1
Clusters, structural embeddedness, and knowledge: A structural embeddedness
model of clusters
Shaowei He
Centre for Urban and Regional Studies
University of Birmingham, U.K.
Email:
sxh219@bham.ac.uk
-Paper to be presented at the DRUID-DIME Winter PhD Conference, Skoerping
, Denmark, 26
th
– 28
th
January, 2006
Abstract: The paper is concerned with network structure and its relation to clusters
development with reference to knowledge creation and diffusion. Consciously trying
to avoid ‘sloppy thinking’, it revisits the concept of ‘structural embeddedness’ and
uses it as the analytical lens. The paper demonstrates that both cohesive internal
linkages or ‘closure’ and diverse external linkages or ‘range’ are important for
clusters development. While ‘closure’ is important for fine-grained information
transmission and action coordination, ‘range’ brings into the cluster novel information
and knowledge and therefore prevents the cluster being locked-in. As clusters and
firms face both uncertainties in knowledge access and in action coordination, the
network structure needs to strike an appropriate balance between ‘closure’ and ‘range’.
The specific network structure of a cluster depends on the specific uncertainties it
faces. When the prominent uncertainty is in accessing relevant knowledge, the
balance may need to lean towards ‘range’. When the greatest uncertainty is in
securing the coordination, the balance may need to lean towards ‘closure’. Finally, the
paper shows that the appropriate mix of ‘range’ and ‘closure’ is contingent on clusters
difference and cluster life cycles, which influence the uncertainties facing a cluster.
Key words: clusters, knowledge, structural embeddedness, internal linkages, external
linkages, range, closure,
2
1. Introduction
Arguably cluster theory has evolved towards an embeddedness approach,
emphasizing the social context in knowledge creation and dissemination and
innovation. Consequently, the analysis of networking relationships has become the
core of various accounts of clusters. However, the arguments in current literature are
to a large extent biased toward local embeddedness or geographic proximity for
promoting and sustaining the kinds of relationship necessary for innovation. It has
been noticed that a good part of the existing literature looks at clusters as isolated and
self-contained entities (Breschi & Malerba, 2001). In particular, the growing emphasis
of tacit knowledge and collective learning process helps to legitimize the arguments
such as ‘localized capabilities’ (Maskell and Malmberg, 1999) and regional-specific
assets, for example ‘untraded interdependencies’ (Storper, 1995), which are based
upon intra-regional linkages. While there is recognition of the importance of external
relationships and linkages, especially where clusters are export-oriented, the nature of
extra-regional linkages is ‘characterized explicitly or implicitly as arm’s-length’ and
the external world is described at most ‘as a market presenting competitive challenges
that must be met through improved organization and effort within the cluster’
(Humphrey & Schmitz, 2002: 1019).
This bias towards intra-regional linkages might be explained by the interest in the
‘new regionalism’ (Lovering, 1999; MacLeod, 2001). In the ‘new regionalism’, it is
argued that most of the essential determinants of economic performance reside in
regions which are characterized as “containing the full range of activities required to
produce finished products for the world market, or at the very least retaining the core
functions’ (Humphrey & Schmitz, 2002: 1019). By presuming the importance of the
‘local’ or the ‘region’, innovation is viewed ‘as a result of social processes which
depend on close interaction and network linkages in localized production contexts’
(Bathelt, 2003: 769-770). Consequently, much of earlier work ends up with ‘find[ing]
indicators that conforming this’ (Wolfe & Gertler, 2004).
However, it has been pointed out how problematic this ‘new regionalism’ is and how
rare it is to find a self-referential region, as regions are strongly dependent on national
3
institutions and other external influences and lack important political decision-making
competencies (Lovering, 1999; Hess, 2004; Bathelt, 2003;). In addition, numerous
empirical studies have demonstrated that external or global linkages are as much
important, if not more than, as internal linkages in the innovation processes (for
instance, Hendry et al, 2000; Saxienian & Hsu, 2001; Britton, 2003; Simmie, 2004;
Leibovitz, 2004).
Despite the well-documented importance of external linkages in innovation and
economic development (Grabher, 1993; Simmie, 2004) and a few recent attempts to
conceptualize them into cluster models (Wolfe & Gertler, 2004; Bathelt et al., 2004),
they are still ‘weakly theorized’ in the current cluster literature (Humphrey & Schmitz,
2002). There still lacks systematic and dynamic conceptualization of ‘embeddedness’
and ‘structural embeddedness’ in particular. If we accept that ‘structural
embeddedness’ is the structure of overall networking relations (Granovetter, 1992)
and define it as the structure of internal linkages and external linkages in the case of
‘clusters’, then the questions we need to ask are: What are the functions of internal
linkages and external linkages in knowledge creation respectively? What is the
relationship between these two kinds of linkages? What is the structure of the two
linkages in clusters and what are its impacts on knowledge flow and innovation
(Staber, 1996; Tracey & Clark, 2003).
It is the objective of this paper, based on the deficit in the literature, to develop a
systematic and dynamic analysis of clusters, using ‘structural embeddedness’ as the
theoretical analyzing lens. The arguments moves as follows: The next section revisits
the concepts of ‘embeddedness’ and ‘structural embeddedness’ in the economic
sociology literature (where they originated), with the aim to avoid ‘fuzzy’
conceptualization of ‘embeddedness’ in regional sciences (Hess, 2004). In section 3, a
structural embeddedness model of clusters is presented, discussing the functions of
both internal and external linkages and the relationship and structure of the two
linkages in regard to knowledge creation and diffusion. Section 4 moves on to discuss
the contingent factors that influence the ‘optimal’ structure of internal and external
linkages for clusters. The final section summarizes the argument of the paper and
presents its implications for policy.
4
2. A brief introduction to embeddedness and structural embeddedness
The concept of ‘embeddedness’ has gained much prominence in regional
development studies including works in clusters. However, it is criticized that the
concept has been applied with ‘spatial fetishization’ (Lewis et al., 2002: 441) and
‘new regionalism’ is almost conceived as ‘the only spatial logic of embeddedness in
an era of globalization’ (Hess, 2004: 167). Particularly, as ‘many studies in the new
regionalism tradition pay attention almost exclusively to local and regional systems of
economic and social relations’ (Hess, 2004: 166) and overlooked the non-local
relations, the ‘structural embeddedness’ elements, which is emphasized much by
Granovetter (1992) is to a large extent neglected in regional studies until recently. It is
therefore worth pausing for a while to revisit the concepts of ‘embeddedness’ and
‘structural embeddedness’ and the basic ideas behind them.
2.1 embeddedness and structural embeddedness
According to Granovetter (1985, 1992), embeddedness refers to the fact that
‘economic action and outcomes, like all social action and outcomes, are affected by
actors’ dyadic (pairwise) relations and by the structure of the overall network of
relations’ which he refers to relational embeddedness and structural embeddedness
respectively (Granovetter 1992: 33). Therefore relational embeddedness reflects the
nature of dyadic (pairwise) relations (through strong ties or weak ties), which has
direct effects on individual’s economic action, while structural embeddedness goes
beyond the immediate ties and refers to the aggregated impact of dyadic relations.
By stressing structural embeddedness, Granovetter argues that not only do personal
relations matter, but also the structure of the overall network of relations. If the pair of
individual relations is abstracted out of the wider social context, the ‘atomization’ of
human behaviour ‘has not been eliminated, merely transferred to the dyadic level of
analysis’ (Granovetter, 1992: 34, emphasis original). The overall networks of
relations has subtle and indirect impact on human behaviour. Granovetter (1992)
explains this well:
“A worker can more easily maintain a good relationship with a supervisor who has good relations with
5
most other workers as well. If the supervisor is at odds with the others, and especially if those others
are friendly with one another, they will be able to make life very difficult for the one worker who is
close to the supervisor; pressures will be strong to edge away from this closeness’. (Granovetter, 1992:
35)
There is a general agreement that social structure is a kind of capital that can create,
for certain individuals or groups, a competitive advantage in pursuing their ends (Burt,
2000). Two contrasting arguments emerge in the literature that illustrate functions of
different types of structural embeddedness. The first is Coleman’s (1988) closure
argument, to which many current embeddedness models to regional development
correspond. The other is Burt’s (1992) structure hole argument. Although it is highly
influential in organization sciences and business literature, it is often neglected in the
regional sciences.
