The uncertainty surrounding the debate, however, calls for conceptual discussions (Malmberg and Power, 2006). The objective of this paper is to contribute towards conceptual clarification and theory building by elaborating upon two different, and to a certain degree, contradictory types of cluster approaches: clusters as mate- rial constructions and clusters as discursive constructions. We will outline the characteristics of these approaches and illustrate how the dualism of materialism and discursive constructions in cluster studies can be overcome by integrating ideas from actor-network theory (ANT). Approaching clusters as material constructions means that the use of the cluster concept is encouraged by the observed preceding practice and networking of co-located firms. Thus, a cluster becomes a materialized entity through bottom-up processes. Clusters as discursive constructions imply that the ideas among pol- icymakers, academics, firm representatives and other stakeholders of a potential cluster are actually stronger than the materialized cluster practice. This illustrates a cluster construction dominated by a top-down process. The cluster literature has identified different mechanisms driv- ing cluster growth and cluster development. We also find distinc- tions between various types of cluster construction and cluster origin that can be related to the distinction between clusters as material constructions and clusters as discursive constructions. Policy driven clusters and working or industry driven clusters are some of the categories discussed by Enright (2003), while Atherton and Johnston (2008) distinguish between potential, emerging and established clusters, where ideas and policy institutions are at the forefront of the development of the former. Still, we will argue that there is a lack of understanding in the literature of the co-existence of discursive and material processes in the development of clusters. There is a need for an approach that takes into account the multi- plicity of actors in networks and permits the fluidity of multiple types of relations and spatial configurations in cluster development. We will argue that by including ideas from ANT, there is a potential to overcome the dualism of material and discursive constructions 0016-7185/$ - see front matter Ó 2012 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.geoforum.2012.05.002 Corresponding author. E-mail addresses: [email protected] (A. Fløysand), Stig-Erik.Jakobsen @hib.no (S.-E. Jakobsen), [email protected] (O. Bjarnar). Geoforum 43 (2012) 948–958 Contents lists available at SciVerse ScienceDirect Geoforum journal homepage: www.elsevier.com/locate/geoforum

Transcript of The dynamism of clustering: Interweaving material and discursive processes

Geoforum 43 (2012) 948–958

Contents lists available at SciVerse ScienceDirect

Geoforum

journal homepage: www.elsevier .com/locate /geoforum

The dynamism of clustering: Interweaving material and discursive processes

Arnt Fløysand a, Stig-Erik Jakobsen b,⇑, Ove Bjarnar c

a Department of Geography, University of Bergen, Fosswinckelsgt. 6, N-5007 Bergen, Norwayb Bergen University College, P.O. Box 7030, N-5020 Bergen, Norwayc Molde University College, P.O. Box 2110, N-6402 Molde, Norway

a r t i c l e i n f o a b s t r a c t

Article history:Received 26 January 2011Received in revised form 2 April 2012

Keywords:ClustersActantFeedback loopsMøre maritime clusterHordaland subsea cluster

0016-7185/$ - see front matter � 2012 Elsevier Ltd. Ahttp://dx.doi.org/10.1016/j.geoforum.2012.05.002

⇑ Corresponding author.E-mail addresses: [email protected] (A.

@hib.no (S.-E. Jakobsen), [email protected] (O.

The aim of this paper is to contribute to an understanding of clusters, including both the material anddiscursive dynamism of cluster construction, and shed light on how clusters—once established—affectthe actors, institutions and processes that constitute them. It does this by viewing clusters as an actant,i.e. something that acts or to which activity is granted by others. The empirical analysis examines twoclusters in the public cluster programme Norwegian Centre of Expertise (NCE): the Møre maritime clusterand the Hordaland subsea cluster. It focuses on the type of development paths they are following andhow the material and discursive processes are interweaved in these paths. The clusters are related tothe concept of cluster construction, which is triggered by ideas, representations, policy and industry prac-tice. The Møre maritime cluster is characterized by bottom-up clustering processes and illustrates howthe material practices of firms can trigger clustering processes such as the establishment of a clusterand the identification of a prototype of best cluster practice. On the other hand, the Hordaland subseacluster expresses a top-down process and how the ideal world of academics and policy-making canencourage processes of clustering among co-located firms. Based on these observations of material anddiscursive interweaved clustering processes and how they affect both those who are practicing and thosewho are promoting them, we find it reasonable to argue for a stronger awareness of such feedback loopsin cluster studies.

� 2012 Elsevier Ltd. All rights reserved.

1. Introduction

Clusters and clustering processes have caught the imagination ofacademics, policymakers and business people during the last cou-ple of decades (Asheim et al., 2006; Porter, 2000). Clusters comein many types, sizes and origins, and there is a wide array of clusterdefinitions. Concepts tend to develop a life of their own and thiscertainly applies to the concept of clusters (Martin and Sunley,2003). The uncertainty surrounding the debate, however, calls forconceptual discussions (Malmberg and Power, 2006). The objectiveof this paper is to contribute towards conceptual clarification andtheory building by elaborating upon two different, and to a certaindegree, contradictory types of cluster approaches: clusters as mate-rial constructions and clusters as discursive constructions. We willoutline the characteristics of these approaches and illustrate howthe dualism of materialism and discursive constructions in clusterstudies can be overcome by integrating ideas from actor-networktheory (ANT).

Approaching clusters as material constructions means that theuse of the cluster concept is encouraged by the observed preceding

ll rights reserved.

Fløysand), Stig-Erik.JakobsenBjarnar).

practice and networking of co-located firms. Thus, a clusterbecomes a materialized entity through bottom-up processes.Clusters as discursive constructions imply that the ideas among pol-icymakers, academics, firm representatives and other stakeholdersof a potential cluster are actually stronger than the materializedcluster practice. This illustrates a cluster construction dominatedby a top-down process.

The cluster literature has identified different mechanisms driv-ing cluster growth and cluster development. We also find distinc-tions between various types of cluster construction and clusterorigin that can be related to the distinction between clusters asmaterial constructions and clusters as discursive constructions.Policy driven clusters and working or industry driven clusters aresome of the categories discussed by Enright (2003), while Athertonand Johnston (2008) distinguish between potential, emerging andestablished clusters, where ideas and policy institutions are at theforefront of the development of the former. Still, we will argue thatthere is a lack of understanding in the literature of the co-existenceof discursive and material processes in the development of clusters.There is a need for an approach that takes into account the multi-plicity of actors in networks and permits the fluidity of multipletypes of relations and spatial configurations in cluster development.We will argue that by including ideas from ANT, there is a potentialto overcome the dualism of material and discursive constructions

A. Fløysand et al. / Geoforum 43 (2012) 948–958 949

(Callon, 1999; Henderson et al., 2002; Hetherington and Law, 2000;Latour, 1996, 1999; Law, 1999; Whatmore and Thorne, 1997). Inparticular, the ANT concept of an actant can contribute towards amore dynamic understanding of cluster construction. An actant is‘‘. . .something that acts or to which activity is granted by others’’(Latour, 1996, p. 373). Hence, an actant is the source of action forthe various groups of actors involved. This corresponds nicely withan idea of cluster construction as something that is triggered bothby industry practice and by ideas, representations and policy. Onceestablished, clusters are also continually ‘‘performed’’ in a relationalcomplex of actors, representations and tasks giving rise to changingactor networks, new performances and new representations. Forinstance, we believe that an idea or a strong representation of a clus-ter can materialize a cluster-like practice among industry actors.The theoretical aim of this paper is to contribute towards an under-standing of clusters that includes both the material and discursivedynamism of cluster construction and that can shed light on howa cluster once established affects the actors, institutions andprocesses that constitute the cluster.

Following on from this objective, our theoretical assumptionsare elaborated and nuanced through two empirical cases, theMøre maritime cluster and the Hordaland subsea cluster, both lo-cated in the western part of Norway. We will consider the processesof agglomeration, networking and representation in each of the twoselected cases. Considering the feedback loops for clusters asactants, the two cases communicate the idea of cluster constructionas something that is triggered both by industry practice and byideas, representations and policy. Despite both environmentshaving obtained cluster status in the Norwegian public cluster pro-gramme Norwegian Centre of Expertise (NCE), their constructionprocesses have been quite dissimilar. The Møre maritime cluster(Section 3) has several characteristics of a complete industry clus-ter, with a high level of networking, knowledge sharing and collec-tive learning among its members. This also inspired policymakersto portray the Møre maritime cluster as a prototype for best clusterpractice. The Hordaland subsea cluster (Section 3), however, fea-tures a top-down construction process where cluster-like behaviorhas gradually gained momentum. As such, it confirms the assump-tion that the ideal world of academics and policymakers canencourage clustering practice among firms. In the final part of thepaper (Sections 4 and 5), we further elaborate upon our main obser-vation that cluster construction should not be regarded as some-thing that is based in either ideas or materialized practice.Instead, it should be viewed as a dynamic entity that is a combina-tion of discursive (cluster ideas, cluster policy, cluster representa-tions) and materialized practices (the practice of the industry).Before presenting our cases, we refine our theoretical argument inthe next section.

