Content uploaded by Marc-Hubert Depret
Author content
All content in this area was uploaded by Marc-Hubert Depret on Sep 24, 2019
Content may be subject to copyright.
Growth and Change. 2019;00:1–24. wileyonlinelibrary.com/journal/grow
|
1
© 2019 Wiley Periodicals, Inc.
1
|
INTRODUCTION
The understanding of clusters has been advanced in the past few years by integrating a dynamic per-
spective through “cluster life cycle” (CLC) approaches (Fornahl & Hassink, 2017; Menzel & Fornahl,
2009; Trippl, Grillitsch, Isaksen, & Sinozic, 2015). Boschma and Fornahl (2011) recognized that “the
existence and structure of clusters can only be understood when studying their dynamics over time”
(p. 1295), highlighting the need for a better understanding of how clusters emerge as well as how and
Received: 8 November 2018
|
Revised: 27 June 2019
|
Accepted: 16 July 2019
DOI: 10.1111/grow.12325
ORIGINAL ARTICLE
Capturing cluster life cycle with a mixed‐method
analysis: Evidence from a French cluster case study
BastienBernela
|
MarieFerru
|
Marc‐HubertDepret
CRIEF EA2249,University of Poitiers,
Poitiers, France
Correspondence
Marc‐Hubert Depret, CRIEF EA2249,
University of Poitiers, UFR Sciences
Economiques Bâtiment A1, 2, Rue Jean
Carbonnier TSA 81100, Poitiers 86073,
France.
Email: marc.hubert.depret@univ-poitiers.fr
Abstract
This article contributes to the growing literature on clus-
ter life cycles (CLC) by demonstrating the opportunities
offered by a mixed‐method approach. The combination of
qualitative and quantitative data is useful both to describe
the evolution of network patterns and to provide an under-
standing of the drivers of CLC. Based on the literature, we
rely on a theoretical background integrating the pre‐exist-
ing context (cognitive, institutional, and social) in which
actors involved in the nascent cluster are embedded, in
order to capture the impact of the genesis period on the
cluster‘s subsequent trajectory. The operationalization of
this mixed‐method focused on one case study highlights the
determining period of the genesis, since CLC is rooted in (a)
the building of interpersonal relationships between actors
from science and industry, (b) a longstanding specializa-
tion in a technological domain, and (c) policy opportuni-
ties and guidelines. The cluster trajectory remains driven by
the initial policies, the cluster‘s nascent specialization and
the founders‘ pre‐existing relationships over time, with an
evolving role of each type of embeddedness over time.
2
|
BERNELA Et AL.
why they grow or decline. While in 2006, Owen‐Smith and Powell indicated “new attention is being
turned to both the origins of regional networks and variations in their capacity with an eye toward
understanding the necessary inputs for cluster formation, as well as the initial conditions that shape
their trajectories,” very few works have yet studied the earlier moment leading to the formation of a
cluster (i.e., before this cluster becomes visible and observable as such). Though the development of
theoretical frameworks and the multiplication of case studies,1
little attention has been paid to the
methodological strategy to capture cluster dynamics and the impact of genesis on the subsequent tra-
jectory of the cluster. Existing works are mainly based on quantitative strategies using econometrics
and/or social network analysis, or qualitative case studies using interviews and resource materials,
but there really is still no room for a third way combining quantitative and qualitative data for CLC
analysis. The choice of the method partially relies on data availability, and if we want to develop a
comprehensive analysis of CLC in order to integrate the context in which local actors are embedded,
we need both longitudinal data, sufficiently precise to decrypt the transition between the various
stages of CLC, and relational data to capture the evolving role of relationships between the cluster
stakeholders (Giuliani, 2011).
We propose to participate to the growing CLC literature by implementing a mixed‐method analysis
on a French cluster to better qualify the evolving drivers of CLC since its genesis. More precisely, the
methodological strategy used relies simultaneously on (a) quantitative data about innovation projects
certified by the cluster (established within the framework of the competitiveness cluster (CC) policy)
from 2006 to 2011 and (b) on qualitative materials ranging from the earlier moment leading to the
formation of the cluster to 2017. Qualitative data are collected through a dozen semi‐structured in-
terviews with the cluster members and governance, and through the 2‐year immersion of an author
within the cluster.
The article is organized as follows. Section 2 develops the theoretical background for a cross‐level
understanding of CLC, including the crucial step of genesis. Section 3 discusses the methodological
issues associated with the CLC approach and suggests the relevance of the mixed‐method to meet our
goals. Section 4 introduces the context of our case study and the research design and data collection
procedure, which enables us to build an original and complete data set. Based on a mixed‐method,
Section 5 provides an in‐depth analysis of the emergence and development of the cluster. Section 6
discusses the results before giving policy implications.
2
|
THEORETICAL BACKGROUND
2.1
|
The overlook of cluster genesis
“More knowledge about how new clusters emerge and why cluster grows in particular place is needed”
(Isaksen, 2016, p. 704). Despite some useful contributions (see notably Feldman & Braunerhjelm,
2006; Frenken, Cefis, & Stam, 2015; Isaksen, 2016; Li, Bathelt, & Wang, 2012; Ter Wal & Boschma,
2011), the crucial stage of genesis is not so represented among the CLC literature. However, this clus-
ter genesis is decisive since the drivers during this stage may differ from those that support its ongoing
growth (Bresnahan, Gambardella, & Saxenian, 2001). This aspect also appears all the more crucial as
it can have an impact on the further stages of the CLC.
Considering the role of history, authors generally agree that “cluster formation is a sequential pro-
cess with an evolutionary logic” (Feldman & Braunerhjelm, 2006, p. 3). The development of evolu-
tionary concepts in economic geography—such as path dependency, lock‐in, capabilities, institutions
matter, learning process, and coevolution—highlighted the heterogeneity of cluster growth patterns,
partly due to diverse initial conditions such as resource endowment and knowledge specialization.
|
3
BERNELA Et AL.
This literature strand also evidenced the relevance of small events to explain cluster emergence: local
industries “can be traced back to some seemingly trivial historical accidents” (Krugman, 1993, p.
35), implying that clusters “start out in a particular location more or less by chance” (Maskell &
Malmberg, 2007, p. 612). Alongside these studies, some authors have increasingly considered cluster
emergence “as a process that is ‘individualistic in nature’” (Li et al., 2012, p. 129), insisting on the
role of entrepreneurship and spinoff creation (Dahl, Østergaard, & Dalum, 2010) as a mechanism
facilitating the process of (high‐tech) cluster formation.
We propose to contribute to this literature by delivering a comprehensive understanding of the
genesis period and by capturing the impact of this period on the cluster‘s subsequent trajectory. In
this paper, cluster‘ genesis refers to the earlier moment leading to the formation of a cluster, integrat-
ing the pre‐existing context (cognitive, institutional, and social) in which the actors involved in the
nascent cluster are embedded (see next sub‐section). CLC is affected by the actors involved and their
interpersonal ties; clusters are embedded in institutions that are deeply structuring; the industry or the
technology in which the cluster is specialized heavily impacts its development. The special attention
dedicated to the actors‘ pre‐existing networks, to the initial technological specialization and to the
institutional context fundamentally roots our approach in an evolutionary perspective in which history
plays an indisputable part. In this perspective, we offer a dynamic framework to break down the clus-
ter‘s trajectory into sequences, paying particular attention to the effect of the actors‘ past decisions on
future choices, while identifying potential points of bifurcation.
2.2
|
The embeddedness of CLC
Geographers have used the concept of embeddedness since the early 1990s (see for instance Grabher,
1993) to explain the evolution and economic success of regions built by locally clustered networks of
firms. We consider indeed clusters‘ evolution is simultaneously and mainly embedded in (a) interper-
sonal ties between actors, (b) specific skills and technological areas, and (c) specific institutional con-
text. This suggests developing a cross‐level understanding of cluster embeddedness that corresponds
to the different contexts that surround cluster genesis and development (Hagedoorn, 2006).2
2.2.1
|
Social embeddedness: Pre‐existing interpersonal ties and evolution of
social networks
New Economic Sociology has acknowledged the embeddedness of economic activities in a social
structure: Granovetter (1973) supported the idea that economic activities depend on interpersonal re-
lationships between actors and called this dependence embeddedness. The relevant level of economic
activity is not only that of companies or organizations in general, but that of individual actors and
their relationships. Ferrary and Granovetter (2009) have shown, for instance, the role of interpersonal
ties between entrepreneurs (former PhDs) and researchers in the emergence of the Silicon Valley.