2.2 Closure
Coleman’s (1988, 1990) basic conclusion is that actors are better off in networks with
‘closure ‘ - a dense network in which actors are tied to multiple actors, who are
connected to one another. The first advantage of ‘closure’ is that it affects access to
information and speeds up information transmission. As Coleman (1988: p104)
explains: ‘An important form of social capital is the potential for information inherent
in social relations….A social scientist who is interested in being up-to-date on
research in related fields can make use of everyday interactions with colleagues to do
so, but only in a university in which most colleagues keep up-to-date’.
The second advantage of ‘closure’ comes from the development of trust, norms and
culture, which govern actions in the network. In a closed community, trust develops in
the frequency of interaction, in ethnic and family ties, through religious affiliation and
so on. An example of the Jews-dominated wholesale diamond market in New York
City suffices to explain this (Coleman, 1988). There are close ties among these Jews
as they live in the same community in Brooklyn, have a high degree of intermarriage,
and go to the same synagogues. Precious stones could change hands for inspection
without formal insurance because of the presence of the closed ties within the
merchant community. Without these close ties, expensive insurance devices must be
6
present to ensure the transaction to take place.
‘Closure’ governs actions within the community not only through trust, but also
through sanctions and monitoring. The news spread so quickly in the closed
community that bad behaviour and malfeasance accumulate to bad reputation in the
community which excludes the offender from the community. The economic and
social costs of developing a bad reputation is so high that one seldom dare to offend
(Granovetter, 1992).
In addition, the pressure against cheating not only arises from direct sanctions, but
also from the fact that ‘closure’ is more efficient than sparse relational networks at
generating norms, culture, rules. ‘Closure’, through its long history of mutual
interaction, has accumulated a set of ‘institutions’ such as conventions, rules (explicit
or implicit) and culture. Individual members have absorbed a set of standards from the
community because they have been in the community for so long, which means
cheating is ‘literally unthinkable’ (Granovetter, 1992).
Finally, we should also be reminded that the social capital derived from closure not
only facilitates certain actions, it also constrains others. The norms in the diamond
wholesale market for example may prevent members from transacting with dealers
outside of the community. Also the obligation to the community may ‘prevent
members from participating in broader social networks’ (Woolcock, 1998: 158).
2.3 structure hole
Rather than emphasizing the benefits of being within a cohesive network or ‘closure’,
Burt (1992, 1997, 2000) argues that social capital also emerges from another
particular type of network – network rich in structural holes. As ‘closure’ is often
associated with strong ties, structure holes are mainly concerned with weak ties or
nonexistent ties. Granovetter (1973) argues that weak ties are essential to the
information flow between otherwise disconnected groups. Burt (1992, 1997) extends
the argument. He claims that a structural hole exists wherever there is a gap between
the otherwise disconnected groups. For Burt, it is structural holes that breed potential
7
benefits. Hence once they are bridged, whatever by strong ties or weak ties
1
, there are
advantages for the actors who bring the otherwise disconnected partners together.
The first advantage concerns information access. As discussed in the ‘closure’
argument, ties to multiple actors, who are connected to one another, provide
redundant information, as similar information circulate in the network (Granovetter,
1973; 1992). A structural hole however, indicates that the people on either side of the
hole circulate in different flows of information – nonredundant information, which is
more additive than overlapping (Burt, 1992). Therefore, an actor who spans the
structural hole has access to both information flows. This could be demonstrated in
the Figure 1(see the appendix):
In Figure 1, actor X in network A exists in a ‘closure’, which provide reliable,
fast-transmitted information. However, everyone in the network tends to know the
same. In network B, there are structural holes between groups K, L, M and N, while
actor X spans the structural holes. Therefore X get access to four distinct information
sources. The potential nonredundant information advantage that the structural holes
generate has been confirmed in one of Granovetter’s (1973) widely known study –
‘the strength of weak ties’, which demonstrates that white-collar workers find better
jobs faster because of the weak ties that link otherwise disconnected social groups.
In addition, actors spanning structural holes are awarded brokerage benefits, because
the otherwise disconnected partners could only communicate through them (Burt,
1992). Therefore they act as intermediaries or brokers who facilitate exchange flows
across the network and broker tensions between other actors, ‘which gives them
disproportionate say in whose interests are served when the contacts come together’
(Burt, 2000: 10). The brokerage role in turn presents opportunities for entrepreneurial
activities (Burt, 1992). The brokers may be able to extract superior terms of trade
because of possible control benefits (Gulati, 1998).
Despite the nonredundant information and brokerage benefits, there is downside of
structural holes: the sparse network in which structural holes exist cannot provide the
1
Burt (1992) agrees that usually it is weak ties that span structural holes.
8
governance mechanism to impede opportunism which is present in ‘closure’ (Rowley
et al., 2000).
3. A structural embeddedness model of clusters
If we view clusters as networks of firms and associated institutions, then there is
much that structural embeddedness can tell about how particular network structure
can lead to mutually adaptive learning processes, with consequences for innovation.
In this section, a structural embeddedness model of clusters is presented, examining
the role of internal linkages and external linkages as well as the impact of networking
structure on cluster development, particularly in regard to knowledge creation and
diffusion.
A cluster is defined here as a network of proximate firms and associated institutions,
linked by traded interdependence and untraded interdependence (Storper, 1995)
around a core activity. It has a densely connected critical mass in a geographical area
and is bridged through external linkages with the global production/innovation
network. Ongoing interactions between internal actors and/or external actors
determine the dynamics and evolution of the cluster (internal actors means actors
within the geographically critical mass while external actors refers to those outside the
critical mass).
The definition describes a cluster as a network of internal actors and external actors.
Therefore the network relationships could be grouped into internal linkages (linkages
between internal actors) and external linkages (linkages between internal actors and
external actors). Compared to those between internal actors and/or external actors,
interactions between internal actors within the critical mass are usually very intensive
and frequent. Hence a cluster could be conceived as a dense network of internal actors
within the critical mass, linked through relatively sparse external linkages to other
clusters or organizations and therefore embedded in the global production networks or
global value chains (See figure 2). In this sense, a cluster is an ‘open system’ (Bathelt,
2003).
Figure 2 network structure of a cluster (sorry, I have to draw it manually)
9
It is also worth noting that the definition talks about internal linkages and external
linkages rather than intra- and extra-regional linkages, as the critical mass of a cluster
rarely fits in well with the boundary of a region. External linkages may well include
linkages with actors outside the critical mass but within the same region. On the other
hand, some extra-regional actors may well be within the cluster critical mass.
As it is well acknowledged that economic actions are embedded in social relationships
and the ongoing network of social relationships, the following analysis adopts the
language of ‘structural embeddedness’ and put social relationships between actors and
networks of these social relationships at the center of the analysis. The focus is on the
relevance of these relationships/linkages and structure of these linkages to knowledge
creation, diffusion and exploitation. The analysis is not an attempt to explain the
whole story behind clustering. Rather, the aim is to emphasize the upsides and
downsides of both internal linkages and external linkages in knowledge generation
and innovation and therefore the need to strike an appropriate balance between them.
3.1 clusters, knowledge and network structure
internal linkages, closure, and knowledge
Firms and other organizations in a cluster are embedded in a dense critical mass of
internal linkages. The dense network of internal actors is in effect a ‘closure’ in
Coleman’s (1988) term. Each actor is not necessarily strongly connected to everyone
else. However, there are many strong ties and there are multiple links between actors
at multiple levels so that each members of the ‘closure’ is located ‘in a dense
communal web of overlapping affiliations and obligations’ (Owen-Smith & Powell,
2004).
Within the cluster critical mass, a network of communication and information
linkages develops among the densely connected internal actors as the result of the
history of prior interactions. Information and knowledge is transferred in the daily
interactions and face-to-face contacts. The rich history of social interaction within the
‘closure’ generates multi-level relationships between individuals and organizations,
10
which link actors in multiple ways (as business partners, friends, agents, etc), offering
a great deal of channels through which information is transmitted from one end to
another (Uzzi, 1997). Over time, an ‘information and communication ecology’
(Bathelt et al., 2004), or ‘buzz’ to use Stoper & Venables’ (2002) terminology, is
formed in which information and knowledge is transferred with high velocity and is
continuously updated.
Within the ‘closure’, trust is developed when extra effort is voluntarily given and
reciprocated (Uzzi, 1997) and over time it is enhanced as misunderstanding and
suspicions are gradually eliminated through the ongoing interactions (Gulati, 1998;
Maskell & Lorenzen, 2004). In addition, ‘closure’ can also serve as a basis for
‘enforceable’ or ‘deterrence-based’ trust (Gulati, 1998). The potential sanctions
means that, even there is not real trust developed among actors in the ‘closure’, they
can still behave as if they trust each other (Granovetter, 1985; Maskell & Lorenzen,
2004).
