Intermediation in European aerospace clusters: a configurational approach

ABSTRACT

This paper investigates which intermediary practices among European aerospace clusters support their diverse membership. Using a configurational approach, we find nine successful sets of practices spanning cluster members with varying levels of internal resources. We unpack important trade-offs, with no single approach to intermediation being ideal in every context. However, the presence of mutual trust is necessary, hinting that its importance in brokered networks has been understated by past research. Furthermore, equifinality exists even for similar types of members, suggesting the need for future research to be attentive to the existence of multiple viable approaches to cluster intermediation.

KEYWORDS:

• intermediation
• clusters
• agglomeration
• qualitative comparative analysis (QCA)
• internal resources

JEL:

• L14
• M0
• R11
• R58

1. INTRODUCTION

The bundling of expertise and other types of resources through inter-organizational networks can aid organizations in fulfilling their goals – from innovation, to increasing profitability (Baraldi et al., 2012; Giuliani, 2013a). These networks can be small or span up to hundreds of members, with larger networks having an increased potential for collective efficiencies but becoming more challenging to coordinate. This has led to the proliferation of intermediaries working to bring actors together (Provan & Kenis, 2007). Whereas some narrowly focus on streamlining science–industry knowledge transfer (e.g., research and technology organizations, and technology transfer offices; De Silva et al., 2018; Villani et al., 2017), others manage relations within diverse incubators and science parks (Etzkowitz, 2008; Hernández-Chea et al., 2021), or much larger geographical clusters (Giest, 2015). The latter type of intermediary is among the most complex, and can service over 100 co-located members that range from resource-deprived small and medium-sized enterprises (SMEs) to resource-rich multinationals (Lindqvist et al., 2013). The last three decades have seen over a thousand intermediated clusters emerge within the European Union, including Aerospace Valley in Toulouse, BICCnet in Bavaria and Maritime London (European Cluster Collaboration Platform, 2019).

In recent years, our understanding of how cluster organizations and other intermediaries emerge and are sustained has improved (Clarke & Ramirez, 2014; Ebbekink & Lagendijk, 2013; Ooms & Ebbekink, 2018), particularly for those prioritizing innovation enhancement among their membership (De Silva et al., 2018; Howells, 2006). In a similar vein, there is increased attention to the impact of intermediaries’ characteristics and practices (Mueller & Jungwirth, 2016; Schüßler et al., 2013), and to conditioning factors such as the presence of a common goal (Agogué et al., 2017). However, much less consideration has been given to the characteristics of the individual organizations subject to intermediation, with only sporadic acknowledgements of different types of members having different needs (e.g., Vernay et al., 2018; Villani et al., 2017). This is exemplified by the scholarly emphasis on universal success factors such as intermediaries having a well-defined collective aim and critical mass of members (Klofsten et al., 2015; Laur et al., 2012). However, there are hints to significant variation in intermediation characteristics, especially among clusters (Ingstrup, 2013), and to disparities in how different members are affected by them (Villani et al., 2017), casting doubt on the catch-all formulations of best practices that characterize the literature (Bergek & Norrman, 2008; Sydow et al., 2011).

In addition, while the regional economics literature stresses that clusters differ in their contextual conditions and that these heterogeneously affect incumbent organizations (Frenken et al., 2015; Giuliani, 2007), little is known about how this combines with varied intermediation practices. This knowledge gap is exacerbated by the related, yet distant (Schmitz & Nadvi, 1999; Yström & Aspenberg, 2017), intermediation scholarship largely ignoring cluster-specific conditions: from their geographical composition, to the networks that individual members maintain, and the institutional frameworks in which they are embedded (Porter, 1998). Yet, both the internal characteristics of member organizations and the combinations of cluster conditions that they are subject to shape the context in which intermediation occurs, and the demands placed on it. To illustrate, members with strong internal resource bases may benefit from their extra-cluster networking being facilitated, especially when they are part of a small and specialized cluster with few relevant local networking opportunities (Fitjar & Huber, 2015; Huggins & Thompson, 2014). However, weaker organizations could instead be assisted by intermediaries prioritizing intra-cluster networking, which is not only less costly than maintaining distant partnerships but can complement weaker members’ knowledge bases (Boschma, 2005; Huggins et al., 2012; Laur et al., 2012).