Similarly, Powell, Koput, and Smith‐Doerr (1996) emphasized the role of relationships between star
scientists and firms in the concentration of R&D partnerships and the clusters of specific technolo-
gies in the U.S. More recently, Ter Wal (2013) noted also in Sofia Antipolis‘ development the more
coherent local network based on inventors‘ relationships. This crucial role of social embeddedness for
CLC analysis has also recently been stressed by authors from relational economic geography (Bathelt
& Glückler, 2003). Following this approach, we argue for the integration of social embeddedness as
a driver in the emergence and further evolution of clusters. Interpersonal ties and partnerships boost
the emergence and development of clusters because they promote (tacit) knowledge sharing and trust
between stakeholders, reduce opportunistic behaviors, and enhance cooperation and collective learn-
ing (Capello & Faggian, 2005; Carli & Morrison, 2018).
4
|
BERNELA Et AL.
In addition to embeddedness, we must also consider decoupling (Grossetti, 2008), that is, the pro-
cess according to which an initial tie between two individuals turns into an institutionalized relation-
ship between two organizations. The embeddedness theory as presented by Granovetter was the object
of a certain amount of criticism related to Granovetter‘s overestimation of interpersonal relationships
in socio‐economic activities, implying that social embeddedness cannot be viewed in isolation: local
actors can at the same time be embedded in institutions.
2.2.2
|
Institutional embeddedness of actors and the evolution of
public policy
The embeddedness must be also considered within a pre‐existing institutional context. Even though
the concept of “institutional thickness” (Amin & Thrift, 1995) recognizes the role of institutions in
economic activities and highlights “local” conditions within economic geography and regional devel-
opment studies, we prefer using the “institutional embeddedness” term to stress the influence of the
environment and the binding or incentive role of institutions, more than just their existence. In this
context, institutional environment can be a strong support to the emergence and development of clus-
ter, by reducing the uncertainty of the actors in the cluster and promoting their access to markets (in-
cluding financial support). It also provides human capital (e.g., engineers, technicians, managers) and
in some cases future entrepreneurs, besides acting also as cluster enabling factors (Carli & Morrison,
2018; Feldman & Braunerhjelm, 2006).
Institutions generally “include formal organizations (training institutes, associations, state agen-
cies, sponsors, banks, etc.), formal standards and regulations, as well as less visible rules, shared
norms and taken‐for‐granted beliefs” (Sydow & Staber, 2002, p. 218). Institutional embeddedness
refers, for instance, to the continuing reliance upon region‘s research organizations in some clusters‘
dynamics such as Harvard and MIT in Boston (Owen‐Smith & Powell, 2006) or Stanford in the
Silicon Valley (Ferrary & Granovetter, 2009). Policies designed at a national level to initiate regional
specialization can also directly support the formation of clusters and their trajectories. A cluster policy
can be defined as “a set of policy interventions aiming at strengthening existing clusters or facilitating
the emergence of new ones” (European Commission, 2008, p. 28) through the design of incentives
fostering collaborative R&D to strengthen knowledge networks and stimulate the expected benefits
of local knowledge spillover and cognitive complementarity (Broekel, Fornahl, & Morrison, 2015).
Indirect cluster policies must also be taken into account: those include past creations of public struc-
tures such as technology and science parks or research universities (Belussi & Sedita, 2009; Paton &
Kenney, 2010). “While cluster promotion policies [as the French CC policy] are unlikely to succeed
in creating clusters ab initio” (Martin & Sunley, 2003), indirect policies can contribute to developing
the clusters‘ premises and support their development (Njøs & Jakobsen, 2016; Uyarra & Ramlogan,
2017). In this regard, Brenner and Schlump (2011) argue that the effects of policy measures differ
depending on the stages of the CLC: investment in education and specialized sectors has the strongest
impact on the expansion phase, whereas public research has a significant role throughout the entire
CLC; promoting and funding start‐ups is more important in the initial phase, while supporting spin‐
offs appears more important in the expansion phase.
2.2.3
|
Cognitive embeddedness: Initial core competencies and
technological evolution
Clusters‘ actors are also embedded in a cognitive environment deeply structuring. Zukin and
DiMaggio (1990) used the concept of “cognitive embeddedness” to refer to “the ways in which the
|
5
BERNELA Et AL.
structured regularities of mental processes limit the exercise of economic reasoning” (p. 15). This
notion calls attention to economic actors bounded rationality and place‐based knowledge (Grabher,
1993, pp. 8–12). At the firm level, given the tacit, idiosyncratic and cumulative nature of knowledge,
the concept sheds light on the role of core competences (Langlois & Robertson, 1995) and absorption
capacity (Nooteboom, 1999). Such a capacity is something that develops over time, is path dependent
and therefore builds on prior knowledge of another organization‘s capacity. When analyzing clusters‘
dynamics, the concept highlights the bounded rationality of local companies and laboratories and
their initial specialization in a specific technological area. Therefore, we consider “the emergence of
new local industry may not be due to chance or historical accident but stimulated and enabled (…) by
preexisting resources, competences, skills, experiences inherited from previous economic activity”
(Simmie & Martin, 2010, p. 33). This path dependency to prior cognitive endowment of a territory
is empirically observed giving rise to the concepts of regional branching or technological relatedness
(Boschma, 2017). While history matters, “path dependency and resources accumulation are part but
only part of the story” (Feldman & Braunerhjelm, 2006, p. 11). The Silicon Valley case indeed sug-
gests that it is “the dynamic process of creating the industry that created the concomitant location
of the institution ingredients and the social relationships which makes them effective” (Feldman &
Braunerhjelm, 2006, p. 1). This cognitive embeddedness encourages emergence and development of
cluster because it strengthens technological or industrial initial specialization and competitive advan-
tage of cluster (Belussi & Sedita, 2009).
Finally, we consider that these three levels of embeddedness do not work in isolation in shaping
clusters dynamics but have possible interactions (Hagedoorn, 2006). Connections between the dif-
ferent levels can reinforce each other. For instance, partnering history between companies can be
simultaneously affected by sectoral background and national or local policies. In this context, it is
crucial to bring history back into the analysis and to take into account the multidimensional nature of
clusters dynamic.
3
|
THE OPPORTUNITY OF MIXED‐METHOD FOR
CLC ANALYSIS
One could roughly say that recent developments in the empirical literature about CLC are based either
on in‐depth qualitative analysis (Belussi & Sedita, 2009; Giuliani, 2011; Shin & Hassink, 2011) or on
social network analysis (Balland, Vaan, & Boschma, 2013; Crespo, Suire, & Vicente, 2016; Giuliani,
Morrison, Rabellotti, & Pietrobelli, 2010). Both methodologies give new insights regarding their own
advantages and limits for the CLC analysis. Qualitative studies—based on interviews, participatory
approaches, case studies, or focus groups—are considered as appropriate for the evaluation of pro-
cesses and contribute to enhancing information about the institutional environment and to a better
understanding of the complexity of the formation process. Nevertheless, these studies rarely integrate
longitudinal data and consider the various potential drivers of CLC, as they mostly focus on a specific
one. Quantitative studies are relevant to describe how clusters evolve, although they are hampered
by their specific interpretation of the network‘s evolution: they can for instance identify actors who
become more and more central in a network, but do not provide any information to justify this cen-
trality reinforcement. Moreover, quantitative studies are faced with the unavailability of data about
the genesis period, since “the emerging cluster is not actually a cluster” (Menzel & Fornahl, 2009,
p. 225). Overall, most studies focus on cluster outputs that are measurable (mainly through collabora-
tive innovation projects) but inexistent until a later stage in the CLC (Feldman & Braunerhjelm, 2006;
Isaksen, 2016).
6
|
BERNELA Et AL.
Data availability appears as one of the main challenges to develop a comprehensive and dynamic
analysis: we need both (a) longitudinal data going back to the very first stage of emergence and suf-
ficiently precise to decipher transitions between stages and the ongoing impact of initial drivers, and
(b) relational data to establish the existing relationships between the various actors of the cluster and
their evolution over time. In this respect, Balland et al. (2013, p. 761) suggest not only to implement a
social network analysis based on collaborative innovation data, but also to conduct “a more qualitative
approach (…) that could deepen our understanding of the motives behind networking and the role of
more informal personal ties.”
Although it has been recommended by Boschma and Fornahl (2011, p. 1297), the deployment of
“different data sources, ranging from the collection of primary data by qualitative research or ques-
tionnaires, to a multitude of secondary data sources” has rarely been used until now (see Ter Wal,
2013). The combination of qualitative and quantitative approaches through a mixed‐method therefore
appears as a promising approach for an in‐depth analysis of CLC.
Initially used in the 1950s in the social sciences, mixed‐method analysis was developed and rolled
out to other human and social sciences in the 1980s. It is used for research projects that involve the
collection, analysis and integration of quantitative, and qualitative data in a single study (Small, 2011).
In short, it is assumed that quantitative and qualitative approaches are not irreducible but complemen-
tary. These two approaches are combined to exploit their respective strengths: statistical and systemic
results but misinterpretation risks on the quantitative side versus “decoding” of complex processes,
behaviors, or trajectories but an illustrative and contextual analysis on the qualitative side (Starr,
2012). The combination of qualitative and quantitative approaches can be done either through trian-
gulation in order to obtain the convergence and verification of findings via different data, or through
the nesting of additional data to deal with different facets of the same subject.