2. Introducing ‘‘actants’’ into cluster theory

2.1. Cluster conceptualization

Of growing interest in studies of economic practice is the ob-served tendency for many industries to form specialized concentra-tions in specific locations (Asheim et al., 2006; Fløysand andJakobsen, 2007). In general, there is a diverse assembly of clusterdefinitions in the literature and a large number of strategies forthe identification and mapping of a cluster. Porter’s (1998, 2000)work on clusters has proved especially influential and inspired pol-icymakers over the world to use his model as a tool for promotingregional competitiveness, innovation and growth. Innovative clus-ters are seen as a driver of national economic growth (Martin andSunley, 2003). Porter (2000, p. 15) defines a cluster as a ‘‘geographicconcentration of interconnected companies, specialized suppliers,

service provides, firm in related industries, and associated institu-tions (e.g. universities, standards agencies, trade associations) in aparticular field that compete but also cooperate’’. Despite beinghighly influential, his contributions have been heavily criticizedfor lack of preciseness according to the sectorial and spatial charac-teristics of a cluster (Bathelt, 2005; Martin and Sunley, 2003). Othercontributions include Malmberg and Power (2006, p. 57), who havepointed towards some generic criteria for cluster formation: ‘‘Thereshould be a spatial agglomeration of similar and related economicactivity; these activities should be interlinked by relations andinteractions of local collaboration and competition; there shouldbe some form of self-awareness among the cluster participantsand some joint policy action (we are a cluster and we are deter-mined to develop together)’’.

Still, clusters vary in type, size, origin and dynamics. Athertonand Johnston (2008) discuss how clusters can develop from apotential cluster to an established cluster. In a potential cluster,firms start to recognize that there may be scope for workingtogether. The possibilities for firm co-operation are latent, and cer-tain mutual recognitions are needed to get development underway. When such mutual recognitions have been triggered, thecluster can move into the phase of an emerging cluster. In thisphase, firms start to collaborate and processes of inter-firm knowl-edge sharing and learning exist in the initial stage. A critical factor,according to Atherton and Johnston, is the development of a com-mon approach to inter-firm collaboration among clustered firms. Ifthese initial collaboration projects turn out to be successful, theemerging cluster will eventually turn into an established cluster,where the structure and processes of group collaboration can beformalized. The transformation from a potential to an establishedcluster is not a straightforward process and can be guided andachieve momentum by various types of public cluster developmentinstruments, especially in the early phase of cluster development.

This approach is closely linked to some recent studies on clusterlife cycles. In a relative stylistic manner, Menzel and Fornahl(2010) distinguish between four stages in the life cycle of a cluster.In an emerging cluster, there are few companies, and firms arerelatively heterogeneous. In the growth phase of a cluster, thenumber of firms increases because of cluster dynamics, and theprofile of the cluster firms is more distinct. In a sustaining cluster,there exists a strong regional competence base and a fierce rivalrybetween clustered firms. The growth of the cluster is also dimin-ishing. In the final stage, i.e. a declining cluster, there is a decreasein the number of companies and employees. The development ofthe cluster is also hampered by closed networks and negativesentiments.

Another point of view, focusing more on the origin of the cluster,is the separation between an endogenous or self-organized clusterand an intentional cluster (Yeung et al., 2005). The origin of the firstone is closely linked to classical agglomeration and cluster effects(knowledge spillovers, collective learning, rivalry, complementar-ity, proximity, etc.), while an intentional cluster is closely linkedto policy strategies and government support in its preliminaryphase. Such an idea of policy driven clusters is further developedby Richardson (2010, p. 59), who defines this as clusters that‘‘. . .owe their origin to the direct action and strong commitmentof governmental actors who set the conditions for the clusterscreation either in response to an industrial crisis, or to promoteregional growth’’. Policymakers in many countries have tried toreplicate successful regional clusters by designing, promoting andcreating clusters, some of them lacking an industrial foundation(Martin and Sunley, 2003). Enright (2003) develops a more diversecategorization of various types of clusters Still, the distinction be-tween clusters developed from below through the mechanisms ofknowledge sharing, low transaction costs, localized trusts and in-ter-personal relationships, and cluster development from above,

950 A. Fløysand et al. / Geoforum 43 (2012) 948–958

through ideas, political commitment and strategies are evident inthe categorization. His concept of working clusters (a critical massof knowledge, expertise and resources that creates agglomerationeconomies), latent clusters (a co-location of firms that forms a crit-ical mass, but that have not developed the level of interactionnecessary to benefit from co-location) and potential clusters (clus-ters with only some of the elements necessary for a working clus-ter) illustrates industry driven clusters at various levels ofdevelopment. On the other hand, his concept of policy driven clus-ters (clusters chosen by the government for support, but which lackthe critical mass of firms or favorable conditions) and ‘‘wishfulthinking’’ clusters (assembly of firms that lacks both critical massand any particular source of advantages) is closely linked to the roleof ideas and policy institutions for promoting cluster thinking andcluster development.

Still, we will argue that the cluster literature fails to recognizethe co-existence of discursive elements (cluster theory, cluster pol-icy, cluster representations) and material elements (co-location offirms, networking, etc.) that characterize cluster development inthe real world. There is also a tendency to reduce the developmentof a cluster to a linear development starting with the co-location offirms in an emerging cluster and ending up in a dynamic workingcluster with political institutions of particular importance in theinitial stages. Our point of departure is that cluster developmentmust be viewed as a fluent and complex process that does notfollow a pre-given trajectory. Cluster ideas, cluster policy andmaterialized cluster practice are interwoven in the various phasesof the development giving rise to changing representations andchanging networks within the cluster. To capture this materialand discursive dynamism of cluster development, it can be fruitfulto look into ANT, and especially the concept of an actant.

2.2. ANT and cluster studies

ANT has gained increased popularity within the social sciencesfor its potential to overcome the dualism of material and discursiveconstructions (Callon, 1999; Henderson et al., 2002; Hetheringtonand Law, 2000; Latour, 1996, 1999; Law, 1999; Murdoch, 1998;Whatmore and Thorne, 1997). It has been argued that ANT offersa more comprehensive view of actors, networks and relations thanthe network approaches dominating the social sciences in recentdecades (Bosco, 2006). It takes into account the multiplicity ofactors in networks and permits the fluidity of multiple types ofrelations and spatial configurations (Fløysand and Jakobsen,2011). Regarding studies of business systems or industrial clusters,there is also some evidence of the rising importance of ANT (see,for instance, Coe et al., 2008; Dicken et al., 2001; Hendersonet al., 2002; Jones and Murphy, 2011; Powell, 2007; Yeung,2000). Yeung (2000) argues that ANT allows for a focus on howthings are ‘‘stitched together’’ across distinctions and how net-works are being reshaped by those with the power to influencethe strategy and activities of others. He uses ANT as a practiceoriented perspective to understand the impact of globalization ten-dencies on Asian business systems, and analyses how these busi-ness systems are being enrolled into, and influenced by, globalnetworks elsewhere. ‘‘Globalization tendencies’’ are identified asthe main agents of change (Yeung, 2000, p. 426). The ideas ofANT have also informed the so-called global production network(GPN) perspective and analyses of the globalizing economy andits impacts on territorial development (Coe et al., 2008; Hendersonet al., 2002). ANT is acknowledged for its relational appreciation,where entities can only be understood through their relationsand connectivity to other entities. Its dynamic understanding ofterritory, as something that is in a constant state of flux, is alsoregarded as favorable. Other studies have used ANT to develop aframework that can provide a coherent and consistent approach

to the study of economical practice (Jones and Murphy, 2011).However, it is argued that ANT lacks an understanding of the struc-tural preconditions and the power relations that shape productionnetworks (Henderson et al., 2002). Thus, these are early daysregarding the use of ANT in studies of business systems and clus-ters, and there is a need for a clarification of why and how ANTcan inform studies of industrial clusters.

In many ways, ANT has been treated as an ontology rather thanas an analytical framework (epistemology) for application inresearch. Thus, when practiced in research, ANT has helped guideresearch projects rather than served as a theoretical framework.We find this understandable but defensive. What is clear is thatANT requires us to study the world by analyzing the heterogeneityof actions (Callon, 1999; Law, 1999). The focus on performance(practice) and effects of relations (networks) is in itself not particu-larly original from the perspective of other approaches. What canmake ANT distinctive from other perspectives is the potential forswitching from actors to actants as the primary study objective. Thisintervention seems to be overlooked in many studies referring toANT (see, for instance, Yeung, 2000) but is of significant importanceif ANT is to make any difference to research on cluster formation. Anactor is normally understood to be an organization or a conscioushuman being, while an actant compromises these actors and muchmore. It is not simply an actor but an integral structural elementupon which meanings are produced and practices unfold.