The primary outcome of trust within the ‘closure’ is that it promotes greater exchange
of knowledge and access to privileged and fine-grained information, which is difficult
to obtain in other network structure (Uzzi, 1997). Trust not only affects information
access, it also helps to smooth interactions and develop flexible expectations among
partners (Gulati, 1998), all of which generate favorable conditions for the success of
collaboration, joint-problem solving and collective learning.
Apart from trust, ‘closure’ is conducive to the development of other institutions like
norms, conventions and routines, which are established and modified through
day-to-day interactions, regular meetings, collaborations and joint-problem solving.
They help firms to understand their partners, to develop trust and flexible expectations,
and to reduce uncertainty in knowledge transfer and economic transaction (North,
1990; Lawson & Lorenz, 1999; Bathelt, 2003). Therefore the institutional framework
serves as an enabler that facilitates the process of joint-problem solving and provides
the basis of collaborating and learning between cluster actors.
The previous discussion is a familiar story readers have seen in numerous studies in
regional sciences. However, there are further points that this paper would emphasize.
11
The first concerns the mechanism by which ‘closure’ transmits information.
Owen-Smith & Powell’s (2004) ‘irrigation’ metaphor is very illustrative in this sense.
In a ‘closure’, fine-grained information is transferred between actors that are strongly
connected. Trust derived from previous interaction means one side is more inclined to
release information and the other side more willing to accept it (Reagans & McEvily,
2003). Apart from this point-to-point transmission, information could also be
transferred within the ‘closure’ through ‘irrigation’. Therefore within a ‘closure’,
actor A need not to be directly connected to actor C to get information. As long as
there is a co-partner B between them, B would pass by the information from C to A
and vice versa. By the same mechanism, other actors within the ‘closure’ could also
be ‘irrigated’ with the information originally from C. Hence the ‘closure’, which is
rich in relational linkages, breeds potential collective information benefits for all the
members within the community. As a result, firms not only benefits from greater and
finer information exchange from direct contacts, but also could tap into the
knowledge pool within the critical mass through ‘irrigation’.
The second point that deserves attention is the fine-grained nature and quality of the
information being transmitted in the ‘closure’ (Uzzi, 1997), particularly in the context
of learning and innovation. Actors within the ‘closure’ are likely to be familiar with
each other’s practices and behaviors. They may prescreen or sift information
according to their partners’ needs before passing over the information, because
through prior relation they know what their partners need. In addition, trust within the
community reduces the risk for the receivers to accept the information. In other words,
the receiver would accept the information because it believes the information is sent
by ‘good will’. Moreover, familiarity also makes it easier for the receivers to correctly
interpret the information that is passed over. Over time the ‘closure’ may even
develop its own language through which very detailed information could be
transferred. The information could be so subtle and implied than overtly expressed in
conversation that it is hard for outsiders to understand. This is illustrated in Uzzi &
Lancaster’s (2003) study on a local group of bank loan managers, which demonstrates
that knowledge circulated in ‘closure’ could be very ‘private’ but critical to the
learning process.
12
the downside of ‘closure’
Despite the well-acknowledged and documented benefits of being in the ‘closure’, it
has been warned recently that there is a downside associated with densely networked
firms. Research in the business literature has demonstrated that the same processes by
which the ‘closure’ ‘creates a requisite fit with the current environment can
paradoxically reduce an organization’s ability to adapt’ (Uzzi, 1997: 57).
When embedded in dense internal linkages in a ‘closure’, firms are connected with
one another through various relationships. While this network structure offers the
advantage of getting fine-grained information, it can also reduce the flow of new or
novel information into the ‘closure’. Because multiple links to the same network
partners means there are many redundant ties within the ‘closure’ but few or no links
to the outside members who can potentially contribute innovative ideas (Burt, 1992).
In other words, there is a danger for the ‘closure’ to be ossified without the inflow of
new or novel information.
In addition, as time goes on, trust, conventions, norms, and other institutions, which
are derived from the intensive interconnectedness, could become the very obstacles
for further growth or the seed of decline. Within the ‘closure’, there is a tendency for
homogeneity (Granovetter, 1992). Managers, engineers and workers tend to think and
behave in a similar way and to adopt similar cognitive frameworks or ‘mental models’
(Pouder & St. John, 1996) to solve problems, conforming to the institutions. These
‘mental models’ and institutions, once being formed, tend to persist, even in the face
of contrary evidence (Pouder & St. John, 1996). Therefore when faced with an
environmental jolt, firms within the ‘closure’ may not be able to recognize the need
for radical change. They are less likely to challenge institutional norms. Hence the
network structure of ‘closure’ may limit cluster firms’ ability to anticipate and react to
environmental shocks and therefore their innovative potential. In other words, there is
a real danger of homogeneity in relation to innovation (Tracey & Clark, 2003). This is
the process through which a cluster decline over time, from ‘hot spots’ to ‘blind spots’
(Pouder & St. John, 1996).
13
external linkages, range, and knowledge
Firms and other actors within the cluster critical mass are connected with each other
by multiplayer relationships, which consist of the ‘closure’. At the same time, the
‘closure’ is linked to the outside world through some external linkages established by
some of its internal actors. They are the ones that bridge the ‘structure hole’ – the gap
between other internal actors and external actors. They are therefore hole spanners.
Without them, the critical mass would be insulated from the outside. In figure 2,
actors a, b and c are hole spanners. They link other internal actors with external actors.
At the aggregate level, they are also the channels of communication through which
the whole internal critical mass is connected with the external world. However, the
structure of these external linkages is likely to be different with that of the internal
linkages. The overlapping, multilevel and cohesive relationships within the ‘closure’
are less likely to be found in the network of external linkages. There could be some
strong ties involved in external linkages. However, it is rare to observe actors in the
external linkages to be connected with one another. In another word, external linkages
are more likely to be independent from each other. Reagans & McEvily (2003) use
‘range’ to describe the structure of the external links outside a cohesive group which
is rich in structural holes. Therefore, while ‘closure’ reflects the extent to which
internal actors are connected with one another, ‘range’ indicates the extent to which
the ‘closure’ spans structural holes between other clusters or groups. The different
network structure implies that ‘range’ may function differently from ‘closure’ in
terms of information and knowledge flow.
According to Burt (1992, 2000), hole spanners bridge different flows of information.
By building external relationships with actors outside the cluster critical mass, hole
spanners tap into nonredundant sources of information. This gives hole spanners huge
advantage. Through the internal mechanism of the ‘closure’, the nonredundant
information from structural holes will finally diffuse across the critical mass’: other
internal actors benefit the information through direct links with spanners and through
referral and irrigation. The nonredundant information then becomes redundant after
the process.
14
The external linkages that span structural holes could be strategic alliances, R&D
collaboration projects, mobile labourers and so on. Hole spanners could be innovation
centers of international companies, subsidiaries of TNCs, international logistics firms,
and global component suppliers. They could also be key scientists and researchers
who travel and meet other people around the world. These hole spanners therefore
bring into the ‘closure’ new market information, new development of technologies in
other places and in other but related technological areas, new ways of organizing
production, new method of cost management, new production standards, etc.
It is this novel information that deserves our special attention. We have known
‘closure’ is good at circulating more reliable, more accurate but redundant
information. In contrast, information transmitted through structural holes is more
likely to be new and novel. Remember how important that new information is to
innovation as innovation is very much about breaking up old tradition. The novel
information coming from outside also provide stimulus to prevent the cluster from
being ‘locked-in’. We have known ‘closure’ generates homogeneity over time.
Cluster actors tend to conform to traditional rules and conventions formed through
previous experience and interaction, using old ‘mental models’ to solve problems.
Quite likely it is only through the injection of new ideas and knowledge through hole
spanners can cluster actors challenge institutional norms, discard old ‘mental models’,
and form new ones.
Burt (1992) based his structural hole at the individual person or firm level. However,
we could extend the argument to the aggregate group level or the cluster level. For
example, again in figure 2, there is a structural hole between cluster B and C. Only
through cluster A are they connected. Therefore cluster A is a hole spanner and it
enjoys novel and diverse information and knowledge coming from both cluster B and
C. According to Kaufmann & Tödtling (2001), the novelty of exchanged information
fades over time as ‘partners become locked into well-established routine interactions’
(p796). Therefore one of the propositions for clusters would be to build up a broad
range of diverse external linkages, widening the ‘range’. The wider the ‘range’, the
broader and more diverse information the cluster embraces. Also note that cluster A
also act as a broker between cluster B and C as otherwise B and C are disconnected.