Applied to CLC analysis, the implementation of mixed‐method relies on the collection of existing
quantitative data (actors involved in collaborative innovation projects) combined with historical and
qualitative data. More precisely, we suggest to first carry out a social network analysis from quantita-
tive data to characterize the network‘s properties at the various stages of the CLC, and to then comple-
ment the analysis with a qualitative approach in order to enter the black box of networks. We can thus
interpret statistical trends and decipher the cluster‘s complex trajectory. By reintroducing “real‐life
experience and bibliographical events that leave traces, qualitative data offers more depth and a better
understanding of quantitative data” (De Federico de la Rúa & Comet, 2011, p. 9). This data serves to
take the context into account, to bring content to the analysis of network structure and to explore the
reasons for change (D‘Angelo, Ryan, & Tubaro, 2016).
4
|
CASE STUDY AND RESEARCH DESIGN
4.1
|
Operationalization of a mixed‐method for CLC analysis: Case study of
Elopsys and the French cluster policy
Our methodological proposition is applied to a specific case study: Elopsys, a cluster specialized in
four high technology sectors related to hyperfrequencies— microwaves, photonics, secure networks,
and digital interfaces—and located in the French region of the Limousin. This is a relatively small
region, located in the center of the country. The region‘s scientific productivity is high compared to its
rank in R&D expenditure and patent filing, with a higher level of dependency on public organizations
than the average (Appendix 1). Regional activity in the sector covered by the cluster today accounts
for 18% of the region‘s industrial employment: the regional development agency for the Limousin
identified more than 70 businesses in this sector, employing around 7,500 people encompassing all
|
7
BERNELA Et AL.
levels of qualification. This sector generates more than 25% of the region‘s exports and is based on
extensive public and private research carried out locally. SMEs are over‐represented in this cluster
(Appendix 1).
Elopsys was labeled as a cluster in 2005 in the context of the still ongoing French cluster policy.
In September 2004, the French Inter‐Ministerial Committee on Regional Planning and Development
launched a call for “Competitiveness Clusters” (CC)—Pôles de Compétitivité in French, defined as
a “combination, in a given geographical area,3
of companies, training centers and public and private
research units engaged in a partnership designed to create synergies around common innovative proj-
ects” (https ://compe titiv ite.gouv.fr/en/home-853.html). In July 2005, 71 CCs were set up, each in a
given area of France, around a specific sector (including IT, electronics, biotechnology, wood indus-
try, and clean technology).
Once clusters are officially set up, the national policy carries out a formal evaluation to assess
the performance of each CC and decides whether to renew or not the cluster‘s label: three successive
phases have already been implemented, each characterized by evolving policy guidelines and incen-
tives. During the first phase of the policy (2006–2008), the focus was on the clusters‘ ability to generate
collaborative R&D projects within their ecosystem, within the boundaries of the R&D areas defined
above: clusters had to prove that they could be “project factories.” The second phase (2009–2012)
introduced recommendations for inter‐cluster collaborative projects, which translated in a rise in the
number of projects co‐certified by two clusters or more. Finally, the third phase (2013–2018) added
strong incentives to boost exploitation and markets: clusters are required to become “products of the
future factories,” turning collaborative R&D projects into marketed innovative products and services.
This CC policy agenda is very structuring for the life cycle of Elopsys, to the point that it leads us
to make the distinction between a cluster as a phenomenon and a cluster as an organization (Figure 1).
In 2005, Elopsys is labeled by a national policy, endowed with a dedicated governance and an oper-
ational support team and becomes eligible for national windows to finance innovation projects: the
cluster exists as an organization, independent and recognized as such in the regional and national
landscape. This somewhat ad hoc organization does not emerge from nowhere but from a fertile re-
gional context: although Elopsys did not exist as an official cluster before 2005, it existed before as a
phenomenon and as a local ecosystem under construction around the theme of hyperfrequencies. This
period corresponds to the genesis one, that is, the preconditions leading to cluster formation, before
the cluster becomes visible and observable as such. It can be assumed that the official labeling of the
cluster through the national public policy led to the strategic formulation of the nascent specializa-
tion and initial relationships into a territorial scientific programming. Methodologically, as reported
in Figure 1, the structuring of the network in the form of a CC in 2005 generated the inventory in a
database of innovation projects certified by Elopsys (see next sub‐section), which may be subject to
quantitative processing. Such data are not available for the genesis period, implying that the study of
the preconditions for the cluster emergence is purely qualitative.
FIGURE 1 Capturing cluster life cycle
8
|
BERNELA Et AL.
4.2
|
Data and method
The competitiveness cluster policy provides local incentives to collaborations between public re-
search organizations and companies. Collaborative innovation projects are launched through a two‐
stage procedure. During the first stage, the CC certifies the most promising R&D projects; once
they are first certified, the second stage consists in searching for funds from various sources: (a) the
National Research Agency, (b) governmental funds dedicated to CC projects, (c) European funds, and
(d) regional funds.
We used quantitative data related to the projects certified by the CC from 2006 to 2011 (i.e., from
the start of the policy to the year for which latest data are available). This data comes from the cluster‘s
internal data, which made it possible to obtain exhaustive data on certified projects (not all of them
have been funded, according to the two‐step procedure described above). However, as this data are not
available in open access, we were only able to collect it during the period of immersion in the cluster
carried out by one of the authors of this article. This data includes information about project members
(location, type of institution). To work with fine‐grained data, the authors had to verify the data, and
in particular the location of the plants actually involved in projects, in order to avoid the over‐repre-
sentation of multi‐site companies and large public research organizations.
By the end of 2011, the cluster had acquired around 100 members and certified 250 projects: if
we compare these figures with those for other French CCs (Appendix 1), the cluster is of an average
size and offers a relevant setting to analyze collaboration within clusters. Using this data, we built
a database of the 250 certified projects involving nearly 500 participants, 1,000 participations, and
3,000 bilateral ties. Table 1 shows the quantitative data used on project level, actors (participant/par-
ticipation), and collaborations.
Based on this data, we were able to provide descriptive statistics about the projects (number and
weight of co‐certified projects), participants (distribution by technological domain, location and type
of organization) and partnerships (geography of dyadic relations within projects). This set of variables
and indicators—summarized in Table 2—allows to describe the evolution of other cluster patterns:
Are projects increasingly co‐certified? Are partners increasingly local? Are collaborations increas-
ingly conducted between science and industry? etc. The collected quantitative data also allow to con-
duct a social network analysis (Appendix 2). When computing a network analysis, nodes represent
participants and ties represent dyadic relations between two partners involved in the same collabora-
tive project (even if Bernela and Levy (2017) showed that all partners do not collaborate with each
other inside a project). Projects are represented in one‐mode networks, and single‐partner projects (33
cases of start‐up creations) are not excluded from the analysis as we do not focus on collaboration pro-
cesses but rather on cluster evolution: these are represented by isolated nodes until they participate in
a collaborative project. We characterize the structural properties of the network and the position of ac-
tors, using classical indicators such as average degree, density, and centrality scores (see Appendix 2).
For the qualitative approach, we collected data from two main sources:
1. Immersion of one of the article's authors within the cluster for a 2‐year period (2012–2013),
providing opportunities to take part in the life of the cluster and team discussions, to attend
executive board meetings and steering committee meetings, as well as information and pre-
sentation days. This made it possible to collect crucial information on the cluster's strategic
vision, governance and key actors. We were also able to constantly compare the knowledge
acquired on the development of the cluster with information from the ground.
2. From 2012 to 2017, we conducted semi‐directed face‐to‐face interviews with three successive man-
agers of the cluster about its history, its development and the key issues it faced. These interviews
|
9
BERNELA Et AL.
allowed us to collect unique data on the cluster's genesis and on the periods for which no quantita-
tive data is available (before the creation of the cluster and after the end of the author's immer-
sion in the cluster). Complementary interviews were conducted with nine researchers involved in
thirteen certified projects. These interviews mainly focused on the role played by the cluster in the
formation and implementation of the project. More precisely, for each collaboration studied, we
focused on the origins of the collaboration in order to identify how the partners' networking was
initiated and whether the CC itself or members had a role in this linkage.
Interview data were recorded and transcripted, while data from board meetings were recorded in man-
uscript notes. These textual data are then the subject of a content analysis allowing us to reconstruct the
period prior to the creation of the cluster (period not visible quantitatively); they are also and above all
allowed to give meaning or interpret elements observed quantitatively during the creation and develop-
ment of the cluster (i.e., verification and clarification) thanks to the identification of discourse elements
in the form of verbatim.
5
|
ELOPSYS' LIFE CYCLE
5.1
|
From a cluster as phenomenon to a cluster as organization: A nascent
specialization based on pre‐existing relationships and supported by policies
(genesis‐2005)
We focus here on the earlier moments of the cluster, which is unobservable quantitatively: although
Elopsys is a policy‐driven cluster, its story begins well before 2005. In 2004, a national call for
Competitiveness Cluster was launched by the French government (see above). The Limousin, like
other French regions (decentralized administrative entities), was encouraged to consider cluster
projects.
