Network diversity, distance and economic impact in a cluster: visualising linkages and assessing network capital

2.1 Agglomeration, clustering and networking

From theory, the spatial agglomeration of firms results from different types of external benefits (Marshall, 1921): a more extensive pool of labour may emerge, specialised inputs may be developed and local knowledge flows can be enabled, potentially generating benefits (Brosnan et al., 2016). In practice, substantial variation in the impacts of such externalities on firms has been estimated (Beaudry and Schiffauerova, 2009; de Groot et al., 2015). Prior research indicates the relative impact of inter-firm linkages seems largest for inputs (i.e. input-output or value-chain linkages), with labour linkages the next most important and knowledge spillovers the weakest (Ellison et al., 2010). Notwithstanding the range of related research, Diodato et al. (2016, p. 2) argue that agglomeration impacts remain “poorly understood”.

Knowledge spillovers are problematic as a concept as knowledge itself is so broad (Sammarra and Biggiero, 2008). Ambiguity over the impact of knowledge spillovers and their variation over the industry life cycle (Audretsch and Feldman, 1996) has the added complication that market imperfections also generate unintended spillovers (Scotchmer, 2004).

Attempting to separate out the variety of interacting factors affecting the performance of agglomerations and firms within them is challenging. One means to this end is through investigation of the place-based collaborative networks in which firms are engaged. Specifying different types of collaboration is also an option, through distinct categorisation of technological, market, industry-specific and managerial knowledge (Simonin, 1999). Such inter-organisational knowledge exchanges are recognised as playing an important role in economic growth through endogenous effects (Antonelli et al., 2011; Grossman and Helpman, 1994). The growth of cities, for example, has been explained in endogenous terms “stemming from a city’s capability to invest in a range of intangible assets, in particular human capital” (Huggins, 2016).

Endogenous growth and increasing returns are evident in the agglomeration concept developed by Porter (1990) i.e. “cluster”. In fact, in Porter’s (1990, p. 131) initial formulation of the concept, while geographic proximity was identified as important, the focus was rather on the system of evolved linkages, relationships and processes connecting businesses i.e. “industries related by various links of various kinds”. The processes through which increasing returns might be generated include scale effects, network effects, learning effects and other interaction effects (Arthur, 1988). Within collaborative clusters all sources of increasing returns matter “with the potential for realising scale effects and learning effects magnified by the potential of interaction and networking effects” (Brosnan et al., 2016, p. 508). Research into network contexts, therefore, contributes to understanding the nature, structure and impacts of knowledge flows in a variety of networks.

Across cluster-focussed research Speldekamp et al. (2020) note that while understanding of the contribution of clusters to economic performance has improved, the significant contradiction remains across empirical results with respect to how clusters generate economic growth or innovation (Wolman and Hincapie, 2014). Positive benefits of regional clusters has been reported (Hospers and Beugelsdijk, 2002; Delgado et al., 2014), while ambiguity is evident with other findings of limited positive productivity effects and no strong innovation effects (Duranton, 2011).

In policy contexts, network-based policies that target regional development have been found to generate different results in different contexts (Martin et al., 2011; Falck et al., 2010). Acknowledging that poor networking sets limitations on knowledge flows (Breschi and Lissoni, 2001), there is widespread agreement that developing co-operative relationships across firms and other agents is an important policy goal (Schott and Wickstrom Jensen, 2016; Huggins, 2000). As argued by Graf and Broekel (2020, p. 12), it is necessary to consider “what type of network failures are actually present” to better understand how networks function or exhibit dysfunction and to target appropriate policies.

Over time market failure arguments in favour of traditional industrial policy have been supplemented by network failure rationales for regional cluster-type policies (McCann and Ortega-Argiles, 2013). Network failures have been studied alongside cluster life-cycles (Suire and Vicente, 2009; Brenner and Schlump, 2011) with such failures associated with economic decline. Where markets fail to produce sufficient productive knowledge, policy options target network expansion through innovation incentives, reducing risks in under-appropriation of knowledge, upgrading human capital and improving general knowledge infrastructures (Scotchmer, 2004; Vicente, 2017). Solutions to network failures include increasing network density through, e.g. clustering supports. A more detailed analysis of networking activities sheds light on where weaknesses lie.