In other words, it is likely that the success of cluster intermediation is affected by multiple factors coming together in complex arrangements, and that cluster conditions and intermediation practices can have asymmetrical impacts on different types of cluster members. It is even possible that there are multiple pathways to successful intermediation, and conditions in clusters may substitute for one another (Molina-Morales et al., 2019). Geographical diversity may, for instance, provide similar knowledge inputs as extra-cluster networks (Speldekamp et al., 2020b). To accommodate this potential, we adopt a configurational lens allowing for conjunction, asymmetry, and equifinality (Misangyi et al., 2017). Specifically, we address the following research question: Which combinations of cluster conditions, intermediation practices, and internal resources cohere with high goal attainment by cluster members?

Our paper proceeds with a theoretical framework outlining how the conditions in clusters and intermediary practices may heterogeneously impact cluster members based on their internal resource strength. Subsequently, our empirical setting of European aerospace clusters is presented together with our methodological approach: fuzzy-set qualitative comparative analysis (fsQCA). Following the analysis section, we discuss the implications of our study for practitioners and future research.

2. THEORETICAL BACKGROUND

Clusters, defined by Porter (1998, p. 78) as ‘geographic concentrations of interconnected companies and institutions in a particular field’, offer organizations access to complementary external resources. Our interest in this paper is in intermediated clusters, and more specifically the intermediary actions enabling members to make the most of clusters’ geographical, network and institutional conditions, and achieve their individual goals. The intermediaries under study are organized through network administrative organizations (NAOs) – a form of brokerage that is capable of a large array of practices and is ideally suited to servicing a voluntary, diverse membership base with actors ranging from universities and SMEs, to multinationals (Cristofoli et al., 2014; Provan & Kenis, 2007).

2.1. Geography

At the core of the cluster concept is the notion that geographical conditions have a positive impact on member organizations (Cruz & Teixeira, 2010). The cluster literature prioritizes the role of urbanization and localization economies (Jacobs, 1969; Marshall, 1920). Urbanization denotes a diverse concentration of economic activities, coinciding with high-quality infrastructures and general business services. In contrast, localization emerges when many organizations with a similar focus co-locate, resulting in the presence of specialized infrastructures, labour market pooling, and the emergence of specialized services. Furthermore, as knowledge is difficult to fully protect and contain within organizations both urbanization and localization coincide with geographically bounded knowledge spillovers (Beaudry & Schiffauerova, 2009). For the former these spillovers are diverse, and for the latter they are industry-specific, with both types occurring through, for example, labour mobility and chance meetings (Bathelt et al., 2004; Feldman, 1999). Spillovers also spur the creation of spinoffs, which are created by entrepreneurs leaving an existing organization and creating their own, and can thus trigger a self-reinforcing clustering process (Golman & Klepper, 2016).

However, concentrations of economic activity can have negative consequences, or provide no net gain, as they bring congestion, inflated land prices, increased competition for inputs such as labour, and knowledge leakage (Grillitsch & Nilsson, 2017). Cluster studies increasingly suggest that the net effect is shaped by incumbents’ level of internal resources (Hervas-Oliver & Albors-Garrigos, 2008). Three competing hypotheses exist on what level of internal resources allows organizations to benefit most from the geographical conditions in clusters (Frenken et al., 2015). The first suggests that it is organizations with strong resources, as they have the greatest capacity to absorb knowledge (McCann & Folta, 2011). However, they run the risk of leaking knowledge to competitors (Rigby & Brown, 2015). Instead, it may be organizations with weak resources who benefit most. They suffer little from leaking knowledge, and have the most to learn (Grillitsch & Nilsson, 2017). Alternatively, organizations with moderately strong internal resources may be in the best position to benefit by being able to capture and use spillovers while suffering only slightly from leaking knowledge (Hervas-Oliver et al., 2018). Crucially, and as we will develop further below, members’ level of internal resources and the geographical conditions to which they are subject likely determine which intermediation practices are beneficial.