Historically, the Limousin region has specialized in porcelain from the 18th century, which has
led to the fame of “Limoges porcelain” (Perrier, 1924). It was therefore rather natural for the region
to apply for a cluster in this sector. Despite this pre‐existing local knowledge, the region's political
TABLE 1 Variables under study by scale of analysis
Project
n=250
Actors
Collaboration
n=2,866
Participant Participation
n=471 n=1,140
Co‐certifying
Size
Technological domain
Funds
Renewal (new/former)
Membership
Geography
Type of organization
Note: In gray, the studied variables by scale of analysis.
10
|
BERNELA Et AL.
TABLE 2 Descriptive statistics on projects, actors, and collaborations
Period 2006 2007 2008 2009 2010 2011 Total
Project
Total (cum.) 30 (30) 40 (70) 33 (103) 36 (139) 42 (181) 69 (250) 250
Co‐certified projects (%) 2 (7) 6 (15) 4 (12) 8 (22) 5 (12) 33 (48) 58 (23)
Size (%)
<5 21 (70) 27 (68) 23 (70) 17 (47) 23 (55) 39 (57) 150 (60)
≥5 9 (30) 13 (33) 10 (30) 19 (53) 19 (45) 30 (43) 100 (40)
Average number of par-
ticipants by project
3.8 3.8 4.2 4.5 5.1 4.9 4.5
Technological domain (%)
Microwaves 13 (43) 22 (55) 19 (58) 16 (44) 21 (50) 32 (46) 123 (49)
Photonics 14 (47) 9 (23) 7 (21) 14 (39) 7 (17) 20 (29) 71 (28)
Secure networks 3 (10) 6 (15) 2 (6) 0 (0) 7 (17) 4 (6) 22 (9)
Digital interfaces 0 (0) 3 (8) 4 (12) 4 (11) 7 (17) 10 (14) 28 (11)
Transverse 0 (0) 0 (0) 1 (3) 2 (6) 0 (0) 3 (4) 6 (2)
Funds (%)
Local 6 (20) 8 (20) 2 (6) 4 (11) 5 (12) 5 (7) 30 (12)
CC policy 3 (10) 3 (8) 6 (18) 8 (22) 5 (12) 9 (13) 34 (14)
Research agency 21 (70) 28 (70) 25 (76) 24 (67) 32 (76) 50 (72) 180 (72)
Europe 0 (0) 1 (3) 0 (0) 0 (0) 0 (0) 5 (7) 6 (2)
Participant
New (cum.) 77 (77) 61 (138) 48 (186) 56 (242) 88 (330) 141 (471) 471
Member (%) 16 (21) 8 (13) 3 (6) 4 (7) 4 (5) 9 (6) 44
Geography (%)
Local 19 (25) 13 (21) 4 (8) 6 (11) 9 (10) 15 (11) 66 (14)
National 56 (73) 48 (79) 44 (92) 50 (89) 70 (80) 91 (64) 359 (76)
International 2 (3) 0 (0) 0 (0) 0 (0) 9 (10) 35 (25) 46 (10)
(Continues)
|
11
BERNELA Et AL.
Period 2006 2007 2008 2009 2010 2011 Total
Type of organization (%)
Science 48 (62) 24 (39) 27 (56) 26 (46) 41 (47) 57 (40) 223 (47)
Industry 29 (38) 36 (59) 19 (40) 25 (45) 40 (45) 71 (51) 220 (47)
Other 0 (0) 1 (2) 2 (4) 5 (9) 7 (8) 13 (9) 28 (6)
Participation
Total (cum.) 115 (115) 150 (265) 137 (402) 163 (565) 214 (779) 341 (1,120) 1,120
New (cum.) 77 (77) 61 (138) 48 (186) 56 (242) 88 (330) 141 (471) 471
Formera (cum.) 38 (38) 89 (127) 89 (216) 107 (323) 126 (449) 200 (649) 649
Renewal rateb (%) 33 59 65 66 59 59 58
Member (%) 47 (41) 50 (33) 46 (34) 51 (31) 49 (23) 104 (30) 347 (31)
Geography (%)
Local 50 (43) 56 (37) 46 (34) 51 (31) 52 (24) 95 (28) 350 (31)
National 63 (55) 94 (63) 91 (66) 112 (69) 152 (71) 210 (62) 722 (64)
International 2 (2) 0 (0) 0 (0) 0 (0) 10 (5) 36 (11) 48 (4)
Type of organization (%)
Science 73 (63) 84 (56) 81 (59) 97 (59) 132 (62) 194 (57) 661 (59)
Industry 42 (37) 65 (43) 54 (39) 60 (37) 74 (34) 132 (39) 427 (38)
Other 0 (0) 1 (1) 2 (2) 6 (4) 8 (4) 15 (4) 32 (3)
Collaboration
Total (cum.) 219 (219) 319 (538) 307 (845) 372 (1,217) 588 (1,805) 1,061 (2,866) 2,866
New (cum.) 206 (206) 242 (448) 234 (682) 255 (937) 467 (1,404) 830 (2,234) 2,234
Formera (cum.) 13 (13) 77 (90) 73 (163) 117 (280) 121 (401) 231 (632) 632
Renewal rateb (%) 6 24 24 31 21 22 22
Membership (%)
Member–Member 30 (14) 19 (6) 19 (6) 23 (6) 8 (1) 52 (5) 151 (5)
TABLE 2 (Continued)
(Continues)
12
|
BERNELA Et AL.
Period 2006 2007 2008 2009 2010 2011 Total
Member–Non‐member 89 (40) 107 (34) 123 (40) 139 (37) 180 (31) 358 (34) 996 (35)
Non‐member–Non‐
member
100 (46) 193 (60) 165 (54) 210 (57) 400 (68) 651 (61) 1,719 (60)
Geography (%)
Local–Local 37 (17) 30 (9) 19 (6) 35 (10) 18 (3) 61 (6) 200 (7)
Local–National 86 (40) 110 (35) 121 (39) 109 (29) 153 (26) 241 (23) 820 (29)
Local–International 2 (1) 0 (0) 0 (0) 0 (0) 18 (3) 59 (5) 79 (3)
National–National 93 (42) 179 (56) 167 (55) 228 (61) 377 (64) 456 (43) 1,500 (52)
National–International 0 (0) 0 (0) 0 (0) 0 (0) 6 (1) 171 (16) 177 (6)
International–
International
1 (0) 0 (0) 0 (0) 0 (0) 16 (3) 73 (7) 90 (3)
Type of organization (%)
Science–Science 86 (39) 104 (32) 93 (30) 125 (34) 188 (32) 247 (23) 843 (29)
Industry–Industry 47 (22) 73 (23) 82 (27) 67 (18) 134 (23) 278 (26) 681 (24)
Science–Industry 86 (39) 140 (44) 124 (40) 151 (40) 213 (36) 430 (41) 1,144 (40)
Other 0 (0) 2 (1) 8 (3) 29 (8) 53 (9) 106 (10) 198 (7)
aIf an actor participates in two projects during the same year, we consider that he is new the first time and former the second time, explaining how we can find former partners since the first year.
bRenewal rate=former/total.
The data in (italic) are percentages.
TABLE 2 (Continued)
|
13
BERNELA Et AL.
opposition started a debate about the appropriateness of exclusively supporting this local industry,
claiming there was not only ceramics in the Limousin.
Hyperfrequencies were indeed strongly represented locally given the specialization of an academic
laboratory and of some of the largest local companies (Photonis, Thalès, Anovo, etc.). More precisely,
Ircom—an academic laboratory and Creape—a technology transfer center—played a major role in
the local development of hyperfrequencies by collaborating heavily with academic entrepreneurs,
who were involved in applied research transferable to the industry. This trend was encouraged by a
Science Park (Ester technopole), which was created in Limoges in 1993. This park hosts cutting‐edge
skills, support mechanisms and specialized services, and provides a fertile environment for the eco-
nomic development of the Limousin. A researcher talked about the park's benefits in terms of innova-
tion, noting that “the unity of place and action is convenient since we are able to carry out complete
experiments here.” The Regional Council therefore commissioned a consultancy firm to deliver a
study about a cluster specialized in hyperfrequencies. A researcher from University of Limoges was
also seconded in his position to prepare the “cluster application form.” Finally, the Regional Council
decided to apply for a project based on this high‐tech cluster.
The competitiveness cluster was officially created in July 2005 under the institutional name of
“Elopsys” and its headquarter was located on the Ester science park. A crucial restructuration of
this scientific ecosystem occurred thereafter. First, Creape, the former technology transfer center,
was restructured and merged to become larger and officially recognized as a regional center through
the designation Cisteme (Center for Telecommunications, Electromagnetism and Electronics Systems
Engineering). This institutional restructuring was supported by regional innovation policies. Cisteme
was actively involved in the elaboration of the cluster's collaborative projects, and also acted as a me-
diator between the research sector and the industry in the region.