2.2 Proximity and network capital

Geographic proximity has been found to offer no immutable guarantee of benefits from agglomeration or local interaction (Bathelt et al., 2004; Tallman and Phene, 2007). Rather, many useful flows of knowledge have been identified through distant rather than local networks (Ceci and Iubatti, 2012; Fitjar and Rodríguez-Pose, 2011). More distant knowledge sources tend to feature for innovation-based links (Davenport, 2005). When benefits from knowledge networks arise, they appear to depend on a range of institutional, cognitive, organisational and social proximities (Boschma, 2005; Tödtling et al., 2011; Fritsch and Kauffeld-Monz, 2010). Membership of knowledge-sharing networks, however, spatially configured, rather than proximity, represents a distinct dimension of network impact.

As envisaged here, network capital includes scope for greater spatial reach than in, for example, the related “social capital” concept, which tends to be built up and concentrated across communities in spatial terms and consists of assets such as goodwill, belonging and social intercourse (OECD, 2001). Where social capital is developed it results in trust and keeps people connected in ways where they can live and work productively together. As Huggins (2010) clarifies, social capital focusses on individual actors within inter-personal networks (following Putnam, 2000) and it may contribute to the creation of socially beneficial resources. However, it is essentially built up without expectation of the results generated from relational interactions. In contrast, network capital is a firm-centric concept, defined as investments in strategic and calculative relations to access knowledge to enhance expected economic returns (Huggins and Thompson, 2015). Investing deliberately or as Williamson (1993) terms it – calculative – involves an expectation of economic return (Belussi and Sedita, 2012).

In addition to the economies associated with agglomeration that generate different impacts for firms, organisations also differ in their ability to convert collaborative interactions into profitable outcomes. Firms’ absorptive capacity – their ability to exploit external knowledge – is complementary to external knowledge acquisition (Laursen and Salter, 2006; Cohen and Levinthal, 1990; Hansen and Birkinshaw, 2007) and considered to be multi-dimensional given the separate processes it encompasses (Volberda et al., 2010; Ferreras-Mendez et al., 2015). This implies that the underlying relational dynamics of firms engaging in collaboration, irrespective of other benefits, generates benefits from interactions and is worthy of investigation (Smith et al., 2020). It also points to the potential in separating out resources used by firms in investing in the creation and maintenance of networks when assessing their benefits. In this way, whether the expectation of economic return from investments in network capital has been fulfilled may be examined.

2.3 Visualising networks: framing assessment of network capital

Visual representation of a network requires linkage types to be identified. Some research has acknowledged the important roles of Input and Output linkages (including e.g. Porter, 1990, 1998b; Sölvell and Protsiv, 2008; Sölvell et al., 2009). In the examination of various “cooperative arrangements” for “knowledge and information sourcing” as the basis for network capital, Huggins and Weir (2007, p. 713) identify linkages to include those with other firms; suppliers; clients; competitors; consultants; R&D laboratories; and higher educational institutions. Porter (1998a, p. 78) highlights the importance of linkages for productivity improvement and identifies partners including “governmental and other institutions, such as universities, standard-setting agencies, think tanks, training providers, and trade associations, who provide specialised training, education, information, research and technical support”. Value chain or transactional approaches can also be considered with linkage categories derived from related literature (Marshall, 1921; Porter, 1998a; Contractor and Lorange, 2002; Leydesdorff, 2012).

Within each linkage, category lies potential interaction that may be organised around activities, actors and/or resources. In short, no standard mapping or visualisation techniques have emerged with some scholars calling for mapping conventions to be used [e.g. Gardner and Cooper (2003) for supply-chain research]. For this research a functional approach was selected, based on the literature. The selected list of linkages, provided below, illustrates that firms may choose to engage in networks with a range of partners, in the business realm and beyond, into governmental and support institutions:

• Government agency linkages (GA): all forms of linkages to government departments and agencies including state support for enterprise; e.g. regional authorities and local gov. agencies.

• Industry association linkages (IA): all memberships and relationships with organisations for collaboration; e.g. IA groups, chambers of commerce, cluster organisations.

• Industry peer linkages (IP): formal and informal relationships with companies in similar or related industries, e.g. related via shared technologies or targeting complementary markets.

• Input linkages (IN): links with suppliers of raw materials, goods and services with a critical impact on the end product or service of the surveyed firm.

• Output linkages (OU): customers and channel sellers – both goods and services. Outputs may be with individual customers or assigned to customer segments and regions.

• Research and development linkages (R&D): include R&D relationships between companies and with academic and research institutes.

• Specialist service linkages (SS): relationships with vendors supplying essential services unavailable in-house (outside of inputs) e.g. services specific to an industry, distribution, IT, consultancy, marketing, financial and legal services.