According to Latour (1996, p. 373), an actant is ‘‘. . .somethingthat acts or to which activity is granted by others’’. This means thatan actant can include human beings, organizations and non-humanentities. Furthermore, ‘‘something’’ can only turn into an actant if itcan be detected as the source of action. As such, actants by defini-tion have a structuring role on actors, but to overcome the dualismof ideas and practice, actants are also described as products ofagency. They are ‘‘performed’’ by and through relations and under-stood to be the effects of relations enacted through networks ofhumans and non-humans. This makes actants not only sources ofaction that guide future practices and systems of relations but alsophenomena constructed through and by practice and systems ofrelations (Latour, 1996). We believe that the notion of an actantallows us to explore ‘‘the heterogeneity of elements which infectthe course of things’’ (Gomart and Hennion, 1999, p. 226). In otherwords, it makes way for an analysis that opens up for elaboratingon becoming over being and hence the processual character ofthe development of an entity, such as a cluster. This includes ana-lyzing feedback loops and the interactive character of the object inquestion (Hess, 2004).

Accordingly, by reading cluster as an actant, there is a possibilityof substituting simplified stage models of cluster developmentwith a more fluid understanding and an emphasis on varietiesand dynamics in cluster configurations. This is important not onlyfor mapping the cluster, but also for understanding how clusters,once established as a representation, affect the very processes thatmaintain and (re)construct them. We will further elaborate onclusters as actants in the section below.

2.3. The cluster as an actant

As already mentioned, the ANT approach encourages us to studythe world from the view-point of the performance and effects of rela-tions. Hence, it can contribute to a clarification of our understandingof cluster formation. The existing cluster literature (see, for instance,Asheim et al., 2006; Atherton and Johnston, 2008; Bathelt, 2005;Enright, 2003; Porter, 2000; Malmberg and Power, 2006) pointstoward some basic elements that are present in a cluster. First, firmsand other actors have to take part in task-solving processes produc-ing relations that in some way or another are constraining andenabling the agency of the co-located firms. Processes leading to

A. Fløysand et al. / Geoforum 43 (2012) 948–958 951

increased competitiveness and innovation capacity serve as anexample. Second, the firms need to collaborate in local networksthrough performances that should also have a strong influence onthe establishment of trust-based learning practices such as the shar-ing of information from the external networks of the firms. Finally,co-located firms have to be associated with particular cognitive cat-egories. Clusters need to be conceptualized both within and aroundthem: they need to be named and narrated. Questions of interesthere are whether and how names are institutionalized, and in thecases where strong representations of clusters are established,how this interferes with the (re)production of network practicesand notions about ‘‘us’’ and ‘‘them’’ and other questions of interestinfluencing the development of the clusters such as the capacity toinnovate.

What seems clear from the discussion above is that cluster for-mation to some extent overlaps with how ANT represents the con-struction processes of actants. The concepts also share an emphasison relational conditions and practice. So far so good, but the coreidea of ANT is that actants represent phenomena that overridethe discursive-materialism dualism. Actants are not only con-structed through materialized performance, but are also orderingsthat affect our ideas regarding how to act when they are main-tained and renegotiated. We believe this is an issue that has beenoverlooked in cluster theory. In the real world, clusters do not existin pure form but continue to develop. The result is that some clus-ters are dominated by SMEs and others by TNCs, some dominatedby research-based knowledge and others by industry knowledge,some closely linked to political initiatives and others more endog-enous, some controlled by local capital and others by FDIs, somerecognized by a specialized knowledge base and others by a strongheterogeneity, and some characterized by renewal and adaptationwhile others are in recession. Cluster formations can also be thesource of representations, and through feedback loops such repre-sentations can be reconstructed even before the entity has reachedthe level of a working cluster. Consequently, clusters as an actantprovide insights into how the idea of a cluster is maintained andrenegotiated through practice.

The remaining question is how such dynamism of clusters interms of changing actors, relations of power and economic perfor-mance can be captured. Malmberg and Power (2006) argue thatthere is value in using a less categorical understanding of clusters,acknowledging the gap between theoretical concepts and practice.As we see it, processes of maintenance and reconstruction of clus-ters can best be analyzed when viewing them as a relational com-plex of (i) multiple actors, (ii) cluster representations (tangible andintangible) and (iii) tasks. Different actors associate various tasksand representations with a given cluster. The tasks to be solvedand the representations that are encouraged will sometimes becontradictory and in conflict, and sometimes overlapping and inharmony. For example, firms in a cluster are normally concernedabout productive and economic tasks. They want their participa-tion in a cluster to contribute to the value creation of their partic-ular firm in the form of increased innovation and profit. Nationalauthorities also depend on clusters as a tool to improve innovationcapacity and to strengthen the international competitiveness of theindustry. While the concern of national governments is the well-being of the national economy, local authorities very often wantclusters to promote regional development and to strengthen theimage of the area (Martin and Sunley, 2003). Public innovationcompanies depend on dynamic clusters as best-practice examplesor blueprints for the development of regions lagging behind, legit-imizing the founding of public cluster programmes. R&D institu-tions look to clusters for industry partners in innovation projects,while university colleges appraise cluster firms as career possibil-ities for graduate students or as recruitment areas for short-termindustry-relevant courses.

Along with such task-solving processes, clusters are establishedas a socio-spatial and territorial category within and outside thecluster through processes of representation. Accordingly, processesof representation are concerned with how clusters are narratedboth in theory (referring to concepts such as tacit knowledge, localcapabilities, and context specific rules of conduct) and in everyday(economic) life by branding clusters using names and symbols. Insummary, our theoretical argument is that a cluster can be ob-served as an actant when it becomes the source of action for thevarious groups of actors involved.

The discussion above has illustrated how cluster formations arelinked to processes of spatial agglomeration, networking and rep-resentations (Enright, 2003; Malmberg and Power, 2006). Onceestablished, clusters are also continually ‘‘performed’’ in a rela-tional complex of actors, representations and tasks giving rise tofeedback loops and changing networks, performance and represen-tation. These are overlapping processes, but in the presentationthat follows, we will keep them separate.

The empirical part of our analysis consists of two cases: theMøre maritime cluster and the Hordaland subsea cluster. The re-search question is about what kind of development paths the clus-ters are following and how material and discursive processes areinterweaved in these development paths. One of the paths illus-trated is development from below, i.e. from materialized clusterpractice into cluster policy-making; another is development fromabove, i.e. from ideas and cluster policy-making into materializedcluster practice. In both cases, the dynamism of cluster develop-ment is triggered through various forms of feedback loops. Onceestablished as a representation, clusters are an actant that affectthe processes that maintain and (re)construct them.

Our empirical analysis draws on a research project funded by theResearch Council of Norway and based at the Institute for Researchin Economics and Business Administration. Fieldwork has beenundertaken in the clusters since 2008, using document analysisand semi-structured interviews. We have carried out interviewswith managers of clustering firms (9 in Møre maritime and 7 inHordaland Subsea) and with cluster facilitators and policymakers(4 in Møre maritime and 4 in Hordaland Subsea). The main themesin the interviews have been networking between firms, processes ofidentification and the role of clusters facilitators and policy makers.The analyses are also based on information from web pages, reportsand evaluations conducted by the cluster facilitators. In addition,we rely on recently conducted surveys in the clusters mapping link-ages and geographical embeddedness of firm practice. In the Møremaritime cluster, the survey data were produced by MøreforskningMolde (Hervik et al., 2007, 2009). For the Hordaland cluster, we relyon a survey we conducted in 2009. By using different registers ofbusiness enterprises and information from the cluster facilitator,we identified 132 firms in the Hordaland subsea cluster. Six of thesefirms reported back that they were no longer active within thesubsea segment, reducing our working population to 126. In ourweb-based survey, we received 74 usable returns, representing aresponse rate of 58.7%.

We commence the case presentation by looking into processesof agglomeration, networking and representation in each of thetwo selected cases. We also discuss feedback loops for clusters inaction (Section 3). In Section 4 the various development pathsand dynamism of the two cases are compared.

3. Clusters and clustering practice

3.1. The Møre maritime cluster

3.1.1. Processes of agglomerationThe origin of the Møre maritime cluster, located in the counties

of Møre and Romsdal in western Norway, dates back to the early

952 A. Fløysand et al. / Geoforum 43 (2012) 948–958

1900s. Since then, the region has developed a rich tradition of ship-building, fishery services and coastal transport with strongco-operative links to fishermen and other local business (Bjarnaret al., 2006). In the 1960s and early 1970s, the fishery sector wascharacterized by major technological innovations and firms inthe Møre area developed a unique ability to adapt to changingmarket conditions. The demand for new fishing vessels eventuallydeclined during the 1970s because of overfishing and reducedstocks. However, with the discovery of oil in the North Sea, newmarket possibilities emerged. The milieu took advantage of thisand specialized in building supply vessels for offshore oil platforms(Halse and Bjarnar, 2011). In 2009, the cluster consisted ofan agglomeration of approximately 200 companies within themaritime sector, including 17 shipping companies, 13 shipyards,12 ship design consultants and 159 equipment suppliers (seeTable 1). Between 2007 and 2009, the number of equipmentproducers grew from 139 to 159, while the number of firms inother sectors was relatively stable (Hervik et al., 2009). The num-ber of man-years was approximately 20,000 in 2008, an increaseof over 30% since 2005. Turnover peaked at 50 billion NOK in2008, reflecting a doubling of the value of market transactionsfrom 2005 (Hervik et al., 2009). Thus, the milieu includes a group ofco-located firms that easily fulfil the criteria of a spatial agglomer-ation of similar and related economic activities.