In parallel, the Xlim lab was officially founded in 2006 through the merger of four labs, including
Ircom: the lab was recognized as a spearhead of research on hyperfrequencies nationally speaking. Its
researchers were very active in the preparation of the application form for Elopsys to the CC policy,
mobilizing their interpersonal relationships with companies and their high level of involvement in
local structures. For instance, the cluster capitalized on the industrial collaborations built up by Xlim,
while “the pre‐projects of the cluster were based on a brainstorm carried out by the Xlim lab on indus-
trial requirements,” as explained by the second head of Elopsys. Consequently, the organization of the
laboratory around six research departments structured the cluster's early technological specialization.
The cluster's founders—who were key individuals initially connected to Xlim, Cisteme, or companies
within the cluster's specialization—encouraged this fertile restructuration of the scientific ecosystem.
This was the case with the cluster's current head and the former head of Cisteme. In this respect, a
researcher stated that, “in my mind I see no difference between Cisteme and Xlim. My contact per-
son is the same for both,” meaning that in the early stages, the individual level matters more than the
organizational.
Regarding the genesis period, we can observe the crucial role of cluster policies: local and national
ones, but also direct and indirect ones (see above). Despite the region's historical specialization, the
regional policy heavily supported the nascent high‐tech capabilities and bet on emerging science–
industry relationships, which had already been favored by the creation of the science park. Cluster
patterns (specialization, concentration, and partnerships) were heavily reinforced. The genesis period
also shows an interaction between the three levels of action: social embeddedness of actors, indus-
trial specialization, and institutional support. The cluster's emergence reaffirms the continued role of
indirect policies (i.e., the institutional restructuration of technological and scientific structures) and
the role of interpersonal relationships of the cluster's founders. This led to the reinforcement of the
cluster's core competencies.
14
|
BERNELA Et AL.
5.2
|
Legitimation from a local standpoint: Evolution of policy
rationales and reinforcement of the pre‐existing relationships (2006–2008)
Once labeled, the main challenge for the nascent cluster was to demonstrate the formal existence of
a dynamic ecosystem of innovation. In this context, we can observe a fast increase of the cluster size
(Table 2): the number of projects was significantly high from the outset (30 projects certified in 2006
and more than one hundred in the first 3years), and the number of actors involved in the cluster grew
rapidly (186 different participants in this period). The current head notes that “this momentum was
only possible because the cluster was able to immediately translate intentions into real projects.” He
also explains that “the Elopsys strategy was to legitimate the concrete existence of the cluster; by
funding these first projects, we proved both its capacity to create collaborative projects thanks to local
resources and partners, and its territorial impacts in terms of innovation and employment.” He finally
adds that “at the end of its first three years of existence, the cluster reached a critical size, allowing it
to meet policy requirements.” In the national cluster assessment, Elopsys was successfully qualified
as a “project factory” and ranked among the 39 CCs (out of 71) that achieved the policy objectives.
This growth was mainly based on pre‐existing fertile science–industry relationships and on the
core scientific competencies detailed in the previous section. A researcher we interviewed confirms
that “the first projects relied on a solid core of previous partners who knew each other prior to the
official creation of the cluster.” Descriptive statistics confirm that the projects certified during this
first phase were based on the pre‐existing ecosystem: (a) these projects focused on microwave and
photonics technologies (90% of total projects during the first years) that form the scientific core of
the cluster, and (b) they were mainly conducted by Elopsys members and local partners. Members
are over‐represented in terms of participation volume compared to participant volume (Table 2): for
example, local partners in innovation projects accounted for 25% of participants in 2006 and for 43%
of participation, meaning that they participated in many projects. This early CLC was characterized
by the very intensive participation of key actors including Xlim teams. In addition, regarding the total
number of projects, the six lab teams are over‐represented since “researchers from Xlim systematically
certify their scientific projects,” as explained by one project leader. The centrality scores throughout
the period highlight the constant importance of the historical actors of Elopsys (i.e., continued pres-
ence of long‐standing actors who renew previous partnerships). The key industrial actors (Appendix
2) confirm the crucial role played by Xlim; these actors all share a joint‐laboratory with the lab: the
Thales research chair, the Alcatel Thales lab after 2004, the Thales Alenia Space after 2006 and the
CEA since 2014. The role of Xlim is also evidenced by the SNA, with isolated nodes (22 nodes) that
partly represent the creation of spin‐off: 12 start‐ups have been established by Xlim researchers within
the studied period (and 11 have been located in the ESTER science park). While this result could at a
first glance be seen as a weakness of the cluster considering that these nodes are isolated, these are in
fact closely linked to the scientific core of the CC through lab spin‐offs. This comment highlights the
relevance of qualitative analysis to shed light on quantitative figures.
The “local dimension” also appears important in this period of growth when considering funding
sources (Table 2): one‐fifth of projects benefited from local funds in the first year (compared to 7%
in the last). The regional administration was strongly committed to this cluster policy in the pursuit of
local development. The nascent cluster turned this interventionist local policy into a winning strategy:
the current head of the cluster considers that “the partnership was deliberately locally oriented at the
beginning and aimed at promoting the territory and generating intellectual property.”
Finally, data indicates that the cluster continued to be research‐driven because of: (a) the weight
of scientific organizations in total participations (around 60% over the period), and (b) the pre-
dominance of project applications to the National Research Agency. This is partly explained by the
|
15
BERNELA Et AL.
central position occupied by Xlim (as each of the six teams and Cisteme were involved in more than
10 projects).
5.3
|
Bifurcation: The era of openness (2009–2012)
Once the local foundations of Elopsys had been laid, the cluster continued to grow during this period:
the number of new projects doubled in comparison with the former period. The acceleration in the
number of certified projects, participants and ties in 2010/2011 led to a decrease in the renewal rate
at the end of the period. In volume, the number of former partners surpassed that of new ones from
2008. The renewal rate reached a peak in 2009 (66%) and then decreased (Table 2). This openness did
not prevent historical key actors from remaining central (Appendix 2). Explaining the formation of a
project during this second period, a researcher notes that “among the nine partners, six new partners
were added to the three historical members of Elopsys.” We can consequently note a tendency toward
an increase in project size (from an average of four partners in phase 1 to an average of five partners in
phase 2) and changes in terms of project funding: while projects were initially based on local funds, at
a later stage they were more frequently developed thanks to national and European funds and through
bigger collaborative projects (which are necessary to apply for European programs). This movement
of openness is observed statistically at various scales: institutional, geographical and technological.
We observe first a tendency toward institutional openness: the proportion of member‐member part-
nerships decreased to reach 5% of total ties at the end of the period (Table 2), reducing the risk of an
institutional lock‐in of the cluster. This trend is partly explained by a growing number of co‐certifica-
tion practices in response to incentives for inter‐clustering in the second phase of the policy. Following
these national guidelines, 50% of projects were co‐certified at the end of 2011 with other French CCs
as PEC (Limousin), S2E2 (Centre), Images & Réseaux (Bretagne) and Minalogic (Rhône‐Alpes).
This strategy of openness was also initiated by some key players of the cluster such as the CTTC
(center for technology transfer in ceramics); in this regard, the head of the cluster mentioned that “it
helped to bring the two CCs of Limoges closer.”
Inter‐clustering incentives led to the geographical openness of Elopsys. We can observe the declin-
ing share of local funds counterbalanced by European projects, which were inexistent earlier (Table
2). Applications for European projects led to a higher share (25%) of international partners at the end
of the second phase. Symmetrically, local actors become the smallest share of actors involved.
Finally, we can observe a tendency toward technological openness: projects specialized in digital
interfaces accounted for around 15% in the second phase while this technology was absent in the first
one. In 2009, against a backdrop of economies of scale, the region prompted Elopsys to integrate
this technology field of e‐design (sector of digital interfaces), whose local activity was driven by the
Regional Council until then. This indirect cluster policy involving restructuration led to major reper-
cussions on the network's structure and specialization: as explained by the head of the cluster, “this re-
structuration led to the formation of a fourth area and therefore to a new growth trend in terms of both
members and employment within the cluster.” In this context, a team from the Xlim laboratory work-
ing on signals, imaging and communication, became a central actor from 2009 onward (Appendix 2).
This tendency toward openness ran counter to local injunctions, demonstrating the role of the vari-
ous levels of public policy. These different incentives created some tensions between local institutions
and industrial actors involved in the cluster. More generally, this stage of development was mainly
structured by cluster policies, both direct and indirect, at the national and regional levels. To meet
these new recommendations, Elopsys relied on the cluster's initial key actors (and their interpersonal
relationships) but new rationales led to deeper changes (i.e., expansion of the network both geograph-
ically and technologically) implying a deviation from the initial trajectory.
16
|
BERNELA Et AL.