2.2. Networks

Networks are similarly a fundamental aspect of clusters (Broekel et al., 2021; Porter, 1998). In intermediated clusters there is a brokered, ever-evolving network of members for which intermediaries can facilitate local and non-local interconnectedness (Schüßler et al., 2013; Ter Wal & Boschma, 2011; Yström & Aspenberg, 2017). Networks are especially important for the pursuit of innovation, which may be a goal in itself or a means to achieve another (e.g., profitability or growth), as they can provide access to knowledge and capabilities (Bathelt et al., 2004; Boschma, 2005; Giuliani, 2013a).

The extent to which organizations choose to network with co-located or distant partners, and thus the intermediation practices they would benefit from, relates to matters such as their level of internal resources and the conditions in clusters (Giuliani et al., 2019; Gordon & McCann, 2000; Speldekamp et al., 2020a). Co-location tends to coincide with shared norms and a strong similarity in knowledge bases (Boschma, 2005; Huggins & Thompson, 2014). Additionally, organizations working in the same regional context may encounter similar obstacles, and their ability to meet face-to-face allows for complex, interactive problem-solving and rich information exchange (Giuliani, 2013b). This facilitates collaboration, requiring fewer internal resources to leverage but has the potential to limit the novelty of the accessed knowledge (Fitjar & Huber, 2015). Therefore, engaging with local partners may be useful to weak members, yet unattractive for strong organizations that stand more to gain from working with distant partners (especially in underperforming clusters; Giuliani et al., 2019). Members with a moderate level of internal resources may be able to leverage either. Nevertheless, expansive and diverse clusters are more likely to contain local partners with knowledge that is relevant even to the most knowledge-rich organizations (Fitjar & Rodriguez-Pose, 2011; Menzel & Fornahl, 2010).

Despite having the potential to boost organizations’ performance, maintaining inter-organizational networks is a costly endeavour, especially when engaging with distant partners (Huggins & Thompson, 2014). Furthermore, as the returns are uncertain there is the risk of underinvestment (Graf & Broekel, 2020). Those member organizations seeking to strengthen local networks will benefit from intermediaries organizing activities such as match-making sessions with other cluster incumbents (Yström & Aspenberg, 2017). This exposes them to potential partners, lowering search costs, while decreasing coordination costs due to their shared cluster membership (Bell & Zaheer, 2007). Intermediaries can also facilitate non-local networking, for example, by organizing shared exhibition booths at international trade shows, lowering the cost of entry, and by brokering meetings with external parties. The Aviation Forum in Hamburg (2020), for instance, has such joint stands. Although some intermediaries may be able to balance both, many will focus on either facilitating extra- or intra-cluster networking (Comunian & England, 2018; Human & Provan, 2000), making it likely that whichever they prioritize will lead to different levels of satisfaction among their membership.

2.3. Institutions

The third aspect of clusters that is impactful to incumbent organizations is the quality of the institutional environment, which is both formal and informal in nature (Rodríguez-Pose & Storper, 2006). Formal institutions are shaped by matters such as the quality of regulations and control of corruption, which affect the achievement of organizational goals (Rodriguez-Pose & Di Cataldo, 2015). This is distinct from informal institutions, which relate to the norms and habits that develop over time to create trust between organizations (Moulaert & Sekia, 2003), and coincide with the presence of mechanisms to punish unwanted behaviour (Dei Ottati, 1994). Both types of institutions are geographically bounded due to regional administrative control over laws and their enforcement, and the historical context and economic structure from which norms emerge (Boschma, 2005; Rodriguez-Pose & Di Cataldo, 2015).