3.1.2. Processes of networkingThe Møre maritime cluster is also characterized by relations and

interactions of local collaboration and competition. Shipbuilding inthe cluster is characterized by tailor-made end products of highlyunique ships. Such shipbuilding is largely craft-based productionof rather individual solutions. Specific demands and specificationswill be attached to each ship. This makes the production very com-plex, involving a wide range of components, work operations, pro-fessions, suppliers and subsuppliers. The production process has a‘‘fluid’’ character, because the interaction between managers,workers, specialists and suppliers may vary from ship to ship.The low degree of standardization also means that a lot of capitalis tied to each individual product. The production process thus de-mands a well-developed collaboration practice throughout theprocess. The cluster is operating mainly through local/regional net-works. Because of the individuality and craft-based character ofeach newly ordered ship, the construction process has in itself adistinct character of incremental innovation based on extensiveflow of knowledge in day-to-day practical work. Through learn-ing-by-doing, engineers, workers, fishers and sailors discovernew, smart solutions that lead to adaptations and improvementsin construction and equipment. Although some vital technologicaldevelopments have emanated from more top-down verticalrelations, the horizontal networks and flow of knowledge withinregional scaled networks are most important and have also beenessential for taking advantage of global opportunities (Bjarnar,2009).

The local context is also of vital importance in terms of outputrelations. The local production environment and market is of greatimportance, and actors within the different segments deal withdemanding customers within the local territory. Valued in terms

Table 1The structure of the Møre cluster 2008. Source: Hervik et al. (2009).

Numberof actors

Turn-over(billion NOK)

Man-years Result(%)

Shipping companies 17 (14 OSV) 12.4 5960 32.4Shipyards 13 14.2 4750 3.6Ship design consultants 12 0.823 430 37.9Equipment producers 159 22 8750 7.7

of contract prices, the ship consultants in 2007 channelled con-tracts to local firms amounting to 38% of their turnover. This ratiohas been steadily rising over recent years, again indicating thatinteraction between actors has been strengthened (Hervik et al.,2007, p. 12). In 2006, 42% of the value of shipbuilding contractsstemmed from ships delivered to the 14 ship owners in the cluster,versus 22% in 2002. For these 14 companies, this local supply ofnew ships covered over 76% of their total expenditures on newships (versus 62% in 2002). The 12 ship design consultants areamong the key actors acting as a kind of sales corps for the localshipyards and suppliers. In their development of projects, theyinteract extensively with ship owners, shipyards and the supplyindustry. If a project is contracted, the ship design consultantsselect shipyards and equipment producers that co-operate in con-structing a total ‘‘package’’. However, this does not occur withoutcompetition. Only the design consultants who ‘‘package’’ winningcontracts will have their initial costs covered. All in all, the dynam-ics of the Møre cluster are heavily based on formal and informallinks between cluster members, and the regional scale of theinteraction has until recently been remarkably strong.

3.1.3. Processes of representation/identificationThe local embeddedness of the cluster can also be noted when it

comes to how it has been represented among the cluster partici-pants and in joint policy actions. Actors within the cluster stressthat, in particular, horizontal networking has been vital for a feel-ing of belonging. Special value is attached by the informants to thesocial proximity between management and workers, and betweencapital and labor. Labor and especially the management of locallycontrolled firms feel committed to develop the firms in line withnot only the interests of their particular firm, but also the interestsof their community, cluster and region (Bjarnar, 2009).

The clustering environment received formal cluster status and acluster name in 2006 as part of a national initiative to stimulateregional clustering processes in Norway. The cluster became oneof six industrial environments obtaining an NCE status. The NCEprogramme is a public innovation programme run by InnovationNorway offering financial support to internationally oriented clus-ters. After 3 years of operation, the cluster summarizes their exper-tise as follows:

‘‘The Møre maritime cluster on the west coast of Norway isamong the few complete maritime clusters in the world, withseveral companies in each segment of the value chain. The re-gion’s tough climate, and even tougher competition, has turnedthe cluster’s 200 companies into global leaders in the design,building, equipping, and operation of the world’s most innova-tive offshore vessels specializing in advanced marine opera-tions’’ (NCE Maritime, 2009, P.2).This and similar representations have been circulating in differ-ent media:‘‘The Møre maritime cluster is a complete cluster and the worldleader in advanced marine operations. The cluster, consisting ofapprox. 200 companies from all segments of the value chain, isthe largest and most well-established industrial cluster in Nor-way, and was among the first six who received the prestigiousstatus of Norwegian Centre of Expertise, represented by theorganization Norwegian Centre of Expertise—Maritime’’(http://No.wikipedia.org/wiki/NCE_Maritime 27.08.2010) (inNorwegian, translated by the authors).‘‘Norwegian Centre of Expertise is a massive national commit-ment to the most powerful growing industrial clusters in Nor-way. Because of its strong position in the world of advancedmarine operations, the Møre maritime cluster received the Nor-wegian Centre of Expertise—Maritime’’ (http://www.ncemari-time.no/default.aspx?menu=142 27.08.2010) (in Norwegian,translated by the authors).

A. Fløysand et al. / Geoforum 43 (2012) 948–958 953

In the literature, the Møre cluster is widely associated withendogenously driven cluster dynamics and a strong regional iden-tity (Amdam and Bjarnar, 2010; Bjarnar et al., 2006; Hervik andJakobsen, 2001, 2001). This is also clearly expressed in differentrepresentations of the cluster in different media:

‘‘Demanding clients have forced the industry to constantlyadapt and develop more advanced vessels, first for the fishingfleet, and gradually, for the oil industry, which required special-ized and more technologically advanced vessels. Geographicalproximity, good collaboration and communication between alllinks of the value chain—ship owners, design companies,equipment suppliers and shipyards—all play a role. The resultis the world’s leading maritime industry’’ (http://ekstranett.innovasjonnorge.no/templates/Page_Meta____56541.aspx 27.08.2010).‘‘The Sunnmøre region centred in Ålesund is home base for aunique concentration of maritime businesses, as well as special-ized research and educational institutions. Together these formthe Møre maritime cluster, which is among the very few com-plete maritime clusters in the world’’ (http://No.wikipedia.org/wiki/NCE_Maritime 27.08.2010) (in Norwegian, translated bythe authors).‘‘NCE Maritime is committed to facilitating competence in theregion through close cooperation with educational and researchinstitutions, as well as their own projects. NCE Maritime hascontributed to the development of a Master’s Degree Pro-gramme in Product Development and ship design at the Aales-und University College. . . . In December 2008, as an effort tostrengthen further the maritime expertise in the region, NCEMaritime was among the first in the industry to sponsor a pro-fessorship position at the Aalesund University College’’ (http://www.ncemaritime.no/default.aspx?menu=142 27.08.2010) (inNorwegian, translated by the authors).

3.1.4. Feedback loops in operating clustersThe clustering processes described above have turned this

industrial environment into a relational complex that severalactors, representations and tasks are dependent on. Through differ-ent kinds of feedback loops, this cluster construction affects thoseconstituting the cluster. This dynamism includes, for instance, howthe cluster becomes a source of action for policy actors. Thoseresponsible for public cluster programmes in Norway discoveredearly the candidature of the maritime business environment inMøre and Romsdal, when they were looking for geographicallyproximate groups of interconnected companies that fulfil Porter’sconcept of regional clusters. This discovery was linked partly to re-search performed by Møre Research, which introduced a Porter in-spired cluster approach when analyzing developments in themaritime business environment in the 1990s and early 2000s.The research group was reporting more or less yearly from 1998(see, for instance, Hervik and Jakobsen, 2001; Hervik et al., 2009).These reports highlighted the importance of the clustering effectas a competitive advantage for the regional business environment.This stimulated national and regional authorities to approve it asan established cluster, and portrayed it as a prototype when thefirst Norwegian cluster programme (ARENA) was introduced in1999 (see, for instance, Jakobsen et al., 2007).

There also exists other evidence of how actors through variousfeedback loops associate different tasks and representations withthe cluster. The cluster has its own cluster facilitator, through itsstatus as an NCE. The facilitator depends on the cluster as a centerof gravity for connecting global competence networks operated byactors such as Rolls Royce Marine to local institutions such as theAlesund University College. The building of formal competenceand research networks is also supported by the cluster’s hub firms.Moreover, explicit branding projects are carried out advancing the

cluster meta-level action, together with a series of regional, na-tional and international seminars. Interestingly, projects like PyroAS are using the success story of the cluster to promote moreextensive national collaboration within the maritime industry.