We can note that, in this growing period, the network appears increasingly integrated (Appendix
2): its development does not reduce the quality of information dissemination, in terms of network
structural properties: (a) the average degree has risen during the period, revealing that nodes are ever
more connected to each other, (b) the density has naturally decreased as the network has grown, and
(c) the average distance and diameter are however stable over time, highlighting the integrated pat-
terns of the network.
5.4
|
Market logic and new regional opportunities (2013–2018)
According to the successive Elopsys heads, the development of the cluster is driven by the industriali-
zation of the projects implemented since the early 2010s, in coherence with the third phase of the CC
policy (see above): following a national policy recommendation to progress from a logic of techno-
logical development to a market rationale, an increase in industry–industry collaborations and applied
research projects was observed, as noted by the cluster's current head. Elopsys had to shift from being
“project factories” to “product of the future factories”: “we are currently in a phase to secure the mar-
keting of products that have been designed within the projects we have supported.”
In addition, the organization of Elopsys is today impacted by the institutional reform of French
regions (the number of regions decreased from 22 to 13). The Limousin region merged with two other
regions (Aquitaine and Poitou‐Charentes) to become the Nouvelle‐Aquitaine, whose administrative
capital is Bordeaux. “It will be an opportunity to broaden our scope of skills by working in collabora-
tion with CCs from the regions concerned. We are already in contact with them.” Since the interviews,
Elopsys has merged with Route des Lasers, another French CC located in Bordeaux. This merger has
given rise to Alpha‐RLH4
in 2017 and has led to structural changes for the future CLC. This consti-
tutes both the end of the independent trajectory of Elopsys and the beginning of a new CLC (). In this
context of renewal, the cluster's trajectory upstream of the merger could be considered as the genesis
of the new cluster (as an enlarged organization)—and as such, as a determining period for the future
of Alpha‐RLH. More specifically, this policy‐driven reorganization will necessarily lead to a new
tendency toward openness and to an increase in size. For the members, the complementarity of the
two clusters' specializations could lead to improved skills, with greater opportunities in terms of inno-
vation partnerships. Nevertheless, difficulties may arise within this bigger multi‐site cluster regarding
coordination, in particular with its new governance and geographic reorganization.
This last period confirms the existence of a bifurcation in the cluster's initial trajectory: Elopsys
can no longer rely only on its core scientific competencies and its initial network: the new institutional
constraints and opportunities require to integrate into a larger network (i.e., industrials and actors from
Nouvelle‐Aquitaine) than the one than had been built over time.
6
|
DISCUSSION AND CONCLUSION
The goal of this article was to capture the crucial role of the genesis period on the CLC through a
mixed‐method analysis. By applying this mixed‐method to the study of one specific cluster, we have
been able to highlight the key and evolving role of the cognitive, social, and institutional dimensions
in the genesis of the cluster and its subsequent trajectory.
Table 3 summarizes the life cycle of Elopsys and highlights its overall trends. It corresponds al-
most to the one described by Carli and Morrison (2018) in their recent case study: they distinguish
three phases in the CLC which are emergence, take‐off, and renewal.5
|
17
BERNELA Et AL.
Based on interviews with local stakeholders and database about innovation projects certified, we
observe that the genesis of the cluster is rooted in the conjunction of these three types of embedded-
ness. (a) The nascent specialization of the local industry and the specific skills of laboratories appeared
crucial in the cluster's emergence. (b) Nevertheless, this process would not have been enough without
the support of regional public policies (i.e., the local science park and regional funds dedicated to high‐
tech). (c) The cluster's nascent specialization was also deeply based on pre‐existing interpersonal rela-
tionships between academics and local industrials notably based inside the Ester regional science park.
Despite an obvious specialization in porcelain, academics already highly connected with the industrial
sector, decided to defend locally an application for the call for CC in the hyperfrequencies sector, by
mobilizing their personal address book. The emergence of this science‐driven cluster is therefore made
possible by the active intervention of academic gatekeepers: the social embeddedness partly compen-
sates for a regional technological trajectory that did not necessarily predestine the labeling of Elopsys.
Our findings confirm that clusters are characterized by a high level of social embeddedness and that
this one is formulated in a cohesive and rather close social environment in which people exchange
knowledge through informal social networks (Ter Wal & Boschma, 2011). Once certified in 2005, the
cluster focuses its efforts on animating a local community around its technological core (microwaves
and photonics). This is the take‐off period, based on a logic of setting up collaborative innovation
projects (with academic spin‐offs, technology transfer centers, and firms), and generously supported
by public funding (whether regional or national). In three years, Elopsys demonstrated the existence
TABLE 3 Cross‐analysis of embeddedness and CLC
Genesis (before
2005)
Take‐off (phase
1, 2006–2008)
Openness (phase
2, 2009–2012)
Renewal (phase
3, 2013–2018)
Social embeddedness Interpersonal
relationships of
pivot academic
entrepreneurs
Reinforcement of
science–indus-
try partnerships
based on the
solid core of
interpersonal
relationships
National incentives
for inter‐cluster-
ing and growth
of non‐local
partnerships
Decoupling pro-
cess and network
scaling‐up
Institutional
embeddedness
Creation of a
regional science
park (Ester
technopole) in
1994
Restructuration
of labs (Xlim)
and tech-
nological
transfer centers
(Cisteme) to
accompany the
Elopsys strategy
Spin‐off from
Xlim lab
Territorial reform
of French re-
gions in 2016
National call for
Competitiveness
Cluster in 2004
and official
labelling of
Elopsys in 2005
Technological
diversification in
2009 due to an
arbitrage of the
Regional Council
Fusion of Elopsys
with Route des
Lasers in 2017
Cognitive embeddedness Historical spe-
cialization in
porcelain
Focus on the pre‐
existing ecosys-
tem and the core
competencies of
incumbent labs
and firms
New rationale
of national
cluster policy:
from project‐
technology to
product‐market
Development
of research in
hyperfrequencies
18
|
BERNELA Et AL.
of a dynamic scientific and industrial ecosystem around microwaves. In this context, the cognitive
embeddedness strengthens the technological lock‐in of cluster (Hassink, 2010).
Then, once consolidated on a local basis, the cluster will open up geographically (extralocal part-
nerships), socially (new interpersonal ties outside the initial and local network), and technologically
(technological diversification) in response to policy incentives. Finally, the territorial reform and
administrative reorganization of the French regions will somehow force the cluster to renew itself:
Elopsys merges with Route des Lasers (a cluster whose headquarter is in Bordeaux) to form a single
(new and interregional) cluster in the Nouvelle‐Aquitaine region on the theme of lasers in a product‐
market strategy. This phase is characterized by a process of decoupling and scaling‐up of the initial
and local network. In line with other recent studies (Martin & Coenen, 2015; Trippl et al., 2015), this
case study suggests finally that a multiscalar (territorial) approach is needed to take into account the
different factors behind CLC.
These findings call for several comments. The foundations under which the cluster came into
existence have had a strong impact on its evolution (Feldman & Braunerhjelm, 2006; Isaksen, 2016):
finally, we could say that the “cluster as an organization” stems from the “cluster as a phenomenon.”
The cluster's trajectory continues to be driven by the initial policies and is still based on its initial tech-
nological/cognitive specialization and on pre‐existing interpersonal relationships of founders (Belussi
& Sedita, 2009). Nevertheless, the evolution of institutional agendas plays such a structuring role that
it produced a bifurcation in the cluster's initial trajectory (Brenner & Schlump, 2011). Direct national
cluster policies (Martin & Sunley, 2003) had a strong influence by providing recommendations at
each step of the evaluation (incentives for “project factories,” inter‐clustering, and the “product of the
future factories”). Indirect (local/regional) policies (Uyarra & Ramlogan, 2017) also played a signifi-
cant part (a) by heavily financing innovation projects, motivated by local benefits, (b) by constraining
the restructuring of the cluster that led to more technological and cognitive openness, and (c) through
the current institutional context of regional mergers. This is clearly a possible result of the traditional
French spatial planning policy. Our findings recommend also a dynamic approach in understanding
how institutions shape clusters and in considering the evolving rationales for cluster policies.
Last but not least, our results have policy implications. Until now, some authors regret that “little
guidance is provided on the role of policies that are conducive to the formation of clusters, both what
policies to promote and equally important what policy to avoid” (Feldman & Braunerhjelm, 2006).
Developing a comprehensive analysis combining various drivers makes it possible to balance the role
of path dependency created by historical specialization (on which anybody can have an impact) with
the role of deliberate direct policy initiatives. The dynamic perspective adopted shows how import-
ant it is to incorporate a historical view to design effective policies and adapt existing ones. It also
indicates that instruments do not work in isolation (Flanagan & Uyarra, 2016) and therefore that no
single policy is universally applicable: “no blueprints of cluster policies can be given simply because
different contexts require different policies” (Van Klink & De Langen, 2001, p. 454).