The information advantage and brokerage role partly explain the success of Silicon
15
Valley. According to Sturgeon’s (2003) analysis of the electronic industry, there has
been a trend towards spatially dispersed clusters, with Silicon Valley emerging as the
hub for the global network. The explanation from the ‘range’ perspective would be
that Silicon Valley enjoys diverse information sources as other clusters are all
connected with it (according to Bresnaham et al. (2001), the ICT clusters in Isarael,
India, Ireland and Taiwan all had significant ties with the Silicon Valley). It also
enjoys the benefits of its brokerage position, as it command and control the operation
of the whole global production network.
the downside of range
Certainly a network structure with many structural holes has its disadvantages. Firstly,
although ‘range’ may offer new and novel information which would not be circulated
in the ‘closure’ before, information transferred is less likely to be as subtle and
accurate as those diffused within a ‘closure’. Because many linkages that bridge
structural holes are weak ties, partners are less motivated and obliged to transfer
fine-grained information. The information is unlikely to be prescreened and
preprocessed according to the receivers’ needs. The receiver might have to pick up the
useful information from a lot of ‘noise’. Certainly overtime as more efforts are put in,
more interactions will take place and partners may become more familiar with each
other and therefore trust may develop with the relationship becoming more intimate.
However first that entails huge costs because of the distance and the gap between
different knowledge areas. Second, the novelty of exchanged information fades over
time as interactions become routine (Kaufmann & Tödtling, 2001).
The second disadvantage associated with ‘range’
is that it does not offer the same
kind of trust and institutions that present in ‘closure’. Trust between partners that are
distant and in different knowledge areas takes enormous time and efforts (Bathelt et.
al., 2004; Morgan, 2004). Even though trust could be built between individual
partners over time, there is no ‘public trust’ within ‘range’. As external partners are
unlikely to be familiar with each other, there is no sanction and group monitoring
mechanism in the ‘range’. Also as actors in the ‘range’ are less likely to be mutual
interdependent on one another, less likely to have day-to-day and fact-to-face actions
16
and regular meetings, norms and conventions are not easy to develop. The uncertainty
derived from the distrust and unfamiliarity forms the barrier to fine-grained
information transfer and joint-problem solving.
3.2 integration of ‘closure’ and ‘range’
mutual dependence between ‘closure’ and ‘range’
Knowledge flows through ‘closure’ and ‘range’ are indeed two distinct processes, but
they are not independent on each other. On the contrary, they rely on each other.
Firstly, ‘range’ provides opportunities to get access to novel knowledge, which is
critical to clusters development and help the ‘closure’ avoid being rigid and
‘locked-in’. There is similar thinking in Jessop’s (2001) “autopoietic system’ when he
argues that ‘an autopoietic system always co-exist with other systems that constitute
key elements in its environment and depends on them for essential conditions for its
own operation” (p218). It is argued that few clusters possess in the ‘local’ the
complete knowledge base they need to draw upon. On the contrary, the knowledge
flows that feed innovation in a cluster are increasingly being found somewhere else
(Wolfe & Gertler, 2004). Therefore clusters could only be successful if they can build
a variety of channels for accessing relevant information and knowledge from around
the world (Bathel et al., 2004), as demonstrated in the development of Taiwan’s
Hsinchu cluster (Saxenian & Hsu, 2001).
From a different perspective, Tracey & Clark (2003) explain how the ‘closure’
becomes rigid and ‘blinded’ if without external stimulus: individuals and firms are
often convinced of the effectiveness of their existing ways of thinking and operation.
They may continue to reply upon them and effectively make decisions within fixed
frames of reference, as the norms and routines have been taken for granted. Through
this negative ‘single-loop learning’, firms prison themselves in the existing patterns of
behaviour, which impairs their abilities to formulate the necessary strategy. It is
usually through the injection of external stimulus can the existing ‘mental model’ and
institutions be updated, as demonstrated in the development of the Chilean tomato
processing cluster where Japanese multinationals helped Chilean firms develop new
quality management practices and solve the long suffering problem of coordinating
17
between processors and suppliers (Perez-Aleman, 2005).
The other side of the mutual dependence between ‘closure’ and ‘range’ is that the
novel information grasped by ‘range’ needs ‘closure’ to be exploited. This is in fact
one of the central themes of Bathelt’s (2003) social systems model of geographies of
production as he proposes that ‘systems collect information from their (external)
environment, process and interpret this information and, subsequently, derive
operations from this’ (p766). How effective and efficient the ‘closure’ responds to the
new information and the change in its environment depends on its own ‘codes and
programmes’ (Jessop, 2001). Wolfe & Gertler (2004) step further by pointing out the
effectiveness of the ‘international pipelines’ depends on the quality of trust that exists
between the firms in the ‘local buzz’. ‘Closure’ helps to generate trust. According to
Krackhardt (1992), trust can reduce resistance and provide comfort in the face of
uncertainty. This partly explains how ‘closure’ can facilitate the change that ‘range’
brings.
Recent work in the business literature has corroborated the impact of ‘closure’ on the
effectiveness of ‘range’. Gargiulo & Rus (2002) for example, address the issue of top
management performance in terms of ‘access’ and ‘mobilization’: According to their
argument, while ‘range’ confers competitive advantages through maximizing access
to diverse information, ‘closure’ help to secure the mobilization necessary to reap the
benefits created by such access. Therefore, a successful team or group needs to
capture elements of both ‘range’ and ‘closure’ to achieve its goals. Similar
conclusions have been reached by other scholars (for instance, Woolcock, 1998; Burt,
2000; Rowley et al., 2000; Tracey & Clark, 2003; Reagans & McEvily, 2003;).
Perez-Aleman’s (2003, 2005) account of the development of Chilean tomato
processing cluster may suffice to explain this: New quality management practices was
introduced into the cluster through two Chilean firms’ connection with Japanese
companies. However, it was only through collective efforts of the local firms,
including group-based discussions and the established business association – the
Federation of Agro-industrial Food Processors that the novel knowledge spread and
the new practices diffused in the cluster. In this case, the established trust among local
players and the existing networking organizations helped the ‘internalization’ of new
18
knowledge, which in turn improved product quality and industrial productivity. For
Perez-Aleman (2005: 671), ‘local upgrading and growth, however, depend on
substantial indigenous efforts, rather than the classic story of multinationals
transferring technology to a passive recipient setting’.
Trade-off between ‘Closure’ and ‘Range’
While the mutual dependence between ‘closure’ and ‘range’ has attracted some
scholars in regional sciences in recent years (e.g., Sturgeon, 2003; Bathelt et al., 2004;
Wolfe & Gertler, 2004), the trade-off between the two has not been paid enough
attention. However, research in other disciplines suggests that ‘closure’ and ‘range’
could also compete against each other (Granovetter, 1992; Burt, 2000).
Firstly, too strong and too diverse external linkages may threaten the cohesion of
internal linkages and therefore the existence of the cluster (Bathelt et al., 2004). The
introduction of new ways of thinking from outside is usually different from what
exists in the ‘closure’. In addition, they are first brought in by the hole-spanners and
introduced to their direct contacts. Therefore, when the new and alien knowledge,
norms and practices are introduced into ‘closure’ by the hole spanners to their direct
contacts, there is a possibility that these new things would reduce the homogeneity of
bahaviour (Granovetter, 1992) within the ‘closure’ and hence its cohesion, as other
internal actors still stick to existing ‘frames of reference’. The impact of the ‘range’
could be so strong that the foundation of ‘closure’ could be wiped out. Bathelt et al.
(2004) are aware of the danger to clusters when they argue that:
“when actors focus primarily on external linkages, global pipelines begin to dominate the local milieu.
Therefore, less attention is being paid to local communication and information flows and people are
less interested to participate in local broadcasting. As a consequence, the local buzz becomes quieter
and the reasons for firms to locate and remain in the cluster evaporate. Such ‘hollow clusters’ might not
survive in the long-term as firms eventually shift to other locations” (page 48)
Secondly, when connections between internal actors are too dense, ‘closure’ may
become blinded to external knowledge and changes and therefore detrimental to the
effectiveness of ‘range’. This has been explained before and has been shown in many
other studies of the phenomena of ‘lock-in’, ‘blind-confidence’, and
19
‘overembeddedness’ (Uzzi, 1996; 1997; Grabher, 1993b).
