Introduction¹

Pointing to the continued role of geographical clusters ² (agglomerations of specialized economic activity) in innovation processes, economic geographers have been a pervasive voice of dissent from the last decades’ claims that the world is becoming “flat” (Iammarino and McCann, this volume, Chapter 14). Concomitantly, international business scholars have also paid attention to clusters, focusing on multinational enterprises (MNEs) and how these span geographical distance in order to search for new markets, resources, and innovation opportunities, and reaching the same conclusion (Ghemawat 2007). Now, the two disciplines are converging around an understanding of cluster connectedness (the quantitative volume of a cluster’s global connections) and connectivity (the nature and scope of these global connections). This nascent theoretical framework can address some pertinent questions related to global science, technology, and innovation. For instance, why do some clusters benefit more than others from being integrated in the global economy? More specifically, what makes particular clusters able to leverage their outward trade and collaboration to catch up to global technological frontiers or even become global technological leaders – and what determines how the fruits of such innovation are distributed among individuals and organizations in these clusters?

While the internal processes of innovation in clusters are relatively well understood (Maskell and Malmberg 1999; Gertler 2003; Maskell and Lorenzen 2004; Storper and Venables 2004), it is a relatively new recognition that globalization has increased both inter-cluster competition and the innovation-related “gains from trade” that arise from connections between firms located in different clusters (OECD2004; Klagge and Martin 2005; Zademach 2009; Lorenzen and Mudambi 2013). Having been studied under diverse headings such as “global value chains” (Dicken et al. 2001; Gereffi, Humphrey, and Sturgeon 2005; Mudambi 2008), “commodity chains” (Pratt 2008), “production networks” (Coe, Dicken, and Hess 2008) “flows” (Hudson 2005), and “linkages” (Giuliani, Rabellotti, and Van Dijk 2005), the study of connections between clusters is still a nascent research field. In the following sections, we provide an overview of this field well as some pointers of how to promote our understanding of the impact of innovation through analyzing cluster connectedness and connectivity.

Global Connections: Connectedness and Connectivity

Global connections are channels for resource flows in and out of clusters (Amin and Thrift 1992; Bathelt, Malmberg, and Maskell 2004). While such connections facilitate harsher inter-cluster competition (Amin 2002; Humphrey and Schmitz 2002; Sheppard 2002; Morgan 2004; Pratt 2008) they also hold potential for innovation in clusters through infusing them with knowledge, technology, and capital from worldwide sources (Amin 2002; Davenport 2005; Gertler and Levitte 2005). Crossing the boundaries of a cluster can lead to global connections with cutting-edge technologies and knowledge – lead clusters, “centers of excellence” (Cantwell and Janne 1999) or “knowledge hotspots” (Bathelt et al. 2004). As noted by Florida (2005), such hotspots are the “spikes” in the global knowledge landscape: Highly specialized Marshallian clusters with focused innovative activity or global cities with diversified innovation profiles. Innovation systems in industry verticals are composed of globally dispersed, interlinked knowledge hotspots that are hierarchically ordered (Meyer, Mudambi, and Narula 2011). Lower order hotspots service higher order ones and all are orchestrated by apex or lead clusters. Silicon Valley is an archetypical example of a lead cluster in information technology. Other examples include Hollywood (filmed entertainment), London (financial services), Milan (fashion), and Stuttgart (automotives). MNEs’ networks co-evolve with the global knowledge and innovation landscape, so that their subsidiaries in lead clusters typically serve as centers of excellence for their worldwide operations.

On the one hand, the potential for a cluster to innovate on the basis of global flows of knowledge depends on its sheer number of global connections – the scale of the cluster’s connectedness. On the other hand, the ability of the cluster to benefit, in innovation terms, from these connections, is moderated by the connections’ decision-making locus and network structure – the scope of the cluster’s connectivity. Below, we shall develop a typology of connectivity and discuss how different types of connectivity impact a cluster’s innovation potential.