Finally, this case study aims to showing that a complex systemic perspective is appropriate to
study the CLC (Pendall, Foster, & Cowell, 2010), and that CLC tends to follow adaptive pathways,
instead of determinist trajectories of CLC (Martin & Sunley, 2011; Shin & Hassink, 2011). The CLC
is therefore a complex empirical question and needs more and more empirical case studies to develop
a comprehensible framework of CLC.
ACKNOWLEDGMENT
The authors would like to thank the anonymous referees and Barney Warf, editor of the journal,
whose detailed remarks and suggestions helped us greatly in refocusing and improving the initial
|
19
BERNELA Et AL.
version of this article. Our many thanks also go to colleagues who made useful suggestions for im-
proving the paper during the Geography of Innovation Conference in Toulouse, and specifically to
Jérôme Vicente (Sciences Po Toulouse). Finally, we are grateful to the Elopsys cluster team for the
time devoted to this research and the provision of data. All errors and omissions remain the authors'
responsibility alone.
ORCID
Marc‐Hubert Depret https://orcid.org/0000-0003-4215-4473
ENDNOTES
1 See notably the book introduced by Feldman and Braunerhjelm (2006) which includes several case studies about cluster
genesis and life cycle.
2 Different typologies of embeddedness have been proposed since Zukin and DiMaggio (1990)—who distinguish social
from structural embeddedness—without obtaining a real consensus (see for instance Hess, 2004).
3 Public incentives for local innovation partnerships, especially between science and industry actors, were consolidated by a
2007 decree that defines R&D areas and stipulates that “the presence of a company within the area entitles it to additional
funding when it participates in collaborative projects approved by the cluster”.
4 Route des Lasers et des Hyperfréquences in French. http://www.alpha-rlh.com/eng.
5 This time decomposition of CLC phases is inspired also by the CLC evolutionary literature (see notably Martin & Sunley,
2006; Belussi & Sedita, 2009).
REFERENCES
Amin, A., & Thrift, N. (1995). Globalisation, institutional ‘thickness’ and the local economy. Managing Cities: The New
Urban Context, 12, 91–108.
Balland, P. A., De Vaan, M., & Boschma, R. (2013). The dynamics of interfirm networks along the industry life cycle:
The case of the global video game industry, 1987–2007. Journal of Economic Geography, 13(5), 741–765. https ://
doi.org/10.1093/jeg/lbs023
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
Belussi, F., & Sedita, S. (2009). Life‐cycle vs. multiple path dependency in industrial districts. European Planning
Studies, 17(4), 505–528.
Bernela, B., & Levy, R. (2017). Collaboration networks within a French cluster: Do partners really interact with each
other? Papers in Regional Science, 96(1), 115–138. https ://doi.org/10.1111/pirs.12170
Boschma, R. (2017). Relatedness as driver behind regional diversification: A research agenda. Regional Studies, 51(3),
351–364.
Boschma, R., & Fornahl, D. (2011). Cluster evolution and a roadmap for future research. Regional Studies, 45(10),
1295–1298. https ://doi.org/10.1080/00343 404.2011.633253
Brenner, T., & Schlump, C. (2011). Policy measures and their effects in the different phases of the cluster life cycle.
Regional Studies, 45(10), 1363–1386. https ://doi.org/10.1080/00343 404.2010.529116
Bresnahan, T., Gambardella, A., & Saxenian, A. (2001). ‘Old economy’ inputs for ‘new economy’ outcomes: Cluster
formation in the new Silicon Valleys. Industrial and Corporate Change, 10(4), 835–860. https ://doi.org/10.1093/
icc/10.4.835
Broekel, T., Fornahl, D., & Morrison, A. (2015). Another cluster premium: Innovation subsidies and R&D collaboration
networks. Research Policy, 44(8), 1431–1444.
Capello, R., & Faggian, A. (2005). Collective learning and relational capital in local innovation processes. Regional
Studies, 39(1), 75–87. https ://doi.org/10.1080/00343 40052 00032 0851
20
|
BERNELA Et AL.
Carli, G., & Morrison, A. (2018). On the evolution of the Castel Goffredo hosiery cluster: A life cycle perspective.
European Planning Studies, 26, 1–18.
Crespo, J., Suire, R., & Vicente, J. (2016). Network structural properties for cluster long run dynamics: Evidence from
collaborative R&D networks in the European mobile phone industry. Industrial and Corporate Change, 25(2),
261–282.
Dahl, M. S., Østergaard, C. R., & Dalum, B. (2010). Emergence of regional clusters: The role of spinoffs in the early
growth process. In R. Martin (Ed.), The handbook of evolutionary economic geography (pp. 205–221). Cheltenham,
UK: Edward Elgar Publishing.
D’Angelo, A., Ryan, L., & Tubaro, P. (2016). Visualization in mixed‐methods research on social networks. Sociological
Research Online, 21(2), 15–30. https ://doi.org/10.5153/sro.3996
De Federico de la Rúa, A., & Comet, A. (2011). Personal networks, social networks. Bulletin of Sociological
Methodology, 110(1), 5–10.
European Commission. (2008). The concept of clusters and cluster policies and their role for competitiveness and inno-
vation: Main statistical results and lessons learned. Luxembourg: Office for Official Publications of the European
Communities. https ://doi.org/10.2769/67535
Feldman, M., & Braunerhjelm, P. (2006). The genesis of industrial clusters. In P. Braunerhjelm & M. Feldman (Eds.),
Cluster genesis. Technology‐based industrial development (pp. 1–13). Oxford, UK: Oxford University Press.
Ferrary, M., & Granovetter, M. (2009). The role of venture capital firms in Silicon Valley’s complex innovation network.
Economy and Society, 38(2), 326–359. https ://doi.org/10.1080/03085 14090 2786827
Flanagan, K., & Uyarra, E. (2016). Four dangers in innovation policy studies—And how to avoid them. Industry and
Innovation, 23(2), 177–188. https ://doi.org/10.1080/13662 716.2016.1146126
Fornahl, D., & Hassink, R. (2017). The life cycle of clusters: A policy perspective. Northampton, MA: Edward Elgar
Publishing.
Frenken, K., Cefis, E., & Stam, E. (2015). Industrial dynamics and clusters: A survey. Regional Studies, 49(1), 10–27.
https ://doi.org/10.1080/00343 404.2014.904505
Giuliani, E. (2011). The role of technological gatekeepers in the growth of industrial clusters: Evidence from Chile.
Regional Studies, 45(10), 1329–1348. https ://doi.org/10.1080/00343 404.2011.619973
Giuliani, E., Morrison, A., Rabellotti, R., & Pietrobelli, C. (2010). Who are the researchers that are collaborating with
industry? An analysis of the wine sectors in Chile, South Africa and Italy. Research Policy, 39(6), 748–761. https ://
doi.org/10.1016/j.respol.2010.03.007
Grabher, G. (1993). Rediscovering the social in the economics of inter‐ firm relations. In G. Grabher (Ed.), The embed-
ded firm: On the socio‐ economics of industrial networks (pp. 1–31). London, UK: Routledge.
Granovetter, M. (1973). The strength of weak ties. American Journal of Sociology, 78(6), 1360–1380.
Grossetti, M. (2008). Proximities and embeddings effects. European Planning Studies, 16(5), 613–616.
Hagedoorn, J. (2006). Understanding the cross‐level embeddedness of interfirm partnership formation. Academy of
Management Review, 31(3), 670–680. https ://doi.org/10.5465/amr.2006.21318924
Hassink, R. (2010). Locked in decline? On the role of regional lock‐ins in old industrial areas. In R. Boschma & R.
Martin (Eds.), Handbook of evolutionary economic geography (pp. 450–468). Cheltenham, UK: Edward Elgar.
Hess, M. (2004). Spatial’ relationships? Towards a reconceptualization of embeddedness. Progress in Human
Geography, 28(2), 165–186. https ://doi.org/10.1191/03091 32504 ph479oa
Isaksen, A. (2016). Cluster emergence: Combining pre‐existing conditions and triggering factors. Entrepreneurship &
Regional Development, 28(9–10), 704–723. https ://doi.org/10.1080/08985 626.2016.1239762
Krugman, P. R. (1993). Geography and trade. Cambridge, MA: MIT Press.
Langlois, R., & Robertson, P. (1995). Firms, markets and economic change. London, UK: Routledge.
Li, P., Bathelt, H., & Wang, J. (2012). Network dynamics and cluster evolution: Changing trajectories of the aluminium
extrusion industry in Dali, China. Journal of Economic Geography, 12(1), 127–215. https ://doi.org/10.1093/jeg/
lbr024
Martin, H., & Coenen, L. (2015). Institutional context and cluster emergence: The biogas industry in Southern Sweden.