Having strong institutions in a cluster can reinforce or diminish the effects associated with geographical and network conditions, and will differentially impact organizations based on their internal resource strength. Under such conditions, the effects of knowledge leakage and labour market poaching are limited (Speldekamp et al., 2020a). Furthermore, strong institutions will facilitate networking, especially when this does not cross institutional boundaries, with risks from, for example, free-riding being diminished (Rodríguez-Pose & Storper, 2006). The quality of institutions may be less important to members with strong internal resources, as they are better able to navigate challenging conditions. However, it could similarly be argued that these organizations have the most to lose from leaking knowledge (McCann & Folta, 2008), and thus benefit the most from mechanisms preventing this.

NAOs (i.e., the intermediating cluster organizations) can be top-down or bottom-up, and their intermediation, shaped by clusters’ institutional conditions, can be experienced differently (Fromhold-Eisebith & Eisebith, 2005). Top-down intermediation is put in place by governments and likely coincides with a high degree of formalization and increased resources, improving the stability of the cluster and its membership, and leading to a higher quality of services (Provan & Milward, 1995). The direct link to governments can result in supporting measures such as the creation of public research centres and subsidies, which are critical in industries such as aerospace (Hickie, 2006). Although weak and strong cluster members can both benefit, it is likely that only the latter type of incumbent has the capacity to influence government decisions (Kerr et al., 2014). Hence, it can be expected that most weak members will prefer NAOs created through bottom-up processes and predominantly or entirely funded by members, which are free of external control and more sensitive to their needs (Fromhold-Eisebith & Eisebith, 2005; Vernay et al., 2018).

2.4. Constellations of conditions

As the above indicates, the conjunction of cluster members’ level of internal resources, and geographical, network, and institutional conditions shape which intermediation practices are most conducive to the achievement of members’ individual goals. We emphasize here how they form complex ‘constellations’, thereby adopting configurational theory (Misangyi et al., 2017). In this vein, we stress the potential for equifinality, that is, that multiple combinations of conditions lead to successful intermediation outcomes, for conditions to substitute for one another, and for them to have asymmetrical effects in different contexts (Fiss, 2011).

Much of the cluster and intermediation literature has ignored this complexity, especially regarding equifinality (Speldekamp et al., 2020b). However, we noted that although strong organizations may typically prefer intermediaries prioritizing the facilitation of networking outside the cluster, they might be able to benefit equally from a focus on local interactions, for example, in clusters with a diverse membership (Fitjar & Rodriguez-Pose, 2011; Giuliani et al., 2019; Maghssudipour et al., 2021; Menzel & Fornahl, 2010). Organizations of average strength may be in an even better position to do so. At the same time, there are hints in the literature that some weak organizations are not only pursuing local networking, but may additionally be able to successfully leverage long-distance collaborations in clusters that are specialized, diverse, and have strong institutions (Speldekamp et al., 2020a). The confluence of these factors could enable learning as they improve weak organizations’ ability to interpret outside knowledge. Therefore, under certain conditions both an extra- and intra-cluster networking focus by an intermediary can serve to support the achievement of organizational goals for all types of members.

There are several conditions with the potential to substitute for one another, and to have asymmetrical effects under different circumstances. Such relationships are progressively common in both cluster and intermediation research (e.g., Boschma, 2005; Villani et al., 2017, pp. 87–88). In our case, there may be a substitution effect where strong informal rules and localized trust may be able to compensate for weaknesses in formal institutions (Saka-Helmhout et al., 2020). In such clusters, bottom-up intermediation likely offers the best results for all members as it is better able to foster the necessary trust to create informal barriers against opportunistic behaviour than are top-down arrangements (Dei Ottati, 1994; Vernay et al., 2018). However, bottom-up intermediation may be suboptimal for strong organizations subject to well-developed institutions. The direct line to the government that top-down clusters offer may be vital for achieving their goals – not only by mobilizing a lobby, but also through the concomitant reputation effects that increase the ease of establishing extra-cluster collaborations (Fromhold-Eisebith & Eisebith, 2005). This suggests causal asymmetry for this type of cluster member, where bottom-up intermediation is positive under poor institutional conditions but negative when these are strong.