Returning to our core concept, we have argued that actants areorderings that affect the practice and relations of their constructorswhen they are maintained and renegotiated. Viewing the cluster as aperforming actant implies that established ideas and representa-tions of the cluster are challenged and reconstructed through vari-ous types of feedback loops. This can be further illustrated in thelight of internationalization and globalization, which at present posea substantial challenge to representations anchoring the industrialcluster to local history and local networks. Two main challengescan be observed. First, time–space compression facilitates a changefrom the former locally embedded networking towards interna-tional networking where innovation increasingly stems from flowof knowledge through global links (Bjarnar, 2009). Second, thischange in the territorial embeddedness of the network practice inthe cluster creates a need for new representations.

Bell et al. (2009) highlight how actors’ choice sets within clustersare negotiated between hierarchical and relational modes of co-ordination, demonstrating that there may be substantial costs in-curred in breaching strong relational traditions. This negotiatedpraxis is coming to the surface in the Møre maritime cluster. Ship-yards build ships for customers all over the world. Most of the con-struction of hulls has been outsourced to Eastern Europeantransition economies. Moreover, Norwegian equipment suppliershave established branch offices and production units abroad. Forexample, STX Europe, formerly Aker Yards, owns shipyards inFrance, Romania, the Ukraine and Vietnam. The Ulstein Group hassubsidiaries in Brazil, China and several European nations. In gen-eral, ship building is increasingly outsourced to partner yards inlow cost countries. In addition, inward internationalization has in-creased substantially in the cluster and has turned global actorssuch as Rolls Royce, ABB, Bourbon, STX Europe and others into ma-jor players.

This spatial reorganization implies changes in the actor–repre-sentation–tasks relations. Although the Møre cluster is still widelyassociated with endogenously driven cluster dynamics and growthprocesses (Amdam and Bjarnar, 2010; Bjarnar et al., 2006; Hervikand Jakobsen, 2001), knowledge flow is altered in many subtleand problematic ways. Local firms and the region are attached tothe global actors’ market networks (Asheim and Isaksen, 2002).The region is also experiencing substantial investment in local edu-cation and centers of expertise by global actors, including regionalresearch groups that are integrated into worldwide webs of re-search institutions (Oterhals et al., 2008). Actors envisage tighterinteraction between flows of tacit and formal knowledge amongcluster-based actors and formal R&D capacity within the globalcompanies. On the other hand, they find that these processes coun-teract what they believe should be the tasks of their cluster. Forexample, the changes in actor–tasks relations mean that manyproducts are developed in house, and know-how is shared lesswithin the regional business environment. The key players in thecluster increasingly develop their own particular design and build-ing programmes, and compete with each other by protectingknowledge and ideas more than hitherto has been the tradition(Bjarnar, 2009). The requirement for formalization (of contractsand relations), reporting routines and formalized decision-makingprocesses increases enormously as endogenous firms interact withthe global actors.

A plausible hypothesis is that the clusters find themselves in acertain momentum in terms of handling these complex sets ofactors, representations and tasks. Interactions between the clus-ter facilitator, education and research institutions, and differentbranches or network organizations give the impression that

954 A. Fløysand et al. / Geoforum 43 (2012) 948–958

actors at diverse levels and within different institutional settingsare advancing the idea of the cluster as a regionally embeddeddriver of innovation. This produces fertile ground for numerousrepresentations of the cluster. For example, the representationof the cluster as a key driver behind ‘‘regional development’’ isco-produced with representation of the cluster as a ‘‘globalizedenvironment’’. The NCE Maritime organization exemplifies thiswhen summing up the cluster under headings such as A world-leading maritime cluster, Global presence—global attraction, and Awinning mix (see Fig. 1). As we see it, these newer representa-tions act as a strategic framework to legitimize internationaliza-tion both for actors moving activity abroad and for attractingnew global actors into the cluster. Under the duress of interna-tionalization, a transformative direction of practice and cognitioncan be observed as firms in the cluster grow beyond the regionalrealm. The cognitive construction and representation of a regio-nal embedded cluster is challenged by practice connected tonot only incoming global actors in the cluster, but also to formerregionally embedded firms going global. We will now considerour second case, the Hordaland subsea cluster, and elaborate onthe processes of agglomeration, networking, representations andfeedback loops in this cluster formation.

3.2. The Hordaland subsea cluster

3.2.1. Processes of agglomerationThe subsea cluster, located in the county of Hordaland in

western Norway, is part of the offshore subsea sector within thepetroleum industry. Development of new technology is a key factorfor cost efficient and safe extraction of oil and gas resources off-shore. During the last two decades, this new technological frontierof subsea wells has been explored. Oil and gas is transportedthrough pipelines on the seabed, and at present, one-third of theexploitation at the Norwegian shelf is through the use of such sub-sea wells. Subsea operations involve different stages, such as devel-opment and fabrication of new technology, operation of thetechnology and maintenance and modification of the technology.In addition, there are specialized services and consultancy directedtowards each of the stages in the production process.

A WORLD-LEADING MARITIME CLUSTER

Most of Norway’s strength in the field of advanced mardrive from the picturesque town of Ålesund on the couimpressive constellation of over 200 leading maritime cinstitutions that form one of very few complete maritim

CENTRE FOR NORWAY’S MARITIME EXPERTISE

Due to its leading positiowas awarded the prestigMaritime by the Norweghi-tech offshore vesselsadvanced offshore fleet

GLOBAL PRESENCE – GLOBAL ATTRACTION

The cluster operates globally and its companies and pthe companies of the cluster are home-bred, an increaas Rolls-Royce Marine, STX or Bourbon, are seeking t

A WINNING MIX

The reason? – the cluster’s winning mix of high pace informal communication, cooperation, and unique knowoperations for the global offshore industry.

Fig. 1. Representation of a cluster on www.ncemaritime.no (reference

Operating within an emerging sector, it can be appropriate todescribe the Hordaland subsea cluster as more precocious thanthe Møre maritime cluster. The latter represents a complete clusterof similar and related economic activities linked through a complexset of networks and relations. Still, the subsea cluster was alsoamong the six industrial environments that achieved the status ofan NCE in 2006. The geographical scope of the cluster was limitedto the county of Hordaland, where Bergen is the main city. How-ever, unlike the Møre cluster, the development of the subseacluster was not rooted in a shared feeling of belonging to a (subsea)cluster among the firms in the industrial environment. This canpartly be explained by differences in how the core activities ofthe firms correlate with subsea activities. Actually, in the processof formulating the application and afterwards, several attemptswere made to define the industrial content of the cluster; for in-stance, by using selected NACE codes (Econ, 2009). Subsequently,a more practically oriented way of limiting (or rather expanding)the number of firms in the cluster was carried out, such that thesubsea cluster consisted of 161 firms in 2009 (Espelid, 2009). Inaddition to the 83 member firms, this number includes firms inthe region operating within the subsea market. These firms weremapped by screening different registers of firms, contactingselected firms and using the knowledge of key informants. Thenumber of man-years in the cluster was then estimated to be4100 in 2008, increasing from 2900 in 2005. The total turnover ofthe cluster, including only the subsea-related activities of the 161firms, was 11.6 billion NOK, more than doubling from 5.3 billionNOK in 2005 (Espelid, 2009).

As already mentioned, we conducted a survey of the subseacluster in Hordaland in 2009. The survey, including responses from74 firms, demonstrates that the cluster consists of a mix of long-established firms that recently have been shifting some or themain part of their operations towards the subsea segment, afterhaving operated within the petroleum industry for a long time,and newly established firms concentrating only on the subsea mar-ket. The main occupation of the Hordaland subsea cluster is main-tenance, modification and operation of subsea installations. One ofthe firms in our survey is an operating company; 11 can becharacterized as providers of systems, technology and services to

ine operations is concentrated within an hour’s ntry’s west coast. The region is home for an ompanies and training, research and finance

e clusters in the world.

n and its strong global presence, the cluster ious status as Norwegian Centre of Expertise – ian government. Over 75% of the world’s large, are designed there. 40% of the world’s most is controlled by the region’s ship owners.

roducts are known world-wide. While most of sing number of major international players, such heir place in the cluster.

of innovation, prototype orientation, competition, -how within the field of advanced marine

: http://www.ncemaritime.no/default.aspx?menu=291 29.08.10).

A. Fløysand et al. / Geoforum 43 (2012) 948–958 955

operating companies; and 50 are subvendors for these system pro-viders. The remaining 12 firms in our survey have ‘‘other roles’’.The subsea cluster also includes financial institutions, public insti-tutions and education and research institutions that to a varyingdegree link their main activities to the subsea segment. These arenot included in our survey.