• Training linkages (TN): including third parties providing specific TN/learning for employees, e.g. relationships with academic institutes addressing skills needs now/for the future.

Establishing networks of linkages – whatever their classification or typology – requires effort and, therefore, investment and indeed maintaining linkages over time similarly has resource implications. It is useful to use the concept of network capital, defined as investments in strategic and calculative relations to access knowledge to enhance expected economic returns (Huggins and Thompson, 2015). In the context of a range of possible business-relevant linkages, they can be identified as a set of separate investments in network capital or disparate types of network capital. This approach permits the identification of such investments by linkage category offering a route to evaluate their business impact by linkage-type and a firm-specific perspective that may be applied in the context of networks and clusters. To date research on the network capital has not adopted such disaggregated approaches focussing on a span of different linkages.

Assessing investment in network capital is complex and may be proxied using the time (Burnham et al., 2003) to account for the time-input required to create and maintain relations, as well as the frequency of such commitments. Active network management may be needed in some cases more than others and competence in managing external relationships has been identified as the basis for a dynamic capability with consequences for performance (Kale and Singh, 2007). Strong and weak linkages may be measured in terms of time input. Further insight is revealed via the dimension of linkage breadth that includes information on whether the linkage involves more than one contact or by the organisational position/status of the contact. This is useful to establish how a consequential interest might be generated from effective network investment. Such measures of input arguably account for activity, however, rather than impact or outcome. Features that impact on the effect of external knowledge on performance outcomes have included the breadth of linkages established and their depth or intensity (Nieto and Santamaria, 2010; Chen et al., 2013).

Further research is required to better understand those linkages that generate the highest returns, according to Love et al. (2014) who focus on the relationship between the number of a firm’s linkages in the context of decisions to innovate and without consideration of the impact of that innovation. On a related note, Lichtenthaler (2005) also argues that the development of measures of success of external knowledge exploitation are needed that take into consideration the strategies, processes and structures through which firms translate it into commercial propositions.

We propose a network impact framework, in Table 1, as a comprehensive means for understanding the performance that organisations achieve from their networking and external knowledge exploitation efforts, that crucially serves to measure economic success or failure. While social capital is associated with individuals’ capacity to mobilise their individual networks, network capital features organisation-centricity. Social capital can “lubricate” flows of knowledge (Vorley et al., 2012) but it does not determine flows of economically useful knowledge (Huber, 2012). In contrast, network capital development is targeted at an economic advantage as an intentional effort in knowledge interactions is considered important for creating superior knowledge through collective processes (Antonelli, 2008). It is important that impact indicators differentiate between linkages generating benefits in terms of e.g. current mission criticality and future-oriented development. Identification of impact across linkage types (following linkage categories) reveals the extent to which targeted investment generates differentiated economic returns, on the basis of linkage-type.

The business impact is separated into two elements of inputs by and outcomes from deliberate networking. Organisational input (OI) encompasses investment and involvement indicators. Organisational investment is measured through both time commitment and the frequency of contacts required to maintain the linkage. Organisational involvement accounts for two additional indicators; the breadth of contacts in the target organisation and contacts’ proximity to decision-making. In this way, we expand on the basic elements relating to network investment outlined in Grabher and Ibert (2006), Huggins and Weir (2007) and Huggins and Thompson (2015), addressing the nature of the underpinning relationships.

To compare OIs to outcomes, importance and intensity indicators are identified. Identifying separate dimensions of network outcomes is necessary to appreciate the network input-output (or investment-impact) relationship. Importance addresses the criticality of the linkage for the organisation’s operations – capturing linkages that might not be mission-critical but still generate benefit. Finally, intensity measures linkage strength and the expectation of future continuity. If the commitment to current organisational activities is compromised by diverting resources into network investment, network impact may be diminished rather than augmented by over-investing in linkages (Lindner and Strulik, 2014).

4.1 From linkage type to impacts: information and communications technologies cluster considerations

Results of V-LINC analysis are presented in tables and visualisations for 571 linkages identified. The observations yielded by the data generation approach and its ordinal nature was suited to non-parametric tests of differences (Wilcoxon rank-sum tests) that permit measurement of differences between linkage types and across geographies. Across cluster firms, we identify the nature of network capital across linkages and geographies considering impact in terms of both input and outcome.