3. DATA AND METHODS

As indicated above, this paper’s objective is to shed light on the complexity by which successful member-level intermediation outcomes emerge, and how this relates not just to the effects of cluster conditions but also their intermediation practices, and member organizations’ internal resource strength. To achieve this, we matched our configurational theory with fuzzy-set qualitative comparative analysis (fsQCA). FsQCA is a comparative case analysis methodology that discerns the combinations of conditions leading to the outcome of interest (Fiss, 2011). There can be multiple such combinations of conditions resulting in similar outcomes, which are typically called pathways or configurations. The method thereby accounts for the noted potential of conjunction, equifinality, and asymmetry (Misangyi et al., 2017; Ragin, 2008).

3.1. Data collection

Our research focuses on European aerospace: a highly clustered industry with a large number of intermediaries. These clusters, as well as their members, vary significantly in performance (ECORYS, 2009). We were able to enlist the support of six aerospace clusters through the European Aerospace Cluster Partnership (EACP) – a meta-organization with 45 European aerospace clusters as a member in 2019 (EACP, 2019).

Our primary data was collected through interviews with cluster intermediaries, and we administered a survey to their members. The interviews were in-depth and semi-structured, lasting 80 to 180 minutes, and were conducted between December 2015 and May 2016. They concerned clusters’ defining characteristics, membership, and intermediation practices (see Appendix A in the supplemental data online for the protocol followed). All interviews were transcribed verbatim, and coded using ATLAS.ti 8.0 (2020). We used these codes to construct the conditions capturing clusters' intermediation practices, on which we elaborate in the set calibration section. Furthermore, the interviews informed our survey instrument as we asked the cluster intermediaries to reflect on the questions we wanted to pose to their members. It is noteworthy that, in addition to the six clusters that form the empirical focus of this paper, we interviewed four other clusters early in our research efforts to gain a better understanding of the empirical context. These were Aerospace Valley; Farnborough Aerospace Consortium (FAC); the Flemish Aerospace Group (FLAG); and Lazio CONNECT. However, we were not able to reach their membership, and hence excluded them from the rest of our study.

Our second instrument, the survey, was principally based on the 2012 edition of the Community Innovation Survey (Eurostat, 2019a). Our questions concerned cluster members’ organizational characteristics, research and development (R&D) activities, and goals, as well as the extent to which they perceived that their cluster facilitated the achievement of these goals. We also asked respondents to reflect on the informal institutional environment, and to list a maximum of 10 local and 10 non-local innovation partners to capture their networks – the former through a roster-recall, and the latter in a freeform format (Ter Wal & Boschma, 2009). To ensure that our questions were clearly formulated, we did not only present them to the cluster intermediaries but also piloted the questionnaire. This pilot was performed at a cluster that did not participate in the full study, namely the West of England Aerospace Forum (WEAF). We attended one of their showcase events in 2016 and asked eleven of their members to complete the questions, after which we were confident in their quality.

After this phase, we disseminated the questionnaire among the membership of the six participating clusters. (The full questionnaire for one of the participating clusters is presented in Appendix B in the supplemental data online.) In five clusters, the intermediaries sent out links to the online survey via email, and in the remaining case we directly contacted members. Both did not lead to the desired response, resulting in most of the completed questionnaires being collected in person by the first author, for example, at member gatherings and trade shows. In total, we gathered 69 complete responses between October 2016 and July 2017, reflecting the challenging nature of collecting data about organizations’ strategic assets and networks (Baruch & Holtom, 2008).

Approval for the primary data collection described in the above was obtained from Radboud University’s Scientific Advisory Committee, and all participants explicitly consented.

In addition to using our survey and interviews, we added secondary data from three sources to capture clusters’ geographical and formal institutional conditions. For the first, we used Eurostat (2019b) demographic statistics and EACP (2015, 2016) documentation. For the second, we adopted the Quality of Government index developed by Charron et al. (2015).

Table 1 lists the clusters fully participating in our study, and details the response rates for our survey. Note that seven organizations filled in the survey twice, and their least complete responses were filtered. Similar to responses missing vital information, these are not counted in the final response rate.