Thirdly, internal linkages and external linkages compete for investment of time and
resources. Bathelt et al. (2004) are right to point out that the establishment and
maintenance of global pipelines is not free. The process entails enormous amount of
investment and resources, as the partners on both ends need to develop special
schemes to understand each other’s institutional context and a certain degree of trust
in order to engage interaction (Owen-Smith & Powell, 2002). However, it may be
arbitrary to allege that ‘participating in the buzz does not require particular
investment’ (Bathelt et al., 2004: 38). The nature of buzz or ‘closure’ is indeed
spontaneous and fluid. However, this is the result of a long history of interaction,
which takes a good deal of time and effort. To enjoy the fine-grained information
flowed in the ‘closure’, firms need the ‘insider’ identity, which is not free-given.
Reciprocity is present in the ‘closure’ because ‘we have been close for so long that we
expect this of one another’ (Granovetter, 1992: 42). In addition, the establishment of
norms, conventions and common rules should be the result of a long process of
meeting and negotiation. Furthermore, relationships between the internal actors also
need time and resources to maintain; otherwise they may fade over time. To
summarize, the central theme is that clusters and firms may need to choose between
internal linkages and external linkages as time, investment and other resources are
limited.
Integrating ‘closure’ and ‘range’
There has been dispute in the literature in what is the right network structure for
individuals or groups to be in. More sophisticated arguments suggest though, that an
optimal network structure is a combination of ‘range’ and ‘closure’ as both have
advantages and disadvantages (Woolcock, 1998; Burt, 2000; Reagans & McEvily,
2003). It is argued that the ‘closure’ perspective emphasizes ‘dense patterns of local
interaction as the basis for coordination and collective action’, while the ‘range’
perspective focuses on ‘bridge across global divisions as the basis for information
transfer and learning’ (Reagans & Zuckerman, 2001: 512). Similar but more
embryonic thinking could also be found in economic geography, for example ideas of
20
‘neo-Marshallian nodes in global networks’ (Amin & Thrift, 1992; Coenen et al.,
2004) and ‘local buzz, global pipeline’ (Bathelt et al., 2004).
For Gargiulo & Rus (2002), the controversy on the right type of network structure
with which actors are better off conceals a fundamental difference in the assumption
about the type of uncertainty to which ‘closure’ or ‘range’ is a solution
2
. The ‘closure’
perspective implicitly assumes the uncertainty facing actors is how to secure
coordination between cluster members. The uncertainty is reduced if actors are
embedded in a densely connected network that facilitates information flow and trust
generation. The ‘range’ perspective assumes that the greatest uncertainty is how to
economically secure privileged access to relevant knowledge and information. Thus
actors that can bridge structural holes have a competitive advantage in pursing their
interests.
Yet clusters and firms may face both uncertainty in having access to the right
knowledge and information and uncertainty in their ability to secure coordination and
collective actions to exploit the knowledge. An integration model of clusters therefore
needs to incorporate these two uncertainties into the equation. While the ‘range’ of
external linkages ensures the cluster’s access to diverse knowledge and information,
‘closure’ of internal linkages secures the necessary coordination and cooperation to
pursue the opportunities created by such access. For a given cluster, the appropriate
network structure would then be a function of the criticality of these two distinct
uncertainties. When there is no difficulty to get access to the right knowledge and
information but the coordination is crucial, ‘closure’ of internal linkages will provide
the mechanism to overcome the problem. On the contrary, when accessing to the
necessary information and knowledge rather than the need for coordination appears to
be the critical, ‘range’ of external linkages will be the solution.
4. Contingent factors of structural embeddedness
As time and resources devoted to their networks are limited, clusters and firms need
2 Gargiulo & Rus (2002) discuss the uncertainty that ‘network closure’ and structural hole theory implicitly
assume in the case of top managers’ performance rather than clusters.
21
to strike a balance between ‘range’ and ‘closure’. However, because of the complex
relationship between ‘closure’ and ‘range’, the proper match between the type of
uncertainty and the type of network structure is not easy. Recent work indicates that
the optimal mix of ‘range’ and ‘closure’ is contingent on a number of factors
3
, among
which are type of region, type of industry, type of activity, type of knowledge, type of
learning, etc (Ahuja, 1997; Oinas, 1999; Rowley et al., 2000; Breschi & Malerba,
2001). Obviously this paper cannot discuss all these contingent factors. It hence
chooses to address two important factors: cluster difference and cluster life cycle. The
proposition is that, the criticality of uncertainties in knowledge access and action
coordination is different across different clusters and various cluster life stages, which
requires different structures of ‘closure’ and ‘range’ to facilitate cluster development.
4.1 cluster difference
Clusters vary with the core sectors they are around
4
. To a large extent, the difference
between clusters is determined by the difference between the core sectors they are
based upon. Therefore the cluster difference addressed here is very much similar to
sectoral difference, with particular attention being paid to the fact that clusters are
usually broader than sectors.
Previous empirical studies suggest that the degree of importance of ‘range’ and
‘closure’ is different for different clusters or sectors. Freel (2003), for example,
indicates that, novel innovators have greater geographical reach of their innovation
links, while incremental product innovators appear to be more locally embedded. This
seems to suggest that the ‘range’ of non-local linkages is more important for
research-intensive sectors and local linkages are more important for other sectors.
This is echoed in a study of a biotechnology cluster in Scotland (Leibovitz, 2004)
3 According to Owen-Smith & Powell (2004), this is a proposition that has not been well attended
to in literatures.
There are some embryonic ideas though. For example, Bathelt et al. (2004) agree that a mix of ‘local buzz’ and
‘global pipeline’ is necessary to ensure continued growth and innovation. they also argue that the particular mix
can vary across value chains, technologies and markets segments. they suspect some industries require more buzz
while others need more pipelines, but they do not provide any answer to why that should be the case.
4
Clusters vary along at least three dimensions: the core sectors they are based, the space they are concentrated in,
22
which finds that local interactions are surprisingly weak and that many of the key
relationships operate at an international scale. Hendry et al., (2000), in their study on
opto-elctrinics clusters in six locations also demonstrate stronger global relations vs.
local relations. However, opposite arguments and observations exist. Some studies
have shown that the local environment, and the relative propinquity of innovation
partners, is likely to be more important for novel innovators than for incremental
innovators (Baptista and Swann, 1998; Nooteboom, 1999).
It seems that without clear theoretical guidance, it is very easy for empirical studies to
be biased towards either the ‘range’ argument or the ‘closure’ argument because of
the specific nature of the clusters being investigated and the feature of the locality.
The existing literature can hardly answer the question of ‘why there is cluster
difference in terms of relative importance of closure or range’. Probably the best we
could found is the embryonic idea that ‘some industries require more buzz while
others need more pipelines’ (Bathelt et al., 2004). To find the answer of why, one of
the key questions we need to ask first is, from the knowledge and learning perspective,
what differentiates clusters from each other? In this respect, recent research on
Sectoral Systems of Innovation (SSI) provides important food for thoughts (for SSI,
see Breschi & Malerba, 1997; Malerba, 2001, 2002).
Based on the empirical evidence of sectoral patterns of innovative activities and the
notion of ‘technological regime’ (Nelson & Winter, 1982; Winter, 1984), the concept
of SSI lays out five aspects in which sectoral systems of innovation and production
are different from each other: knowledge base and learning processes; basic
technologies, inputs and demand, with key links and dynamic complementaries; type
and structure of interactions among firms and non-firm organizations; institutions;
processes of generation of variety and selection (Malerba, 2002).
From the knowledge and learning perspective, the discussion on knowledge and
learning processes in the SSI literature is of particular interest as it sheds light on why
and how different clusters may have different emphasis on ‘range’ or ‘closure’.
Following Nelson & Winter (1982)’s idea of technological regime, Malerba (2001,
and the time dimension, i.e., the dynamic evolution process.
23
2002) maintains that, accessibility, appropriability and cumulativeness are key
dimensions of knowledge, which differ across sectors. Technology accessibility
reflects the likelihood of gaining knowledge that is external to firms. Accessible
knowledge could be internal and external to the sector. Cumulativeness refers to the
degree by which new knowledge builds upon current knowledge or the possibility of
innovating along specific trajectories. Appropriability of innovation reflects the
possibility of profiting from innovative activities by constraining imitation and other
opportunitistic behaviour. It is argued that these specificities of knowledge and
technological regimes “provide a powerful restriction on the patterns of firms’
learning, competencies, behaviours and organization of innovative and production
activities in a sectoral system” (Malerba, 2002: 254). Support for the argument could
be found in Pavitt’s sector taxonomy, which demonstrates the contrasting sectoral
patterns in terms of sources of technology, cumulativeness in development and
possibilities for appropriation (Pavitt, 1984).