European Planning Studies, 23(10), 2009–2027. https ://doi.org/10.1080/09654 313.2014.960181
Martin, R., & Sunley, P. (2003). Deconstructing clusters: Chaotic concept or policy panacea? Journal of Economic
Geography, 3(1), 5–35. https ://doi.org/10.1093/jeg/3.1.5
Martin, R., & Sunley, P. (2006). Path dependence and regional economic evolution. Journal of Economic Geography,
6(4), 395–437. https ://doi.org/10.1093/jeg/lbl012
|
21
BERNELA Et AL.
Martin, R., & Sunley, P. (2011). Conceptualizing cluster evolution: Beyond the life cycle model? Regional Studies,
45(10), 1299–1318. https ://doi.org/10.1080/00343 404.2011.622263
Maskell, P., & Malmberg, A. (2007). Myopia, knowledge development and cluster evolution. Journal of Economic
Geography, 7(5), 603–618. https ://doi.org/10.1093/jeg/lbm020
Menzel, M. P., & Fornahl, D. (2009). Cluster life cycles—Dimensions and rationales of cluster evolution. Industrial and
Corporate Change, 19(1), 205–238. https ://doi.org/10.1093/icc/dtp036
Njøs, R., & Jakobsen, S. E. (2016). Cluster policy and regional development: Scale, scope and renewal. Regional
Studies, 3(1), 146–169.
Nooteboom, B. (1999). Innovation, learning and industrial organisation. Cambridge Journal of Economics, 23(2), 127–
150. https ://doi.org/10.1093/cje/23.2.127
Owen‐Smith, J., & Powell, W. W. (2006). Accounting for emergence and novelty in Boston and Bay area biotechnology.
In P. Braunerhjelm & M. P. Feldman (Eds.), Cluster genesis. Technology‐based industrial development (pp. 61–83).
New York, NY: Oxford University Press.
Paton, D., & Kenney, M. (2010). The role of the university in the genesis and evolution of research‐based clusters. In D.
Fornahl, S. Henn, & M. P. Menzel (Eds.), Emerging clusters: Theoretical, empirical and political perspectives on
the initial stage of cluster evolution (pp. 214–238). Cheltenham, UK: Edward Elgar.
Pendall, R., Foster, K. A., & Cowell, M. (2010). Resilience and regions: Building understanding of the metaphor.
Cambridge Journal of Regions, Economy and Society, 3(1), 71–84. https ://doi.org/10.1093/cjres/ rsp028
Perrier, A. (1924). Limoges: étude d’économie urbaine. Annales de Géographie, 33(184), 352–364.
Powell, W., Koput, K., & Smith‐Doerr, L. (1996). Interorganizational collaboration and the locus of innovation: Networks
of learning in biotechnology. Administrative Science Quarterly, 41(1), 116–145. https ://doi.org/10.2307/2393988
Shin, D. H., & Hassink, R. (2011). Cluster life cycles: The case of the shipbuilding industry cluster in South Korea.
Regional Studies, 45(10), 1387–1402. https ://doi.org/10.1080/00343 404.2011.579594
Simmie, J., & Martin, R. (2010). The economic resilience of regions: Towards an evolutionary approach. Cambridge
Journal of Regions, Economy and Society, 3(1), 27–43. https ://doi.org/10.1093/cjres/ rsp029
Small, M. (2011). How to conduct a mixed-methods study: Recent trends in a rapidly growing literature. Annual Review
of Sociology, 37, 57–86. https ://doi.org/10.1146/annur ev.soc.012809.102657
Starr, M. (2012). Qualitative and mixed‐methods research in economics: Surprising growth, promising future. Journal
of Economic Surveys, 28(2), 238–264. https ://doi.org/10.1111/joes.12004
Sydow, J., & Staber, U. (2002). The institutional embeddedness of project networks: The case of content production in
German television. Regional Studies, 36(3), 215–227. https ://doi.org/10.1080/00343 40022 0122034
Ter Wal, A. L. (2013). Cluster emergence and network evolution: A longitudinal analysis of the inventor network in
Sophia‐Antipolis. Regional Studies, 47(5), 651–668. https ://doi.org/10.1080/00343 40100 3614258
Ter Wal, A. L., & Boschma, R. (2011). Co‐evolution of firms, industries and networks in space. Regional Studies, 45(7),
919–933. https ://doi.org/10.1080/00343 40080 2662658
Trippl, M., Grillitsch, M., Isaksen, A., & Sinozic, T. (2015). Perspectives on cluster evolution: Critical review and future
research issues. European Planning Studies, 23(10), 2028–2044. https ://doi.org/10.1080/09654 313.2014.999450
Uyarra, E., & Ramlogan, R. (2017). The Impact of Cluster Policy on Innovation. In J. Edler, P. Cunningham, A. Gok, &
P. Shapira (Eds.), Handbook of Innovation Policy Impact (pp. 196–238). London: Edward Elgar.
Van Klink, A., & De Langen, P. (2001). Cycles in industrial clusters: The case of the shipbuilding industry in
the Northern Netherlands. Tijdschrift Voor Economische in Sociale Geografie, 92, 449–463. https ://doi.
org/10.1111/1467-9663.00171
Zukin, S., & DiMaggio, P. (1990). Structures of capital: The social organization of the economy. New York, NY:
Cambridge University Press.
How to cite this article: Bernela B, Ferru M, Depret M‐H. Capturing cluster life cycle with a
mixed‐method analysis: Evidence from a French cluster case study. Growth and Change.
2019;00:1–24. https ://doi.org/10.1111/grow.12325
22
|
BERNELA Et AL.
APPENDIX 1: Positioning of the Limousin and Elopsys in comparison with other
French regions and CCs
Limousin [ranking]
French average [excl.
Ile‐de‐France]
Total R&D expenditure in 2009 (per
inhabitant)
219.22 € (100%) [18th/22] 650.67 € (100%) [472.81 € (100%)]
Private R&D expenditure in 2009 (per
inhabitant)
132.49 € (60.4%)
[18th/22]
415.39 € (63.8%) [297.82 € (63%)]
Public R&D expenditure in 2009 (per
inhabitant)
83.73 € (39.6%) [17th/22] 235.29 € (36.2%) [174.99 € (37%)]
Number of patents filed in 2009 (per mil-
lion inhabitants)
60.92 [21st/22] 174.29 [143.95]
Number of scientific publications pub-
lished in 2009 (per million inhabitants)
498.22 [11th/22] 736.83 [612.50]
Elopsys
Average of the 70
other French CCs
Companies involved (in 2012)
Number of member companies 64 134
Including SMEs 52 (81.3%) 89 (66.3%)
Number of employees (including executives) 3 369 (435) 11 892 (4,046)
Export performance (in 2012)
Export rate of member companies 33% 24%
Export rate of member SMEs 40% 26%
Source: Eurolio dashboard (https://eurolio.univ-st-etienne.fr/) & DGCIS annual survey of clusters (https://competitivite.gouv.fr/
les-56-poles/tableaux-de-bord-statistiques-256.html.
APPENDIX 2: Network analysis over time
2006 2006–2007 2006–2008 2006–2009 2006–2010 2006–2011
Structural properties
Number of
nodes
77 138 186 242 330 471
Number of iso-
lated nodes
4 10 11 10 14 22
Number of ties 206 448 682 937 1,404 2,234
Total number of
components
3 1 1 2 1 1
Size of the big-
gest compo-
nent (%)
49 (63.6%) 128 (92.7%) 175 (94.1%) 227 (93.8%) 316 (95.8%) 449 (95.3%)
Average degree 5.4 7.6 8.9 9.9 10.9 12.2
Density 0.071 0.055 0.048 0.041 0.033 0.026
Average
distancea
2.5 3.0 2.9 2.9 3.0 2.9
(Continues)
|
23
BERNELA Et AL.
2006 2006–2007 2006–2008 2006–2009 2006–2010 2006–2011
Diameter (max.
geodesic
distance)a
4 6 5 5 6 6
Degree centrality
Xlim‐OSA
Cisteme
Xlim C2S2
Xlim‐Minacom
Xlim‐
Photonique
CEA
France Telecom
Xlim‐DMI
Alcatel‐Thalès
Lab.
Alcatel‐Lucent
Xlim‐SIC
INRIA
Closeness centrality
Xlim‐OSA
Xlim‐Minacom
Cisteme
INRIA
France Telecom
CEA
Thalès Com.
Xlim‐
Photonique
Xlim‐SIC
Alcatel‐Thalès
Lab.
Xlim‐C2S2
Radiall Systems
Telecom
Bretagne
Supelec
IETR
Betweenness centrality
Xlim‐OSA
Xlim‐Minacom
(Continues)
Appendix 2 (Continued)
24
|
BERNELA Et AL.
2006 2006–2007 2006–2008 2006–2009 2006–2010 2006–2011
Xlim‐
Photonique
Xlim‐C2S2
Xlim‐DMI
CEA
Cisteme
France Telecom
INRIA
Xlim‐SIC
ENST
Thalès Com.
aIn the biggest component.
We report in this table the actors which are among the top‐ten central in at least two periods.
Appendix 2 (Continued)