4.2 Linkage identification: type and geography

Figure 1 and Table 2 present network capital by linkage category across geographies. Figure 1 displays the geographic pattern of linkages with locational markers (highlighted pins) in each panel representing the respondent firm sample. Local linkages are focussed on the Cork area with evidence of a linkage highway to Ireland’s capital, Dublin. A range of linkages is evident across European and other international destinations.

Table 2 indicates that across the 571 measured linkages, the most frequent were OU (157 linkages), over 50% higher than the next most frequent category, SSs (97). Within the categories of OU, SSs, IN and IAs (ranked 1 to 4 in Table 2, col. 6), 70% of all linkages constituting this cluster’s network capital are represented. Table 2 distinguishes geographic patterns. Across the sample, 33% of linkages are local (190/571) with 27% national. The remaining 40% are evenly dispersed between European and other international locations.

Local plus national linkages dominate several categories especially GA (98% of linkages), IAs (80%), IPs (63%), SSs (70%) and TN (89%) and R&D (61%). The largest international shares, European plus international linkages, are observed for OU (70%), IN (45%), R&D (39%) and IPs (37%). In one linkage category only did international linkages account for more than 50% i.e. in the highest-frequency category, OU. Local plus national linkages represent the majority in this sample with local linkages dominating national in categories of IAs, IPs, R&D, SSs and TN.

A balance favouring local linkages may indicate potential to benefit from knowledge spill-overs if (as often assumed) proximity reduces search and co-ordination costs [encompassed in investment and involvement (IN) indicators]. We examine the extent to which firms in the cluster generate strong outcomes from the most local linkages (Jaffe et al., 1993; Hasan and Koning, 2017). As widely acknowledged both local and global linkages simultaneously feature in international production and consumption webs and especially for innovation-driven growth it is emphasised that international links and international knowledge sourcing are required (Davenport, 2005; Drejer and Vinding, 2007).

To consider the role of distance in explaining frequency/share of linkages, a series of Wilcoxon signed-rank tests was conducted between the shares of aggregate network linkages for each geography. A statistically significant difference is evident between the shares of local-plus-national compared to European-plus-other international network capital. Estimates (significant at the 1% level) indicate the relative importance of the local-plus-national share, using network density or frequency data, suggesting potential for local spillovers. This was supported by statistically significant differences estimated for each comparison of local-to-European (1% significance), local-to-other International (5% significance), national-to-European (5% significance) and national-to-European (5% significance) linkages, respectively. In contrast, the comparison of local to national linkages suggested that the local/national distance discrepancy does not explain the relative frequencies of these aspects of network capital. Evidence of the importance of local and national collaborations are also favoured in the North of England, Greece and Turkey, in related work on network capital (Huggins et al., 2012). Distance and linkage density appear inversely related to aggregate linkages.

4.3 Geographical differences in network capital

Linkage impacts are presented in Table 3 for each category, for all geographies combined and separately. Measures are organised into two impact bands: High plus Medium linkages (H + M) and Low plus Tenuous linkages (L + T).

Approximately 68% of linkages fall into the higher band (H + M) with a range of 41% for R&D to 92% for OU. The top three linkages of highest frequency (i.e. OU, SSs and IN) also rank highest for impact. While less than one third (32%) of linkages are low or tenuous, substantial shares of are reported in this band for R&D (59%), IN (58%), TN (51%) and IAs (51%).

For geography, more balance across bands is evident for local and national linkages. For local linkages, 58% are in the H + M band with 56% of national linkages. For both international measures, higher shares are evident of 84% and 86%, respectively. Aggregate shares by geography provide no evidence that impact declines with distance.

Focussing on the 190 local linkages, 58% are in the higher impact band. The largest shares of linkages here are outputs (89%), IN (86%) and SSs (67%). Local shares are 39% of SS linkages, 27% of IN and 14% of OU. IP linkages are recorded with the lowest impact in 90% of cases. High proportions of lower impact are observed for local R&D (62%) and IA (52%) linkages.

Of the 155 national linkages, 56% are in the higher impact band, including shares of 85% for OU, 82% for IN and 67% for GA linkages. However, 80% or more of the national IA, IP and R&D linkages exhibited a lower impact.

The 119 European linkages reveal 86% in the higher band. Four linkage types exhibit over 80% of linkages in this band: OU (97%), IA (88%), IN (86%) and IPs (83%). Half of R&D and TN linkages fall into the low impact category. The one linkage of GA was low impact.