Extending the argument to cluster difference and exploring its implication to the
structure of ‘range’ and ‘closure’, one could propose that the specificities of
technological regimes and the knowledge base condition the environment within
which clusters and firms operate. A given technological regime and knowledge base
sets the parameters for specific uncertainties in knowledge access and action
coordination. Therefore clusters may be better off with some distinct mix of ‘range’
and ‘closure’ but not with others.
For those clusters characterized by high levels of technological accessibilities,
knowledge is relatively easy to obtain. Clusters and firms hence should endeavor to
build up the necessary coordination mechanism to exploit the knowledge, which
requires frequent and intensive interaction and therefore ‘closure’ of internal linkages.
For those clusters characterized by low level of appropriability of innovation and
opportunistic behaviour, ‘closure’ will help to establish norms, common rules,
regulations, and other institutions, which provide the necessary protection. ‘Closure’
also helps to ease information flows, which is another barrier to appropriability
(Robertson & Langlois, 1995). For those clusters with low level of cumulativeness at
the cluster level, the necessary information and knowledge that is critical to cluster
development is more likely to be found outside the cluster critical mass. Therefore
24
these clusters would benefit more from ‘range’ of external linkages, which provide
the critical new and novel information from many different alternatives.
4.2 cluster life cycle
Obviously the uncertainty problem facing clusters is not static. The structural
embeddedness model therefore needs to consider clusters development over time and
the evolving match between the network structures and the uncertainties they are
aligned to. It is not the case that previous studies paid no attention to the time
dimension of cluster development. Audertsch & Feldman (1996) for example, have
found empirical support for their argument that innovative activity tends to cluster
more during the early stages of the industry life cycle, but has the propensity to be
more dispersed during the mature and declining stages. In the OECD proceedings on
innovative clusters, Peneder (2001) presents a stylized industry life-cycle model to
illustrate the development of culture industry-based clusters. Although these studies
are highly illustrative, it is maintained here that the more direct and relevant subject of
inquiry is cluster life cycle rather than general industry life cycle.
Some cluster life cycle models have emerged in the literature in recent years (for
example, Pouder & St. John, 1996; Klink & Langen, 2001). It is usually proposed that
clusters evolve through stages of emergence, expansion, mature and transition or
decline. While a stage approach has clear limitations
5
, this perspective is useful in
framing the general processes of cluster evolution and the characteristics of each
process. The key point here is that, from a knowledge and learning point of view, each
stage represents a unique, strategic context with particular uncertainties in knowledge
and coordination, which needs appropriate network structure for a given cluster to
successfully survive and grow.
A stylized three-stage model is proposed here: origination, convergence, and
reorientation or decline. The stages are named following Pounder & St. John (1996),
while they are assigned to different characteristics here. The origination stage
concerns the emergence of the core or anchor firms as the result of successful initial
5
For example, how to define the boundaries between stages
25
entrepreneurial efforts and the follow up of a few similar firms and suppliers and so
on in the same location. As the critical mass has not been achieved, the ‘cluster’ at
this stage is not a cluster in the real sense. Although much has been written about the
fully-fledged clusters, how clusters emerge or originated has been less theorized.
Some case studies show that agglomeration economics and external effects that are
often associated with clusters play only a small role in the origination stage. Drawing
on case studies on the emergence of various new clusters, the AIM research (Andriani
et al., 2005) summarizes that different conditions may underpin the emergence of
clusters such as a lead or anchor firm, public sector investment and activities, shocks
and precipitating events, and local demand and market patterns. The main theme of
these case studies is that the engines that foster the rise of a cluster can be very
different from those that keep it going. Indeed, it is argued that the critical factor
common in emerging clusters is the efforts of pioneering entrepreneurs to take
advantage of new technological and market opportunity that had not already been
exploited (Bresnahan et al, 2001; Feldman, 2001). Certainly, both uncertainties in
knowledge accessibility and in coordination are high at this stage. In addition, the
cluster critical mass has not being formed. Therefore it is difficult to discuss the
structure of ‘closure’ and ‘range’ at the origination stage.
At the convergence stage, an agglomeration of firms has begun to from. A critical
mass has come into being and is continuing growing. In addition, the knowledge base
of innovative activities is evolving towards a dominant design (Malerba, 2002). In
other words, knowledge is rather cumulative at the cluster level. Based on past
successful experience, ‘current innovative firms are more likely to innovate in the
future in specific technologies and along specific trajectories’ (Malerba & Orsennigo,
2000:302). Moreover, a mobile labour force, cooperative alliances, personal
relationships, direct observation and monitoring, and also local media lead to a high
level of information exchange among the cluster internal actors (Pouder & St. John,
1996; Maskell, 2001; Bathelt et al., 2004). Face-to-face interaction, the intimacy
involved in the interaction, and the similar ‘mental models’ (Pouder & St. John, 1996)
make the information being exchanged highly interpretable. Therefore knowledge
circulated in the critical mass is abundant and highly accessible. In summary, the
convergence stage could be characterized by high accessibility and cumulativeness of
knowledge. The prominent uncertainty facing the cluster at this stage is how to secure
26
the coordination among members to exploit the opportunities derived from the
available knowledge. As has been discussed before, the cluster will be better off if the
network structure is leaned towards ‘closure’ to establish the necessary institution.
After a ‘long time spans of incremental change and adaptation which elaborate
structure, systems, controls, and resources toward increased coalignment’ (Tushman
& Romanelli, 1985: 215, quoted from Pouder & St. John, 1996: 1205), clusters enter
into the reorientation or decline stage. The availability and richness of information
about internal actors is still high and may still increase over time. However, ‘as
innovation process changes to involve the development of more complex technologies,
the production of these technologies requires the support of sophisticated
organizational networks that provide key elements or components of the overall
technology … increasingly the components of these networks are situated across a
wide array of locations’ (Wolfe & Gertler, 2004: 1077). At this stage, the ‘closure’ is
neither large enough nor heterogeneous enough to provide the necessary knowledge
and resources for more complex innovation and production. Information within the
dense internal network is ‘high in redundancy and low in diversity’ (Hite & Hesterly,
2001). In other words, the self-sufficiency or cumulativeness of knowledge is
declining at the cluster level
6
. Moreover, as internal actors tend to stick to the
established norms, institutions and the ‘mental models’ (Pouder & St. John, 1996),
which are primarily based on what happened in the critical mass, they become more
homogeneous and less sensitive to external stimulus. Therefore information and
knowledge from the outside will be subject to less rigorous scrutiny (Pouder & St.
John, 1996), which effectively reduces the accessibility of external knowledge. In
summary, at the reorientation/decline stage, knowledge has begun to be less
cumulative and the accessibility of knowledge becomes lower. The main uncertainty
facing the cluster is therefore shifting to how to secure the necessary information and
knowledge. As has been discussed before, the only chance for the cluster to
restructure or to rise ‘like phoenix from the ashes’ (Tödtling & Trippl, 2004), is to
have a network structure leaning towards ‘range’.
6
of course, as the cluster grows, there are more external actors being involved, but the necessary horizon could
still beyond the reach of the cluster if no particular efforts is made. and the decline of the self-sufficiency and
27
The above discussion gives a rudimentary answer to Malerba’s (2002:259) question:
“do relationships among agents and networks show a great stability or do they change
over time, and if so, in which direction?” The analysis demonstrates that clusters’
network structure co-evolve with their development and the changing uncertainties in
knowledge and in coordination. As clusters move from origination stage to
convergence stage, they need to build up ‘closure’ to reduce the uncertainty in
coordination and exploit the rich information circulated within the network. As the
cluster become matured and shows signs of decline, it needs special efforts to
diversify the ‘range’ of external linkages, in order to get access to the relevant
knowledge to restructure itself. This is captured in Granovetter’s ‘coupling and
decoupling’ solution to the network structure problem facing ethnic entrepreneurs, “in
which members of economic groups draw initially upon the resources of family and
peers but then attempt to forge broader and more autonomous ties beyond the group
as their need for lager markets and more sophisticated inputs expands” (Woolcock,
1998: 175).
5. Conclusion and policy implications
Starting from the debate around internal and external linkages in clustering, this paper
addresses the relationship between network structure and cluster development. With a
theoretical core based upon ‘structural embeddedness’, a knowledge-based model of
clusters is presented, aiming at clarifying functions and structure of internal and
external linkages in knowledge creation and the learning process.