Even if many of the cluster firms have the subsea market as theirmain market, it is common to combine this with income from othermarket segments. In the survey, only 20% of the firms reported thatthey were operating only within the subsea segment while 26%earned more than 50% of their turnover from the subsea segment(N = 74). This indicates that as a subsea cluster, this group of co-lo-cated firms only partly fulfils the criterion of representing a special-ized agglomeration of similar and related economic activitieswithin the subsea sector. In addition to the subsea market, thisgroup of firms also operates within other parts of the petroleumindustry (74% of the sample), the maritime sector (51%), the marinesector (28%), renewable energy sector (19%) and other sectors(27%). However, as demonstrated in the following paragraphs, theprocesses of networking and collaboration are more extensivewhen we limit our analyses to the subsea activities of theseco-located firms.

3.2.2. Processes of networkingThe dynamics of the Hordaland subsea cluster imply that the

development of new products, services and markets takes placeas an interactive process involving collaboration between differentactors. Our survey reveals a complex set of linkages between firmsoperating subsea installations, firms acting as subvendors for thelatter, and firms providing other types of inputs. When it comesto purchases of inputs related to subsea activities, approximatelyone-third of the firms keep their most important provider ofmachinery and components within the geographical boundariesof the cluster. Not surprisingly, the disposition for using local pro-viders is greater in purchases of services. Cluster purchases accountfor 60% of total purchases in relation to consultancy services, 55%for ICT services and 48% for R&D services. If we look into outputrelations, one-third of the firms are sourcing 50% or more of theirincome from the local market. The survey also finds that approxi-mately 40% of the firms are collaborating with other cluster firmswithin the subsea cluster in the development of new products/ser-vices/markets. Finally, other cluster advantages documented inthe survey relate to how the local industry environment suppliesthe cluster with skilled personnel. Approximately two-thirds ofthe firms report that the county of Hordaland is their main recruit-ment area when employing engineers and researchers (N = 74).

Despite such local linkages, the main spatial pattern of relationsthat can be outlined by the survey data is a multi-scaled networksystem. Table 2 illustrates the balance between local and non-localrelations. The figures reveal that the linkages between firms in thecluster are of importance. However, relations to other segments ofthe petroleum industry, especially within Hordaland, are moreimportant than intra-linkages within a narrow definition of a sub-sea cluster. It can also be observed that firms established before theyear 2000 have developed closer internal cluster relations thanfirms established during the 2000s. This supports an argumentfor viewing cluster construction as an evolutionary and time-con-suming process (Menzel and Fornahl, 2010).

3.2.3. Processes of representation/identificationWhen relating the Hordaland subsea cluster to processes of rep-

resentation, our survey reveals that the feeling of belonging to acluster is still rather modest. Less than one-third of the firms re-ported that they were using their association with the cluster intheir marketing efforts, and only one-quarter believed that externalbusiness partners associated their firm with the cluster (N = 74).

This lack of cluster identification among the members is not some-thing they share with the NCE Subsea organization. On the contrary,the cluster representations of this facilitator largely exemplify howcluster theory is applied to advocate for territorial co-located firmarrangements, manifested in the construction of a center of excel-lence (Enright, 2003; Ffowcs-Williams, 2004; Sölvell et al., 2003).This is expressed on the web page. In advertising the cluster, theregional context in terms of shared history and geographical partic-ularities is curbed, while clustering effects are highlighted. Underthe heading NCE Subsea creates new opportunities, they refer to clus-ter theory, stating that ‘‘International research has demonstratedthat to be part of an industrial cluster . . . is of great importancefor the players involved’’. Furthermore, they seem particularlyfocused on recruiting members, emphasizing that ‘‘NCE Subseacontributes to closer contacts between companies in the subseaindustry’’, and ending the introduction page by stating that ‘‘Thebenefits of joining us as a member are great’’ (http://www.ncesub-sea.no/?page=137&show=163). In this case, the idea of a cluster isactually stronger than materialized cluster practice.

A slightly different representation introduces the cluster to aninternational audience. Under the heading Maintenance – Modifica-tion – Operation, it is concluded that ‘‘The Bergen area in Norwayconstitutes a world leading cluster in subsea technology’’ and that‘‘this formed the basis for the Norwegian government’s appointingof the cluster as a Norwegian Centre of Expertise for subsea technol-ogy’’. It is also stated that the cluster is ‘‘an initiative by the subseaindustry in the Bergen area for the strengthening and international-ization of business, R&D and education’’ (http://eng.ncesubsea.no/?page=254&show=258).

Even if there is a lack of distinct identification among the clustermembers, participating firms are concerned about the develop-ment of the cluster. In our survey, 83% of the firms agree withthe statement ‘‘it is important for our firm that other firms in theSubsea cluster of Hordaland are developing in a favorable way’’(N = 74).

3.2.4. Feedback loops in operating clustersOur second case is interesting in an actant discussion because

what seems to happen is that the order of traditional cluster for-mation processes is reversed. The idea of a subsea cluster is heavilypromoted by local and regional authorities, innovation institutions,and research and education institutions that all rely on Porter’smodel as a tool for promoting national, regional and local compet-itiveness, innovation and growth. In particular, the NCE pro-gramme seems to represent a pivotal factor stimulating clusterformation. It looks like the programme has been launching a clus-ter that at the moment of ‘‘creation’’ did not fulfil the criteria forworking clusters. On the contrary, it seems to be the theoreticaland political ideas linked to the concept of clustering that initiallycontributed most to the creation of a representation of a completeand dynamic subsea cluster.

On the other hand, once this idea of a subsea cluster with all itsadvantages is embraced by actors in the industry, which it cer-tainly has been through the marketing activity of the cluster facil-itator, processes of networking, co-operation and collectiverepresentations seem to be triggered in the subsea cluster. Clusterdevelopment processes are now gaining momentum, and growingnumbers of actors (public authorities, public innovation compa-nies, research and education institutions, industry actors, etc.)are affirming to the idea of a subsea cluster. The cluster as an ac-tant provides feedback on the practices and relations of its con-structors when they are maintained and renegotiated, facilitatingnew measures for cluster development, and new statements ofthe importance of this cluster for the development of the region.

There is evidence of how the idea of a possible cluster can trig-ger cluster-like practices among actors and institutions linked to

Table 2Strength of relations for firms in the Subsea cluster (N = 74). Source: Own survey.

All firms Firms established before 2000 Firms established during the 2000s

Towards other actors in the Subsea cluster in Hordaland 4.24 4.45 4.24Towards actors within the Subsea sector located in other places in Norway 3.96 4.28 3.59Towards actors in other parts of the petroleum industry in Hordaland 4.62 4.85 4.35Towards actors in the petroleum industry located in other places in Norway 4.36 4.72 3.94

Note: Firms reported strength of relations on a scale from 1 (no relations) to 7 (strong relation). The table gives average score.

956 A. Fløysand et al. / Geoforum 43 (2012) 948–958

the cluster, illustrating the interweaving of discursive and materialprocesses in cluster formation. The subsea cluster facilitator re-ports that the number of members in the subsea cluster has morethan doubled since the start up in 2006, and new firms are takingpart in the cluster development projects initiated by the facilitator(www.ncesubsea.no). In an internal evaluation carried out by theNCE Subsea organization in 2010, 60% of the firms reported thatNCE Subsea had facilitated co-operation projects between clus-tered firms (N = 59). Furthermore, 72% of the firms responded thatthe NCE Subsea project had contributed towards knowledgesharing and co-operation between clustered firms and betweenfirms in the cluster and external firms (N = 53). Nearly two-thirdsof the firms also reported that the initiatives of the NCE Subseaorganization had contributed towards innovation through co-oper-ation between clustered firms and R&D institutions (N = 53) (NCESubsea, 2010a). In this recently conducted internal evaluation,some of the firms also elaborated upon increased local networkingamong cluster members because of the initiatives taken by the NCESubsea organization. For instance, they state that ‘‘The meetingpoints are of great value. We have become more families withsome of the other firms. This makes co-operation easier’’ . . . ‘‘In-creased communication among the local firms have opened upnew business possibilities’’ . . . ‘‘We have come closer to otherfirms, it is easier to make contact’’ . . . ‘‘We now pay more attentionto other (local) firms. This opens up new possibilities for co-oper-ation’’ (NCE Subsea, 2010b). In an external evaluation of the orga-nization conducted in 2009, 82% of the firms assessed the efforts ofNCE Subsea as useful or very useful (Econ, 2009, p. 23). In our sur-vey, also conducted in 2009, we found that the majority of thefirms are positive about the cluster-building initiatives carriedout by the NCE Subsea organization (N = 74). In sum, these find-ings, from different sources, point towards an ongoing develop-ment in favor of more cluster-like practices among firms in thesubsea sector in Hordaland. In close collaboration with the indus-try, NCE Subsea was also a driving force in the setting up of aBachelor’s program in underwater technology at the BergenUniversity College, and has also contributed to the establishmentof two new Master’s programmes at the college. Consequently, itseems the subsea cluster as an actant illustrates a reverse processof cluster formation. Legitimization of an initial representation oridea of a dynamic subsea cluster initiates policy and industry prac-tices that in turn strengthen the material and institutional founda-tion of the cluster through various feedback loops.