Table 1. Participating clusters: cluster size and survey response breakdown.

Cluster	Corresponding NUTS region	N cluster members	N usable survey responses
Hamburg Aviation	DE6 (NUTS-1)	133	22
Niedersachsen Aviation	DE9 (NUTS-1)	306^a	16
LR BW	DE1 (NUTS-1)	85	11
NAG	NL (NUTS-0)	108	13
Madrid AeroSpace	ES3 (NUTS-1)	64	4
DAC	ITF3 (NUTS-2)	164	3
Total		860	69

Note: ^aDue to this cluster’s lack of formal membership, this is an estimate based on our interview with the cluster intermediary and EACP (2015, 2016) documentation.

3.2. fsQCA: features and calibration

Although we refer to the methodological overview written by Schneider and Wagemann (2012) for a detailed account of fsQCA, it is useful to highlight its core characteristics. FsQCA both ideally fits our configurational theory, and allows for the exploration of necessary and sufficient conditions leading to successful cluster intermediation (Fiss, 2011). It is noteworthy that fsQCA can analyse a relatively large number of conditions even with small datasets, where 12 to 50 cases commonly include analyses of four to eight conditions (Greckhamer et al., 2013, p. 54). This is a condition our fsQCA meets. Finally, fsQCA can combine qualitative and quantitative data (Misangyi et al., 2017).

These qualities are connected to fsQCA’s set-based nature where cases have different degrees of membership for conditions such as the presence of strong urbanization and localization, as well as the outcome – that is, the presence of successful intermediation. Subsequently, their sub- and superset relationships are analysed (Ragin, 2008). Continuous data are transformed into sets through three anchor points denoting full membership, the crossover point (the maximum point of ambiguity), and full non-membership (Ragin, 2008) – thresholds denoted by scores of 1, 0.50 and 0, respectively. For sets sourced from our survey data, and that is measured via four-point Likert items, we assigned discrete set membership scores of 1, 0.67, 0.33 and 0.¹ For qualitative data we used similar values, adding only the point of ambiguity of 0.50, following best practices in the QCA scholarship (Basurto & Speer, 2012). Table 2 summarizes our operationalizations, set calibrations, and the data relied upon. Note that we conducted our primary analyses on three samples: cluster members with weak, moderately strong and strong internal resources. This enabled us to untangle how different types of members can benefit dissimilarly from intermediation practices. We thereby accounted for the competing arguments in the literature pertaining to the heterogeneous effects of geographical conditions on organizations, which may extend to network and institutional factors (Frenken et al., 2015). More detailed information on these matters is available in Appendix C in the supplemental data online, which considers the measurements and calibration step by step, explaining how we arrived at the anchor points presented in Table 2 through theoretical, external, and sample information. Further, we included an additional analysis of the full sample with only the identification of strong internal resources (or absence thereof) in Appendix D in the supplemental data online. Although this analysis lacks the granularity of the main analyses reported in the results section, it strongly reflects those findings. Finally, please note that all calibrations and subsequent analyses were performed using the Fuzzy-Set/QCA 3.0 program (Ragin et al., 2017).

Table 2. Set operationalizations and thresholds overview.