The model demonstrates that both cohesive internal linkages and diverse external
linkages are important for cluster development. Cohesive internal linkages, or
‘closure’ are important for fine-grained information transmission and generating trust,
norms and other institutions and therefore facilitating coordination and collective
actions. In contrast, ‘range’, or network of diverse external linkages rich in structural
holes, brings into the cluster critical mass novel information, new ways of doing and
thinking, and therefore prevents the critical mass being locked-in.
cumulativeness of knowledge is an inevitable trend as demonstrated by the Silicon Valley case.
28
The relationship between ‘closure’ and ‘range’ is very complicated. On one hand,
‘closure’ and ‘range’ are mutual dependent on each other. ‘Range’ could bring in
novel knowledge, which is crucial to the critical mass. The information through
‘range’ however, needs ‘closure’ to coordinate cluster members in order to develop
action and exploit the novel information. On the other hand, there is trade-off between
‘closure’ and ‘range’. Too diverse external linkages may threaten the cohesion of
‘closure’ and therefore the existence of the cluster. When the internal connections
become too dense, ‘closure’ may become blinded to external knowledge and change
and therefore detrimental to the effectiveness of ‘range’. In addition, as relationships
need time and efforts to build up, maintain and improve, ‘range’ and ‘closure’
compete for the limited resources.
The advantages and disadvantages associated with ‘closure’ and ‘range’ and the
complex relationship between them imply that clusters need to integrate both of them.
According to the integration model, the network structure of a cluster should
facilitates access to a diverse range of contacts outside the critical mass who might
provide the necessary information on one hand, and on the other hand provide
mechanisms to coordinate firms and other organizations within the critical mass to
secure the execution of the intended strategies.
Finding an appropriate mix of ‘range’ and ‘closure’ though is problematic as it
depends on the specific uncertainties facing the cluster. When the prominent
uncertainty is in accessing relevant knowledge, the balance may need to lean towards
‘range’. When the greatest uncertainty is in securing the coordination, the balance
may need to lean towards ‘closure’.
In addition, an appropriate balance is contingent on a number of factors, among which
are cluster difference and cluster life cycle. Different clusters belong to different
technological regimes and have a different knowledge base, and are therefore facing
different degree of uncertainty in knowledge access and in action coordination.
Consequently, the ‘ideal’ mixes of ‘range’ and ‘closure’ are different. Similarly the
uncertainties facing clusters also vary along cluster life stages. Hence the appropriate
network structure should co-evolve with the development of the cluster.
29
The model has important implications for cluster policies. It calls for attention to be
paid to the network structure of internal and external linkages when designing cluster
policies. It proposes the inclusion of external linkages into policy consideration. The
model demonstrates both internal linkages and external linkages could be beneficial
as well as detrimental to cluster development. Therefore it suggests the importance of
striking the right balance between ‘closure’ and ‘range’ in regard to knowledge
creation and interactive learning.
The inclusion of external linkages is worth particular attention of policy makers, as
the majority of current policies are still focusing on local networking and overlooking
the importance of external communication channels (Bathelt et al., 2004). In addition,
the model emphasizes the importance of building up diverse or range of external
linkages. The implication is that it may not be a good idea for clusters to be linked
with one or several dominant external actors.
Finally, the model suggests that the appropriate mix of ‘range’ of external linkages
and ‘closure’ of internal linkages is a subtle balancing act and contingent on a number
of factors. Therefore standard ‘one-fits-all’ (Martin & Sunley, 2003) cluster policy
may not be successful. In particular, the model reminds policy makers that the balance
between ‘range’ and ‘closure’ may be different for clusters based on different sectors
and even for the same cluster but at different cluster life stages.
Acknowledgement:
I would like to thank Stewart MacNeill and Alex Burfitt for their useful comments.
The usual disclaimer applies. Sponsorship from the School of Public Policy
Studentship and the Pat Cam Award is gratefully acknowledged.
30
Reference:
Ahuja, G. 2000 ‘Collaboration networks, structural holes, and innovation: A longitudinal study’,
Administrative Science Quarterly 45 (3): 425-455
Audretsch, D.B. & Feldman, M. 1996 ‘Spillovers and the geography of innovation and
production’, American Economic Review 86: 630-40.
Baptista, R., Swann, P., 1998. ‘Do firms in clusters innovate more?’ Research Policy 27: 525-540
Bathelt, H. 2003 ‘Geographies of production: growth regimes in spatial perspective 1 – innovation,
institutions and social systems’, Progress in Human Geography 27 (6): 763-778
Bathelt, H., Malmberg, A. and Maskell, P. 2004 ‘Clusters and knowledge: local buzz, global
pipelines and the process of knowledge creation’, Progress in Human Geography 28 (1): 31-56
Bathelt H. and Taylor M.(2002): ‘Cluster, Power and Place: inequality and local growth in
time-space’, Geografiska Annaler 84B: 93-109
Breschi, S. & Malerba, F. 1997 ‘Sectoral innovation systems: Technological regiomes,
Schumpeterian dynamics, and spatial boudaries’, in Charles Edquist (ed.), Systems of Innovation:
Technologies, Institutions and Organizations, London: Pinter, 130-156
Breschi, S. & Malerba, F. 2001 ‘The geography of innovation and economic clustering: some
introductory notes’, Industrial and Corporate Change 10(4):817-833
Bresnaham, T., Gambardelia, A., and 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
Britton, J. N.H. 2003 ‘Network structure of an industrial cluster: electronics in Toronto’,
Environment and Planning A, 35: 983-1006
Burt, R.S. (1992): The Social Structure of Competition. In Nohria, N. & Eccles, R. (Eds)
Networks and Organization: Structure, firm and action. Harvard Business School Press, Boston,
MA.
Burt, R.S. 1997 ‘The contigent value of social capital’, Administrative Science Quarterly 42 (2):
339-365
Burt, R.S. (2000): The network structure of social capital. In Robert I. Sutton and Barry M. Staw.
Greewich (eds) Research in Organizational Behavior, CT: JAI Press
Coenen, L., Moodysson, J. and Asheim, B. 2004 ‘Nodes, Networks and proximities: on the
knowledge dynamics of the medicon valley biotech cluster’, European Planning Studies 12 (7):
1003-1018
Coleman, J.S. (1988): ‘Social capital in the creation of human capital’, American Journal of
Sociology, 94, pp. 95-120
Coleman, J.S. (1990): Foundations of Social Theory. Cambridge, MA. Harvard University Press
Cooke, P. 2004b Evolution of regional innovation systems – emergence, theory, challenge for
actions, in Cooke, P., Heidenreich, M. and Braczyk, H-J. (Eds) Regional Innovation Systems (2nd
ed). London: Routledge
Dicken, P. Kelly, P.F. Olds, K. and Yeung, H.W.-C. 2001 ‘Chains and networks, territories and
scales: towards a relational framework for analyzing the global economy’. Global Networks 1:
31
89-112
Feldman, M. ‘The entrepreneurial event revisited: Firm formation in a regional context’,
Industrial and Corporate Change 10 (4): 861-891
Freel, M.S., 2003. ‘Sectoral patterns of small firm innovation, networking and proximity’.
Research Policy 32, 751-770
Gargiulo, M. & Rus, A. 2002 ‘Access and Mobilization: Social capital and top management
response to market shocks’, Working Paper, INSED
Granotvetter, M. (1973): ‘The strength of weak ties’, American Journal of Sociology, 78, pp.