The Decision-Making Locus within Connections: Personal Relationships vs. Organizational Pipelines

Within connections, we draw a fundamental distinction between two different decision-making loci: individual-based and organization-based. The first and most basic type, personal relationships, is operationalized through autonomous individuals, and includes family relations, friendships, and acquaintanceships. Individuals often leverage the social proximity provided by such connections to seek professional opportunities, and when personal relationships span globally, they facilitate both labor mobility and entrepreneurship across clusters. The best examples of how personal relationships may constitute global connections are those between members of global diasporas (Agrawal, Cockburn, and McHale 2006; Saxenian 2006; Zaheer, Lamin, and Subramani 2009). Affiliated to multiple geographical localities by birth, these individuals leverage not just their family and friends, but also cultural similarities and shared national backgrounds, in building “swift trust” (Meyerson, Weick, and Kramer 1996) that facilitates personal as well as professional relationships across several clusters.

The second type of decision-making locus is operationalized through organizations and has been referred to by Bathelt et al. (2004) as pipelines. These include strategic alliances, joint ventures, or ownership spanning different organizations or organizational branches. Even though pipelines, like all connections, are embedded in the individuals who build and operate them (indeed, organizations sometimes hire individuals because of the valuable connections they hold), they differ from personal relationships in important ways. The organizational hierarchy provides particular governance and incentive mechanisms. Consequently, organizational routines are designed to ensure that employees do not act autonomously when operating pipelines. The leverage of such pipelines is aligned with the strategy of the parent organization, not with the interests of individuals.

Organizations maintain pipelines in order to move resources across geographical distance. Mature clusters in advanced market economies are characterized by having a multitude of pipelines built by indigenous organizations. By contrast, in clusters in less developed economies, pipelines are typically brought in by foreign MNEs and operated by local subsidiaries. International business scholars have studied the latter type of pipelines for decades (Mudambi and Navarra 2004; McCann and Mudambi 2005; Beugelsdijk, McCann, and Mudambi 2010). Compared to strategic alliances, subsidiary ownership reduces risk and spatial costs, but also incurs higher setup costs (Gupta and Govindarajan 1991). The general potential of an MNE subsidiary to spur innovation in a cluster depends, on the one hand, on the type of its mother organization. Leading MNEs in the global competitive environment typically undertake more innovative activities in their subsidiaries, while lagging MNEs often undertake routine innovative activities in their subsidiaries by replicating activities performed elsewhere in their organization (Cantwell and Janne 1999). On the other hand, the potential of a subsidiary to spur innovation also depends on the subsidiary’s own type. The literature distinguishes between subsidiaries with competence-exploiting mandates focused on routine replication and local adaptation and unlikely to spark a great deal of local innovative activity and subsidiaries with competence-creating mandates charged with developing new competencies for use beyond the local market, likely to be associated with significant inflows and outflows of knowledge (Cantwell and Mudambi 2005).

The Network Structure of Connections: Centralized vs. Decentralized

A cluster’s configuration of global connections can be seen as a particular social network where local and non-local individuals and organizations are the nodes and their connections are the ties. The structure of this social network impacts how a cluster benefits from local individuals’ and organizations’ global connections (Lorenzen and Mudambi 2013). Of particular importance is the connection network’s degree of centralization (Barabasi and Albert 1999; Wasserman and Faust 1994; Watts and Strogatz 1998). In a social network, the more relations a node has to other nodes, the higher that node’s centrality.³ When a few nodes with high centrality dominate a network, the structure of the network is centralized. By contrast, when all nodes have a comparable number of ties, so that none is central, the network is decentralized.

Using this terminology, when flows into and out of a cluster are mediated by just a few gatekeepers, the structure of the connection network is centralized (Bell and Albu 1999; Bathelt et al. 2004; Giuliani 2007; Graf and Krüger 2010; Malecki 2010). Strong local individuals or organizations that hold many global connections typically also come to dominate the network of local connections. This makes peripheral individuals or organizations in the cluster depend on the central individual or organization for access to each other as well. For example, “industrial complex” type clusters with dominant local organizations (McCann and Mudambi 2005) are likely to progress into centralized connection network structures. This is because the locally dominant organizations receive most of the overtures from organizations outside the cluster seeking to establish inward connections and they concomitantly develop experience in creating and maintaining all types of connections (Wolfe and Gertler 2004).