4. The two cases compared

In summary, our cases illustrate cluster development as a com-plex, relational and fluid process. Viewing a cluster as an actantallows us to overcome the traditional dualism of material and dis-cursive constructions. Our first case, the Møre maritime case, hasseveral characteristics of a complete industry cluster, with a highlevel of networking, knowledge sharing and collective learningamong its members. In the Hordaland subsea cluster, the idea ofthe cluster among policymakers and other stakeholders was ini-tially stronger than the cluster-like practice among the firms and

institutions. However, once established, these clusters are continu-ally ‘‘performed’’ in a relational complex of actors, representationsand tasks giving rise to changing actor networks, new perfor-mances and new representations.

Hence, within each case there are at present at least two mainrepresentations of the cluster. The Møre cluster is associated withendogenously driven cluster dynamics and viewed as a world-leading maritime cluster operating within a globalizing economy.In the subsea case, we can observe a rationality in stressing theinternational dimension of the cluster outwards, while the regionaldimension is being promoted inwards. The explanations for thecoexistence of multiple representations may be many but, in ourcases, they seem to be linked to the fluidity of cluster tasks. Forexample, recent studies within the cluster literature have under-lined the importance of extra-local linkages for cluster develop-ment (Bathelt et al., 2004; Gertler and Wolfe, 2006; Isaksen,2009). These observations have also informed policy practice andguidelines for cluster development in Norway. In both cases, wefind a growing emphasis on how to link the cluster and its partic-ipants to global networks. This process is occurring more easily inthe Hordaland subsea cluster where the local anchoring of the clus-ter is modest compared with the Møre maritime cluster with itslong history of local co-operation and networking. In the latter,the location of a growing number of foreign owned firms withinthe cluster and a spatial rescaling of the network composition oflocal firms seems to be perceived as a threat to the Møre maritimecluster as a genuine regional cluster. In contrast, the main objectiveof the facilitator in the subsea cluster is to assist firms in goinginternational. This is also strongly influenced by the pivotal chal-lenge for the Norwegian petroleum industry to find new marketsabroad at a time when domestic demand is in fluctuation.

The cases also demonstrate that a cluster becomes the source ofaction for policy actors. Porter’s (2000) definition of a cluster as ageographically proximate group of interconnected companies wasan important source of inspiration at the turn of the century whennew policy instruments informed by cluster theory and its emphasison co-location, networking and knowledge flow were introduced inNorway. Porter’s definition is vague when it comes to the geograph-ical scope of the cluster (a city, a region, a state or even neighboringcountries). The solution for Norwegian policymakers was to empha-size a cluster as a regional entity (Jakobsen and Onsager, 2008). Inthis respect, the Møre cluster fits the requirement, partly becausethe research group, which during the 1990s and the early 2000s re-ported on the development of this industrial milieu, applied a Por-ter-inspired regional cluster approach in their analyses (Hervikand Jakobsen, 2001). The Møre cluster was analyzed as a complete(regional) cluster with strong local dynamics and was portrayed asan ideal cluster model by policymakers, exemplified through ARE-NA, the first Norwegian cluster programme (introduced in 1999)(Jakobsen et al., 2007). While the content of the Møre cluster canbe linked to early Norwegian cluster policy, the Hordaland subseacluster has a more recent history. Its composition is linked to therequirements of the second Norwegian cluster programme, theNCE programme (launched in 2006). This programme was intendedto support specialized and internationally competitive clusters withthe best conditions for further growth. Most of the selected clusters

A. Fløysand et al. / Geoforum 43 (2012) 948–958 957

in the NCE programme, with the Møre cluster as an exception, havebeen defined narrowly and include firms within a specific techno-logical area, such as the subsea sector (Isaksen, 2009). To sum up,the cases illustrate how cluster mapping and cluster constructionhave to be viewed as a combination of top-down (cluster theory,cluster policy) and bottom-up processes (the practice of theindustry).

5. Conclusions: rethinking the ANT approach in cluster analyses

This paper has discussed clusters and clustering processes associally constructed phenomena without losing sight of the mate-rial and discursive dynamism of clusters. This was achieved byoutlining how the dualism of materialism and discursive construc-tions in cluster studies is expressed in concepts and terms in thecluster literature. Broadly speaking, definitions of clusters as mate-rial constructions refer to observations of clustering and networkingof co-located firms. Such definitions match up with approachesviewing clustering as a historical bottom-up process. Definitionsof clusters as discursive constructions imply that the idea of a po-tential cluster is in the forefront of the analysis and correspondwith approaches viewing clustering as an instrumental top-downprocess.

We have argued for viewing clusters as an actant, i.e. somethingthat acts or to which activity is granted by others. ANT can be crit-icized for lacking transparency and clarity in the discussion of the-oretical concepts and methodology (Bosco, 2006; Henderson et al.,2002). However, we found ANT, and in particular the concept ofactants, to be a constructive complement to existing cluster theo-ries, because it provides an opportunity to overcome the dichot-omy of material and discursive construction. The concept allowsfor a more fluid understanding, i.e. how clusters affect those consti-tuting the cluster through various forms of feedback loops. Whenanalyzing our cases, we considered processes of spatial agglomer-ation, networking, representations and feedback loops. The pro-cesses of spatial agglomeration refer to the extension of a clusterin time and space, and as such, they link to questions of the geo-graphical embeddedness of clusters. The processes of networkingrefer to the social embeddedness of clusters. A network of firmscan be termed a cluster when it represents a task-solving systemof actors producing relations that in some way or another are con-straining and enabling the agency of its members. Third, clustersare cognitive phenomena. They are conceptualized both withinacademia and in and around the clusters themselves. Accordingly,processes of representation are concerned with how clusters arenarrated both in theory (referring to concepts such as tacit knowl-edge, local capabilities, and context-specific rules of conduct), andin everyday (economic) life by labelling and branding clustersusing names and symbols. Finally, clusters as an actant act backon the processes that maintain and (re)construct them. In anoperating cluster, this gives rise to changing actor networks, neweconomic performance and new representations interweavingboth discursive and material processes.

We focussed on two clusters within the Norwegian NCE pro-gramme to demonstrate our approach. Considering the feedbackloops for clusters as actants, the two cases match nicely with theidea of cluster construction as something that is triggered by ideas,representations, policy and industry practice. The Møre maritimecluster has several characteristics of a bottom-up constructed clus-ter, with a high level of networking, knowledge sharing and collec-tive learning among its members. This also inspired policymakersto portray the Maritime cluster as an ideal or best practice cluster.The Hordaland subsea cluster, on the other hand, expresses a top-down construction process where cluster-like practices have grad-ually gained momentum. As such, it confirms the assumption of

how the ideal world of academics and policymakers can encourageclustering practices among firms. Our discussion demonstrated thatonce a cluster has been promoted and materialized, it affects boththose promoting the cluster and those performing the clusteringpractice. An awareness of such feedback loops is missing and is veryimportant to acknowledge if we are to understand why and in whatways clusters and clustering processes become the source of actionfor cluster stakeholders such as policymakers, academics and busi-ness people. However, it is appropriate to stress that the relationbetween cluster politics and cluster construction is not clear cut.Our analyses have illustrated how cluster-like practices graduallygained momentum in a policy induced cluster initiative, but severalpolicy initiatives have evaporated without encouraging cluster likepractice among the co-located firms.

Policy initiatives have different outcomes and cluster construc-tion follows different paths. Nevertheless, cluster developmentshould be viewed as a process involving both discursive elements(cluster theory, cluster policy, cluster representations) and mate-rial elements (co-location of firms, proximity, cluster labelling).Accordingly, the issue is to not only to map clusters, but also tounderstand how clusters, once established as a representation,act back on the processes that maintain and (re)construct them.Once the idea of clusters has been promoted or materialized, itaffects both those promoting the idea and those actors that consti-tute the cluster. Recognition of these feedback loops is missing inmost of the recent literature on cluster mapping and cluster devel-opment. This article has illustrated how an understanding of clus-ters as an actant can highlight the importance of such fluidity andmultiplicity in cluster construction.

Acknowledgements

The authors are grateful for valuable comments and suggestionsfrom the editor and two anonymous referees. The research wasmade possible by a grant from the Research Council of Norway.

References

Amdam, R.P., Bjarnar, O., 2010. Globalizations and the transformation of clusters: acomparative study of two clusters. Paper Presented at the Regional StudiesAssociation Annual International Conference, Pécs, Hungary, 23–26 May 2010.

Asheim, B.T., Isaksen, A., 2002. Regional innovation systems: the integration of local‘‘sticky’’ and global ‘‘ubiquitous’’ knowledge. Journal of Technology Transfer 27,77–86.