1360-1380
Granovetter, M.(1985): ‘Economic action and economic structure: the problem of embeddedness’,
American Journal of sociology 91, 481-510
Granovetter, M.(1992): Problems of explanation in economic sociology, in Nohria, N. and Eccles,
R.(eds), Networks and Organizations: Structure, Form and Action (Boson, MA: Harvard Business
School Press)
Grabher, G.(1993a): ‘Rediscovering the social in the economics of interfirm relations’, in Grabher,
G.(ed): The Embedded Firm. On the Socioeconomics of Industrial Networks, Routledge, London
and New York, 1-31
Grabher, G. (1993b): ‘The weakness of strong ties: the lock-in of regional development in the
Ruhr area’, in Grabher, G.(ed): The Embedded Firm. On the Socioeconomics of Industrial
Networks, Routledge, London and New York, 1-31
Grabher, G. 2002 ‘Cool projects, boring institutions: temporary collaboration in social conetxt’,
Regional Studies 36: 205-14
Gulati, R. 1998 ‘Alliances and networks’, Strategic Management Journal, 19: 293-317
Hendry, C., Brown, J. and Defillipi, R. 2000 ‘Regional clustering of high technology-based firms:
opto-electronics in three countries’, Regional Studies 34 (2): 129-144
Hess, M. 2004 ‘’Spatial’ relationships? Towards a reconceptualization of embeddedness’,
Progress in Human Geography 28 (2): 165-186
Hite J.M. & Hesterly, W.S. 2001 ‘The evolution of firm networks: from emergence to early
growth of the firm’, Strategic Management Journal 22: 275-286
Hudson, R. (1999): ‘The learning economy, the learning firm and the learning region: A
sympathetic critique of the limits of learning’, European Urban and Regional Studies, 6(1): 59-72
Humphrey, J. & Schmitz, H. 2002 ‘How does insertion in global value chains affect upgrading in
industrial clusters?’, Regional Studies 36 (9): 1017-1027
Jessop, B. 2001 ‘Regulationist and autopoieticist reflections on Polanyi’s account of market
economics and the market society’, New Politics Economy 6 (2): 213-232
Kaufmann, A. & Tödtling, F. 2001 ‘Science-industry interaction in the process of innovation: the
importance of boundary-crossing between systems’, Research Policy 30:791-804
Klink, A.V. and Langen, P.D. (2001): ‘Cycles in industrial clusters: the case of the shipbuilding
industry in the Northern Netherlands’. Tijdschrift voor Economische en Sociale Geografie 92(4):
449-463
Krackhardt, D. (1992): The strength of strong ties: The importance of Philos in organizations. In
Nohria, N. and Eccles, R.(eds), Networks and Organizations: Structure, Form and Action(Boson,
MA: Harvard Business School Press)
32
Lawson, C. & Lorenz, E. 1999 ‘Collective learning, tacit knowledge and regional innovative
capacity’, Regional Studies 33: 302-317
Leibovitz 2004. ‘Embryonic knowledge-based clusters and cities: the case of biotechnology in
Scotland’ Urban Studies 41, 5/6, 1133-1155
Lewis, N., Moran, W., Perrier-Cornet, P., and Barker, J. 2002 ‘Territoriality, enterprise and ré
glementation in industry governance. Progress in Human Geography 26: 433-62
Lonrenz, E.H. 1992 ‘Trust, community and cooperation. Towards a theory of industrial districts’.
In Storper, M. & Scott, A.J. (Eds) Pathways to Industrialization and Regional Development.
Routledge: London
Lovering, J. (1999): ‘Theory led by policy: the inadequacies of the new regionalism’, International
Journal of Urban and Regional Research 23(2): 379-395
MacLeod, G. 2001 ‘New regionalism reconsidered: Globalization and the remaking of political
economic space’, International Journal of Urban and Regional Research 25 (4): 804-829
Malerba, F. 2001 ‘Sectoral systems of innovation and production: Concepts, analytical framework
and empirical evidence’, Paper prepared for the ECIS Conference “The Future of Innovation
Studies”, Eindhoven, September 20-23, 2001
Malerba, F. 2002 ‘Sectoral systems of innovation and production’, Research Policy 31: 247-264
Malerba, F. & Orsenigo, L. 2000 ‘Knowledge, innovative activities and industrial evolution’,
Industrial and Corparate Change 9 (2): 289-314
Martin, R. & Sunley, P. 2003 ‘Deconstructing clusters: chaotic concept or policy panacea?’.
Journal of Economic Geography 3:5-35
Maskell, P.(2001): Towards a knowledge-based theory of the geographical cluster, Industrial and
Corporate Change 10(4), 921-43
Maskell, P. & Lorenzen, M. 2004 ‘The cluster as market organization’, Urban Studies 41 (5/6):
991-1009
Maskell, P. and Malmberg, A. (1999): ‘The competitiveness of firms and regions’, European
Urban and Regional Studies 6(1): 9-25
Morgan, K. 2004 ‘The exaggerated death of geography: learning, proximity and territorial
innovation systems’, Journal of Economic Geography 4: 3-21
Nooteboom, B., 1999. ‘Innovation, learning and industrial organization’. Cambridge Journal of
Economics 23, 127-150
Nelson, R. & Winter, S.G. 1982 An Evolutionary Theory of Economic Change. Cambridge, MA:
Belknap Press
North, D.C.(1990): Institution, Institutional Change and Economic Performance, Cambridge
University Press, Cambridge
Oinas, P. 1999 ‘Activity-specificity in organizational learning: implications for analyzing the role
of proximity’, Geojournal 49: 363-372
Owen-Smith, J. & Powell, W. 2004 ‘Knowledge networks as channels and conduits: The effects
of spillovers in the Boston Biotechnology community’, Organization Science 15 (1): 5-21
Pavitt, K. 1984 ‘Sectoral patterns of technical change: Towards a taxonomy and a theory’.
Research Policy 13: 343-373
33
Peneder, M., 2001 ‘Dynamics of initial cluster formation: the case of multimedia and cultural
content’, in Innovative Clusters: Drivers of National Innovation Systems, OECD
Perez-Aleman, P. 2003 ‘A learning-centred view of business associations: Building
business-government relations for development’, Business and Politics 5: 195-215
Perez-Aleman, P. 2005 ‘Cluster formation, institutions and learning: the emergence of clusters and
development in Chile’, Industrial and Corporate Change 14 (4): 651-667
Pouder, R. & St. John, C.H. 1996 ‘Hot spots and blind spots: geographical clusters of firms and
innovation’, Academy of Management Review, 1996 21 (4): 1192-1225
Reagans, R. and Zuckerman, E., 2001. ‘Networks, diversity and productivity: The social capital of
corporate R&D teams’. Organization Science 12 (4), 502-517
Reagans, R. & McEvily, B. 2003 ‘Network structure and knowledge transfer: The effects of
cohesion and Range’, Adminstrative Science Quarterly 48: 240-267
Robertson, P.L. & Langlois, R.N. 1995 ‘Innovation, networks, and vertical integration’, Research
Policy 24: 543-562
Rowley, T., Beherns, D., and Krackhardt, D. 2000 ‘Redundant governance structures: an analysis
of structural and relational embeddedness in the steel and semiconductor industries’, Strategic
Management Journal 21: 369-386
Saxenian, A. & Hsu, J-Y 2001 ‘The Silicon Valley – Hsinchu connection: technical communities
and industrial upgrading’, Industrial and Corporate Change 10 (4)
Simmie, J. 2004 ‘Innovation and clustering in the globalised international economy’, Urban
Studies 41 (5/6): 1095-1112
Staber, U. 1996 ‘Networks and regional development: perspectives and unresolved issues’, in
Staber, U., Schaefer, N. and Sharma, B. (Eds) Business Networks Prospects for Regional
Development, Berlin: Walter de Gruyter
Storper, M. (1995): ‘The resurgence of regional economics Ten years later: The region as a nexus
of untraded interdependence’. European Urban and Regional Studies 2(3): 191-221
Stoper, M. & Venables, A.,J. 2002 ‘Buzz: the economic force of the city’, Paper presented at the
DRUID Summer Conference on ‘Industrial Dynamics of the New and Old Economy – Who is
Embracing Whom?’, Copenhagen, Elsinore, 6-8 June
Sturgeon, T. J. 2003 ‘What really goes on in Silicon Valley? Spatial clustering and disperal in
modular production networks’, Journal of Economic Geography 3: 199-225
Tödtling, F. & Trippl, ‘Like Phoenix from the ashes? The renewal of clusters in old industrial
areas’, Urban Studies 41 (5/6): 1175-1195
Tracey, P. & Clark, G.L. 2003 ‘Alliances, networks and competitive strategy: rethinking clusters
of innovation’, Growth and Change 34 (1): 1-16
Tushman, M.L. & Romanelli, E. 1985 ‘Organizational evolution: A metamorphosis model of
convergence and reorientation’, in L.L. Cummings & B.M. Staw (Eds), Research in
Organizational Behaviour, Vol. 7: 171-222. Greenwich, CT: JAI Press
Uzzi, B. (1997): ‘Social structure and competition in interfirm networks the paradox of
embeddedness’, Administrative Science Quarterly 42: 35-67
Uzzi, B. & Lancaster, R. 2003 ‘Relational embeddedness and learning: The case of bank loan
34
managers and their clients’, Management Science 49 (4): 383-399
Winter, S.G. 1984 ‘Schumpeterian competition in alternative technological regimes’, Journal of
Economics Behaviour and Organization 5: 287-320
Wolfe, D.A. & Gertler, M.S. 2004 ‘Clusters from the inside and out: local dynamics and global
linkages’, Urban Studies 41 (5/6): 10711093
Woolcock, M. (1998): Social capital and economic development: toward a theoretical synthesis
and policy framework. Theory and Society 27 (2): 151-208
35
Appendix:
Figure 1 closure and structural hole
X
X
L
K
M
N
Network A
Network B