Conversely, when the structure of a connection network is decentralized, flows occur directly between many nodes, all with comparable centrality. Decentralization of global connections is likely to co-evolve with direct relations among local individuals and organizations as well. The example to consider here is non-hierarchical, trust-based clusters without dominant resident organizations, such as the “pure agglomeration” (McCann and Mudambi 2005) or “industrial district” clusters (Markusen 1996).

Figure 10.1 illustrates these two different stylized extremes. As the figure shows, in a completely centralized connection network, all external nodes are connected to a central local node and all flows to and from the cluster nodes must take place through brokerage facilitated by this node (Burt 1992). This individual or organization can effectively operate as a gatekeeper or “guru” (Mudambi, Oliva, and Thomas, 2009), shaping the nature of resource flows into and out of the cluster, as well as capturing the bulk of the rents that arise from these flows. In contrast, in a completely decentralized connection network, every local node is connected directly to external nodes.

Global Connectivity: Archetypes and Examples

A cluster’s connectivity denotes its global connections’ particular configuration of decision-making loci and network structure (Lorenzen and Mudambi 2013). The stylized types of connections’ decision-making loci and network structure developed above combine into four archetypes of cluster connectivity. These are sketched out in Figure 10.2.

The first connectivity archetype is centralized personal relationships. A real-world example is clans – person-based relationships revolving around a few central elders or other leaders (Boisot and Child 1996). Clans may connect different clusters and facilitate the transfer of knowledge and other resources to dominant clan members. Examples of locally dominant clans are often found in East Asia, particularly within Chinese business networks where extended family connections across cities or national borders facilitate privileged access to labor, capital, information, credit, and markets (Todeva 2006; Yeung 1997). Even if this connectivity type has a very long history, it is still prevalent in clusters located in countries with strong family traditions.

Decentralized personal relationships represent the second connectivity archetype. A good example is the relations maintained between a diaspora of migrants from a cluster and their families and friends who remain in the home country. “Global Argonauts” (Saxenian 2006) are important operators of this connectivity archetype; they use their commonalities of language and culture as well as experience and contacts from one cluster in order to start up business ventures in another. Decentralized personal relationships facilitate the transfer of resources across clusters, both knowledge and remittances, often fueling business activity in a home cluster with resources generated abroad (Vaaler 2011). This connectivity type has traditionally been driven by waves of immigration. During the last half century, it has become prevalent in still more clusters due to new transportation and communication technologies allowing diasporas to stay in touch.

The third connectivity archetype is centralized pipelines. This may arise when “flagship firms” (Rugman and D’Cruz 1997) act as gatekeepers between other cluster organizations and global markets, or when “anchor tenants” (Pashigian and Gould 1998; Agrawal and Cockburn 2003), attract customers or investments to a cluster from global markets.⁴ In mature clusters in advanced market economies, flagship firms and anchor tenants are often resource-rich indigenous organizations. However, in nascent clusters, flagships and anchor tenants are likely to be local subsidiaries of foreign MNEs that have a strong and dominant role due to the lack of development of other local organizations. It is worth noting that network position does not always follow from firm size. Rather, it is often a result of a dominant technological capability. As noted by Bell and Albu (1999), smaller organizations or public institutions can function as local technology gatekeepers. This connectivity type is limited to clusters that participate in global trade and production. The most connected such clusters are those home to the most active MNEs.

The fourth archetype, decentralized pipelines arises, for example, when high-technology indigenous SMEs set up “international new ventures” (Mudambi and Zahra 2007), or when MNE subsidiaries collaborate broadly with peers in a cluster, for instance, through R&D partnerships. In this case, MNE subsidiaries do not occupy central network positions. This may be a result of a strategy to be present on an equal footing with local firms in order to gain access to horizontal local knowledge-sharing (Lorenzen and Mahnke 2004), or it may be a result of the cluster being home to a large number of equally big MNE subsidiaries, so that none of them is able to take on a central role as flagship firm or anchor tenant. Since each of the MNE subsidiaries provides a pipeline to global markets, the structure of the connection network is decentralized. Driven by globalization of production and innovation activities, this connectivity type is spreading to more and more clusters that develop small-scale high-tech activities.