Of the 107 international (non-European) linkages, 84% are high impact. All four linkage types featuring above 80% high-impact linkages for Europe demonstrate similar performance here – and all linkages for IA and IPs display higher impact (100%). Over two-thirds of TN linkages are lower impact for this geography.

If distance holds explanatory power for linkages of impact, an inverse relationship would be evident (and statistically significant) between density/frequency and distance for high and medium linkages (390 in Table 3). Wilcoxon signed-rank tests of differences in linkage frequency between geographies were conducted. Tests for comparisons with local linkage frequencies indicated significant differences in linkage frequency for the comparison of local-to-national linkages only (at 5% significance) and local collaborations revealed an average higher frequency of 8% points. Neither local-to-European nor local-to-other international linkages exhibited significance. Therefore, the impact was associated with local rather than national linkages.

Significant differences were also identified for comparisons of national-to-European (at 10% significance) and national-to-other international linkages (at 5% significance). However, results indicated that the impact was associated with a higher frequency of linkages with European and other international locations. On average linkages with European collaborators were 16 percentage points higher and for other international partners, the figure was 19 points. The distance was no hindrance to the impact generated in these linkages.

4.4 Network capital: composition and value-adding impact

Our data allow for differentiating between bonding and bridging network capital [following Putnam (2000) for social capital]. Bonding capital is typified by dense networks with many member ties evident, whereas sparser links characterise bridging capital where new knowledge may refresh available stocks [as with weak ties (Granovetter, 1973)]. The network structure in the ICT cluster reveals bonding network capital for those relatively higher densities evident across four of the eight linkage types examined (OU, SSs, IN and IAs: see Table 3).

Density, as outlined by Vicente (2017) provides limited insight into impact. Density may indicate cohesion within a network but benefits from collaboration may also generate negative lock-in and hinder efforts to attract new members (Crespo et al., 2014). If bridging network capital dominates (in the lower-density linkages e.g. TN, IPs, R&D and GA) such weaker associations may indicate potentially rich opportunities for future brokerage opportunities.

Drilling into linkage quality, we focus on high plus medium linkages (consisting of 68% of all linkages: Table 3) and consider evidence of impact differences across geographies and linkage types. Where firms successfully engage in generating returns from network capital, outcome effects are greater than inputs so net returns are positive. Our measure of network capital allows us to discriminate between those linkages where outcome impact is greater than input, across both geography and by linkage type. Table 4 indicates the outcome and input measures of impact for the eight linkage categories.

In aggregate across the eight linkage categories, outcome dimensions of impact are generally greater than for input, as indicated by shares of impacts scored as higher and medium (73% > 62% from Table 4). This pattern is evident in six categories with only IP linkages demonstrating a greater share of high-and-medium linkages for inputs (58%) than the outcome (51%): for IA links the shares are similar at 55%. There is close alignment between the impacts of inputs into and outcomes from network capital in the case of the most frequent linkage observed, OU, with high shares of linkages at H + M levels (85% and 88%). Substantial misalignment is observed for linkages of IN, TN, GA, R&D and SSs (ranging from differences of 28% to 12%) indicating strong returns to network inputs. Misalignments for the four most frequent linkages (in italics in Table 4) indicate that returns to network capital inputs are amongst the weakest for the most frequent categories (OU, SSs and IAs). Hence, the density or frequency of linkages, does not align simply with impact.

For a spatial perspective, a set of estimations of differences between outcome and input dimensions for each linkage, for each firm, was performed with comparisons across geographies (Wilcoxon signed-rank tests were used). Table 5 presents the results.

Across the four linkage types with the highest frequencies (italicised in Table 5: OU, SSs, IN and IAs) European linkages displayed a statistically significant positive difference between outcome and input dimensions. European linkages vary in their frequency (Table 2) across OU (43%), SSs (13%), IN (17%) and IAs (12%). However, for these linkage types, the impacts on outcome are greater than input impacts.

Even where firms use these linkage types relatively infrequently, the impact of input investments on outcome remains positive. A positive impact is also evident for the IAs linkage. Linkages of OU and IN also exhibit positive returns at the national level. Local linkages with positive returns are identified for SSs and IN. Other international linkages with positive returns are estimated for IAs.

For the less frequent linkages listed in the lower rows of Table 5, more limited evidence is provided in support of positive returns. At both national and local levels, the returns to network capital inputs were positive for TN and GA. In addition, positive returns are observed for IPs from non-European international linkages. The R&D linkage stands uniquely in the absence of positive returns to investment across all geographies.