Asheim, B.T., Cooke, P., Martin, R., 2006. The rise of the cluster concept in regionalanalyses and policy: a critical assessment. In: Asheim, B.T., Cooke, P., Martin, R.(Eds.), Clusters and Regional Development. Critical Reflections andExplorations. Routledge, London, pp. 218–235.

Atherton, A., Johnston, A., 2008. Cluster formation from the ‘bottom-up’: a processperspective. In: Karlsen, C. (Ed.), Handbook of Research on Cluster Theory.Edward Elgar, Cheltenham, pp. 93–113.

Bathelt, H., 2005. Geographies of production: growth regimes in spatialperspectives (II) – knowledge creation and growth in clusters. Progress inHuman Geography 29, 204–216.

Bathelt, H., Malmberg, A., Maskell, P., 2004. Clusters and knowledge: local buzz,global pipelines and the process of knowledge creation. Progress in HumanGeography 28, 31–56.

Bell, S., Tracey, P., Heide, J.B., 2009. The organization of regional clusters. Academyof Management Review 34, 623–642.

Bjarnar, O., 2009. Transformation of knowledge flow in globalizing regional clusters.Working Paper 2010: 2, Molde University College.

Bjarnar, O., Berge, D.M., Melle, O., 2006. Havfiskeflåten i Møre og Romsdal ogTrøndelag bind 2: Fra fri fisker til regulert spesialist. Tapir akademisk forlag,Trondheim (in Norwegian).

Bosco, F.J., 2006. Actor-network theory, networks, and relational approaches inhuman geography. In: Aitken, S., Valentine, G. (Eds.), Approaches to HumanGeography. Sage, Los Angeles, pp. 136–146.

Callon, M., 1999. Actor-network theory – the market test. In: Law, J., Hassard, J.(Eds.), Actor Network Theory and after. Blackwell Publishing, Oxford, pp. 181–195.

Coe, N., Dicken, P., Hess, M., 2008. Global production networks: realizing thepotential. Journal of Economic Geography 8, 271–295.

958 A. Fløysand et al. / Geoforum 43 (2012) 948–958

Dicken, P., Kelly, P.F., Olds, K., Yeung, H.W.-C., 2001. Chains and networks, territoriesand scales: towards a relational framework for analysing the global economy.Global Networks 1, 89–112.

Econ, 2009. Evalueringsnotat NCE Subsea. Econ-notat nr. 2009-018 (in Norwegian).Enright, M., 2003. Regional clusters: what we know and what we should know. In:

Bröcker, J., Dohse, D., Soltwedel, R. (Eds.), Innovation Clusters and InterregionalCompetition. Springer, Berlin, pp. 99–129.

Espelid, K., 2009. Verdiskapning på bunn – en økonomisk undersøkelse avsubseaklyngen i Bergensregionen. Prosjektrapport NCE Subsea (in Norwegian).

Ffowcs-Williams, I., 2004. Cluster development: red light and green light. PaperPresented at the conference ANZRSAI’s ‘Sustaining Regions’, 24 May 2004.

Fløysand, A., Jakobsen, S.-E., 2007. Commodification of rural places: a narrative ofsocial fields, rural development, and football. Journal of Rural Studies 23, 206–221.

Fløysand, A., Jakobsen, S.-E., 2011. The complexity of innovation. A relational turn.Progress in Human Geography 35, 328–344.

Gertler, M.S., Wolfe, D.A., 2006. Spaces of knowledge flows: clusters in a globalcontext. In: Asheim, B.T., Cooke, P., Martin, R. (Eds.), Clusters and RegionalDevelopment. Critical Reflections and Explorations. Routledge, London, pp.218–235.

Gomart, E., Hennion, A., 1999. A sociology of attachment: music amateurs, drugusers. In: Law, J., Hassard, J. (Eds.), Actor Network Theory and after. BlackwellPublishing, Oxford, pp. 15–25.

Halse, L.L., Bjarnar, O., 2011. Cluster transformation from a supply chainperspective: Theoretical models and the case of the maritime cluster in Mid-West Norway. Paper Prepared for the Regional Studies Association AnnualInternational Conference, Newcastle 2011.

Henderson, J., Dicken, P., Hess, M., Coe, N., Yeung, H.W.-C., 2002. Global productionnetworks and the analysis of economic development. Review of InternationalPolitical Economy 9, 436–464.

Hervik, A., Jakobsen, E.W., 2001. Det regionale Maritime Norge. Research Report No.8, Handelshøyskolen BI (in Norwegian).

Hervik, A., Oterhals, O., Bergem, B., 2007. Den maritime næringen i Møre ogRomsdal – en vekstkraftig næringsklynge rustet til omstilling? Rapport 0711,Møreforsking Molde (in Norwegian).

Hervik, A., Oterhals, O., Bergem, B., Johannesen, G., 2009. Status for maritime næringergjennom finanskrisen. Rapport 0905, Møreforsking Molde (in Norwegian).

Hess, M., 2004. Spatial relationships? Towards a reconceptualization ofembeddedness. Progress in Human Geography 28, 165–186.

Hetherington, K., Law, J., 2000. After networks. Environment and Planning D:Society and Space 18, 1–6.

Isaksen, A., 2009. Innovation dynamics of global competitive regional clusters: thecase of the Norwegian Centres of Expertise. Regional Studies 43, 1155–1166.

Jakobsen, S.-E., Onsager, K., 2008. Innovasjonspolitikk for regional næringsutvikling.In: Isaksen, A., Karlsen, A., Sæther, B. (Eds.) Innovasjoner i norske næringer. Etgeografisk perspektiv, Fagbokforlaget, pp. 269–288.

Jakobsen, S.-E., Onsager, K., Rokkan, A., Nesheim, T., 2007. Midtveisevaluering avArena-programmet. SNF rapport 01/07. Samfunns og næringslivsforskning,Bergen (in Norwegian).

Jones, A., Murphy, J.T., 2011. Theorizing practice in economic geography: foundations,challenges and possibilities. Progress in Human Geography 35, 366–392.

Latour, B., 1996. On actor network theory: a few clarifications. Soziale Welt 47, 360–381.

Latour, B., 1999. On recalling ANT. In: Law, J., Hassard, J. (Eds.), Actor NetworkTheory and after. Blackwell Publishing, Oxford, pp. 15–25.

Law, J., 1999. After ANT: complexity, naming and topology. In: Law, J., Hassard, J.(Eds.), Actor Network Theory and after. Blackwell Publishing, Oxford, pp. 1–14.

Malmberg, A., Power, D., 2006. True clusters. A severe case of conceptual headache.In: Asheim, B., Cook, P., Martin, R. (Eds.), Clusters and Regional Development.Critical Reflections and Explorations. Routledge, London, pp. 50–68.

Martin, R., Sunley, P., 2003. Deconstructing clusters: chaotic concept or policypanacea? Journal of Economic Geography 3, 5–35.

Menzel, M.-P., Fornahl, D., 2010. Cluster life cycles – dimensions and rationales ofcluster evolution. Industry and Corporate Change 19, 205–238.

Murdoch, J., 1998. The spaces of actor-network theory. Geoforum 29, 357–374.NCE Maritime, 2009. Breaking Waves, Activity Report 2009. NCE Maritime.NCE Subsea, 2010a. Annual 2010. NCE Subsea.NCE Subsea, 2010b. Report from the Company Evaluation. Unpublished Note. NCE

Subsea.Oterhals, O., Hervik, A., Opdal, Ø., Bergem, B., 2008. Utviklingen i maritime næringer

i Møre og Romsdal – Status 2008. Arbeidsrapport M 0802, Møreforsking Molde(in Norwegian).

Porter, M., 1998. Clusters and competition: new agendas for companies,governments and institutions. In: Porter, M. (Ed.), On Competition. HarvardBusiness Review Book, Boston, pp. 197–287.

Porter, M., 2000. Location, competition and economic development: local clusters ina global economy. Economic Development Quarterly 14, 15–34.

Powell, R.C., 2007. Geographies of science: histories, localities, practice, futures.Progress in Human Geography 31, 309–329.

Richardson, C.J.R., 2010. The internationalization of firms in a policy driven industrialcluster: the case of Malaysia’s multimedia super corridor. Thesis submitted toThe University of Manchester for the degree of Doctor of Philosophy.

Sölvell, Ö., Lindqvist, G., Ketels, C., 2003. The Cluster Initiative Greenbook. BrommaTryck, Stockholm.

Whatmore, S., Thorne, L., 1997. Nourishing networks: alternative geographies offood. In: Goodman, D.J., Watts, M.J. (Eds.), Globalizing Food: Agrarian Questionsand Global Restructuring. Routledge, London, pp. 287–304.

Yeung, H.W.-C., 2000. The dynamism of Asian business systems in a globalizing era.Review of International Political Economy 7, 399–433.

Yeung, H.W.-C., Liu, W., Dicken, P., 2005. Transnational corporations and networkeffects of a local manufacturing cluster in mobile telecommunicationsequipment in China. World Development 34, 520–540.