Of course, what we observe in the real world are mixed forms rather than pure archetypes. Real-world configurations typically include aspects of both individual-based and organization-based decision loci, and network structures with greater or lesser extents of centralization. For example, a family-based business group is a centralized organizational structure in which individual decision locus is crucial; thus, centralized personal relationships play a key role (Khanna and Palepu 2000; Yeung 1997). Universities provide another example. Their two main mission and activity areas, knowledge creation (research), and knowledge dissemination (teaching), have different types of connectivity. When teaching activities span between clusters, this is largely accomplished through bilateral alliances between universities at the organizational level, that is, through pipelines, which may well be centralized. In contrast, cross-cluster research collaboration takes place, in the main, between individual researchers, that is, through decentralized personal relationships.

As mentioned, a cluster’s connectedness denotes the scale of its global connections, while its connectivity denotes the scope of a particular configuration of these connections. While connectedness represents a potential for innovation, it is connectivity that determines whether and how this potential will be realized. This is because the decision-making loci and network structure of the connections of a cluster signify how local individuals and organizations “hook on to” the global production and innovation system. Before we discuss how different forms of connectivity facilitate innovation, let us consider some general ways in which decision-making locus and network structure influence innovation.

The Decision-Making Locus within Connections and Innovation: Emergent vs. Strategic

Innovation through personal relationships is likely to be emergent and span diverse (but often related) technologies and industries, while innovation through pipelines is likely to be strategic and focused in terms of technology and industry.

Creative individuals are more focused on knowledge outcomes than firm performance outcomes (Mudambi and Swift 2009), and hence, connections with individual-based decision-making facilitate innovation activities that are emergent from the bottom up through interactions of many autonomous individuals who experiment, collaborate, merge, learn, spin-off, and even steal from each other. Innovation through personal relationships is often “about not recognizing boundaries or limitations – exploring for its own sake” (Gurteen 1998). Because individuals have diverse objectives, backgrounds, and knowledge bases (Doz, Santos, and Williamson 2004), innovation through personal relations is likely to cross the boundaries between organizations, industries, and technology fields. Such boundary spanning has been demonstrated to be a key aspect of exploratory innovative activities that involve distant rather than local search in technological space (Todeva 2006; Lamin and Dunlap 2011). Hence, personal relationships offer a high scope for diversity and knowledge recombination.

In contrast, connections with organization-based decision-making are likely to target organizational performance objectives, such as the development of marketable products and processes with relatively short time horizons. Thus, innovation emerging through pipelines tends to be strategic. Strategic innovation activities managed and owned by profit-oriented organizations with an explicit innovation strategy are often relatively narrow in scope. Successful organizations tend to link closely to suppliers, buyers and partners, that is, they tend to be very industry-focused (Christensen 1997). This implies that innovation activities through pipelines are likely to be characterized by relatively less technological diversity (Feldman and Audretsch 1999).

The Network Structure of Connections and Innovation: Appropriation vs. Participation

The main share of innovation through centralized connection networks is likely to be appropriated by central individuals or organizations, while innovation through decentralized connection networks is likely to have spillovers because individuals or organizations participate in innovation processes with a comparatively higher degree of reciprocity.

Economic geography as well as international business studies often refer to the innovation effects of social networks, but rarely apply insights from the social network analysis literature.⁵ A general insight in the latter literature is that in any network, central nodes have privileged access to resources (e.g., knowledge and capital) flowing through network ties. On the one hand, this means that central nodes are equipped to perform tasks of benefit to all other nodes in the network, and to allow other nodes to enjoy some innovation-related advantages of being in a “small world” (Watts and Strogatz 1998). In the context of a cluster, central individuals or organizations can, for example, coordinate local value chains, bringing down transaction costs for all participants. They may also undertake activities of risk-taking and the attraction of capital and labor (Feldman 2003). Furthermore, if central individuals or organizations participate to knowledge sharing, knowledge may spill over to other individuals and organizations through network ties (McCann and Mudambi 2005). On the other hand, central nodes are also in a position to appropriate resources in a network, excluding other nodes from sharing. This is particularly relevant for knowledge, being the basis of intellectual property and most value creation by individuals and organizations. In a cluster, all knowledge-bearing local individuals and organizations weigh the fears that local knowledge spillovers may find their way to powerful competitors, against potential future gains from the local upgrading of current (and potential future) partners.

Hence, network structure is a key determinant of whether knowledge and other resources become appropriated or are allowed to spill over. There are theoretical arguments and empirical evidence that centralized network structures are likely to be associated with the appropriation of resources by a few central individuals (Granovetter 1973; Burt 1992; Uzzi and Spiro 2005) and organizations (Uzzi 1997). In the case of global connections with centralized network structure, cluster-wide spillovers are likely to be minimized by central individuals or organizations, because they have both the incentives and the ability to exclude others from most of the rents that arise from the connections that they construct. Thus, the lion’s share of rents arising from innovation is likely to be captured by central individuals or organizations. Decentralized social networks lack strong and dominant coordinating individuals or organizations, but are able to facilitate participation in the sharing of resources among a broad category of individuals and organizations without rendering resources proprietary to any of them, and ensuring that the rents generated through network ties are distributed over a wide variety of actors. In a cluster, collaboration in decentralized networks may suffer from high coordination costs (Foss and Lorenzen 2009). However, as resources can be shifted around and have high value in alternative uses, this network structure propagates flexible specialization in value chains (Press 2008), entrepreneurship (Delgado, Porter, and Stern 2010). The associated innovation processes have many spillovers and a high degree of experimentation (Foss and Foss 2002; Lorenzen and Frederiksen 2005). In the case of global connections with decentralized network structure, both inflows and outflows of knowledge, and the fruits of innovation, occur to serve a multiplicity of interests, rather than the sole interests of the central individuals or organizations.

Connectivity and Innovation: Breadth, Depth, and Spillovers

After these general observations on how innovation relates to decision-making locus and network structure, we shall now discuss how each of the four archetypes of connectivity impacts innovation.

First, the connectivity type centralized personal relationships is likely to facilitate innovation across diverse technology areas and industries. Such in-breadth innovation, however, will typically be appropriated by central individuals and will have limited spillovers inside a cluster. Centralized personal relationships in the form of clans and family-based business groups may serve to coordinate local value chains as well as channel resources to local innovation activities (Todeva 2006; Yeung 1997). A clan or family may spend decades nurturing global connections with adherents or kin in valuable clusters around the world in order to extract value from the structural holes in the networks it spans. Consequently, it has obvious incentives to exclude other local individuals who are not family or clan members, maintaining tight control over the knowledge and other key resources that flow through the global connections it holds, and appropriating most of the innovation-based rents arising through them. As a result, with centralized personal relationships, access to critical resources is likely to be limited to members of the dominant group, reducing both the scope for interaction and the diversity of knowledge. Centralized personal relationships are unlikely to thrive in industries that require constant infusions of new knowledge, since family-based business groups and clans are likely to avoid becoming reliant on outsiders. Hence, they typically operate in traditional industries (Yeung 1997; Khanna and Palepu 2000), where they are likely to leverage global connections to enhance and perhaps extend their existing businesses through management innovations. This is unlikely to enhance knowledge spillovers in a cluster.

Second, the connectivity type decentralized personal relationships is likely to facilitate in-breadth innovation across diverse technology areas and industries with substantial spillovers to the multiple individuals participating inside a cluster. When a cluster is connected by numerous autonomous relationships of migrants (e.g., highly educated Global Argonauts), none of these can act as a gatekeeper. The only way for an individual to profit from his or her connections is to maximize their throughput of knowledge and resources, with the resulting innovation spilling over to other participants in the connection network. The results of such connectivity are, for example, knowledge transfer between diaspora members employed in different clusters (Song, Almeida, and Wu 2003; Hoisl 2007) and high levels of technological entrepreneurship (Saxenian 2002, 2006). Innovation driven by decentralized personal relationships can span broadly, across technology areas, because the personal relationships held by global migrants are typically based on education and professional experience (Saxenian 2006), or simply shared national culture. In contrast to the power-based relationships held by clan elders or heads of family-based business groups, decentralized personal relationships are able to tap into a much wider range of expertise and knowledge, and are more likely to operate in high-technology, asset-light industry environments and to implement technological innovations.

Third, the connectivity type centralized pipelines is likely to facilitate innovation within specialized technology areas and industries. This in-depth innovation will typically be appropriated by central organizations and will have limited spillovers inside a cluster. Centralized pipelines typically involve large sunk investments in facilities and training, which the organizations that built them seek to recoup. Hence, while a centralized pipeline can be effective in facilitating innovation, particularly in high-tech industries where R&D is scale-intensive, the organization will also design its pipelines to capture, to the extent possible, the rents from the resource flows through them. For example, if a flagship firm or anchor tenant MNE subsidiary undertakes repeated interactions with endogenous cluster organizations, fears of spillovers will lead it to retain specialized innovative activities within its wholly controlled local subsidiary and limit its local partners to routinized or “commodity” activities (Andersson, Forsgren, and Holm 2002; McCann and Mudambi 2005). This limits the innovative capabilities of the wider population of organizations in the cluster and ensures that it remains at a lower rung in the hierarchy of the global clusters (Meyer et al. 2011).

Fourth and finally, the connectivity type decentralized pipelines is likely to facilitate in-depth innovation within specialized technology areas and industries with substantial spillovers to the multiple organizations participating inside a cluster. We may find decentralized pipelines when several equally central competing organizations “plug into” a cluster (Lorenzen and Mahnke 2004), or where many strong indigenous SMEs each reach out and establish new international ventures. This connection network structure facilitates flexible and shifting R&D collaborations and innovation projects facilitating trial-and-error experimentation. This type of innovation is particularly prominent in consumer goods and entertainment (“cultural”) industries, which is why we see such industries thriving in well-connected, non-hierarchical, trust-based clusters with a diversity of related and interconnected activities and industries (Maskell and Lorenzen 2004; Lorenzen and Frederiksen 2005).

Concluding Remarks

In this chapter, we have provided an overview of the nascent theoretical framework of cluster connectedness and connectivity. This framework is created by fusing the economic geography and IB literatures and applying insights from the literature on social networks. The framework enables us to develop testable predictions regarding the nature of innovative activity and outcomes in different cluster configurations. We have summed up this argument in Table 10.1.

We contribute to the newly arising literature of microfoundations by explicitly recognizing the role of individuals in macro outcomes. However, we go further by incorporating the fact that individuals may or may not be ensconced within organizations. Organizational pipelines may arise from personal relationships, as when corporations hire people because of their connections. Such personal relationships may then be leveraged for strategic purposes. When individuals operate within firms, individualistic decision-making is both constrained and empowered through the routines and resources of the organization. Thus, while decisions in organizations are made by individuals, they are significantly modulated by the organizational context. Thus, the reverse process of individuals leveraging existing organization-based connections for their personal advantage may be limited by routines such as rules and competition clauses.

Finally, our analysis presents an analytical framework within which we can understand the innovative potential and trajectory of specific clusters. Perhaps our most important prediction is that a centralized network structure may have short-term benefits in terms of the rapid establishment of global connections, but these may be outweighed by the longer-term costs in terms of a limited extent and range of local spillovers.

In this chapter we have laid out a theoretical roadmap for studying global innovation at the intersection of economic geography and international business studies. This is a fertile intersection with numerous research opportunities. On the one hand, economic geographers have an extremely comprehensive and fine-grained approach to space, proximity, and place (Laurer, Lorenzen, and Staber 2012). In contrast, international business scholars tend to treat space in a fairly simplistic manner, often using simplistic domestic/foreign binary approaches. On the other hand, over the last several decades, international business scholars have developed extremely thorough and detailed models of global economic organization at both the macro (country) and micro (firm) levels. Economic geographers, with some exceptions, tend to treat firms as homogeneous (or eschew firm-level analysis altogether). This discussion suggests two avenues for significant advances: (1) introducing a more sophisticated view of space, proximity, and place into international business models and (2) explicitly incorporating the organization of the MNE into economic geography models.

We point to a few specific opportunities for understanding the nature of global innovation within these general research avenues. The sub-national context is only recently receiving attention in the international business literature (e.g., Beugelsdijk and Mudambi 2013; Lamin and Livanis 2013). We suggest that much more can be done by integrating the large body of work on clusters and city regions into international business models. Going in the opposite direction, the literature on clusters and city regions would be significantly advanced by recognizing the critical role of MNEs in the establishment, growth, and decline of geographic locations.