				Fuzzy-set calibration
Causal conditions		Description	Sources	Fully in	Crossover		Fully out
Cluster conditions	Geography: urbanization	Population density percentiles in 2013 at the NUTS level of the cluster	Eurostat (2019b) statistics	90	75		62.5
	Geography: localization	N cluster member organizations in 2016/17 (at the time of our survey)	Interviews and EACP (2015, 2016)	210	160		91
	Networks: local	N partners an organization has within the cluster with which it engages with the primary goal of exchanging or acquiring knowledge to support their innovative activities in 2016/17 (at the time of our survey)	Survey: roster-recall	6	4		0
	Networks: non-local	As above, but pertaining to innovation partners outside the cluster	Survey: freeform	3	1		0
	Institutions: formal	Summary valuation of regional service quality, impartiality and lack of corruption interpolated with the Worldwide Governance Indicators in 2013	QoG Index developed by Charron et al. (2015)	80	65		50
				Discrete calibration
	Institutions: informal	Extent of shared mutual trust with other organizations in the cluster in 2016/17 (at the time of our survey)	Survey: four-point Likert	To a great extent		1
				Somewhat		0.67
				Very little		0.33
				Not at all		0
Cluster intermediation practices	Intermediation: external network focus	Extent to which cluster intermediaries prioritize building external networks over internal networks in 2015/16 (at the time of our interviews)	Interviews	Based on clusters’:		Value between 1 and 0^a
				1) Membership growth
				2) Value proposition
				3) Non-local dealings
	Intermediation: top-down nature	Extent to which cluster intermediation is top-down rather than bottom-up in 2015/16 (at the time of our interviews)	Interviews	Based on clusters’:		Value between 1 and 0^a
				1) Origins
				2) Financing
				3) Concentration of power
Outcome	Successful intermediation	Extent to which clusters facilitate members in achieving their goals in 2016/17 (at the time of our survey)	Survey: four-point Likert	To a great extent		1
				Somewhat		0.67
				Very little		0.33
				Not at all		0

Note: ^aThat is, 1 or 0.67 (entirely and mostly in the set), 0.50 (cross-over), 0.33 or 0 (mostly and entirely out).

Based on the level of member organizations’ internal resources, denoted by the percentage of sales invested in internal research and development (R&D) between 2013 and 2015 (collected via our survey), we conducted analyses on three samples (split through terciles): weak (< 2%), moderately strong (2–10%) and strong (> 10%) members. Appendix C in the supplemental data online details the calibration procedure.

3.3. Understanding causal pathways

To uncover patterns between conditions and the outcome of successful intermediation, fsQCA relies on truth tables. We created one for each of our samples of weak, moderately strong, and strong organizations. These tables consisted of 2^k rows, with k denoting the number of conditions (Fiss, 2011). In a truth table, each row corresponds to different combinations of conditions that are either true or untrue – that is, there is set membership or non-set membership. Moreover, each row has a particular consistency and holds several cases. In fsQCA, the rows of truth tables are reduced by determining a minimum number of cases that should have the expressed solution and a minimum consistency with which cases correspond to this. Our minimum solution frequency was two, fitting our modest sample size, and the required consistency was 0.90 (Schneider & Wagemann, 2012). We chose to set the solution frequency at 2 instead of the more commonly used threshold of 1 (e.g., Santamaria et al., 2021; Schneider et al., 2019) to eliminate extreme cases from our analyses.²

It is important to note that while performing the standard checks (Schneider & Wagemann, 2012), we found evidence that strong informal institutions are necessary for the outcome (reported in Appendix E in the supplemental data online). Moreover, enabled by fsQCA’s causal asymmetry, in one subsample this condition also showcased some necessity for the absence of the outcome. As a consequence, we ensured we did not violate any assumptions on logical remainders during the minimization process described below, and further employed PRI (proportional reduction in inconsistency) scores – an alternative consistency measure that helps eliminate inconsistent solutions. We set the PRI threshold at the commonly used value of 0.70 (Pahl-Wostl & Knieper, 2014). Our reduced truth tables contained 11, four and 15 cases for the subsamples of weak, moderately strong and strong organizations, respectively.³

The truth tables were minimized through the Quine–McCluskey algorithm (Ragin et al., 2017). We used the parsimonious and intermediate solutions, which take a different approach to accounting for logical configurations not in our dataset (Fiss, 2011). The parsimonious solution uses all simplifying assumptions – that is, all logical remainders. Instead, the intermediate solution relies only on easy counterfactuals, which means that it adds causal conditions to configurations known to produce the outcome. It does not employ difficult counterfactuals, which would entail removing conditions from configurations.

Through this approach, we identified three types of conditions: those that are core to a configuration (often referred to as a pathway), those that are peripheral, and those that do not matter (Schneider & Wagemann, 2012). Conditions appearing in both the intermediate and parsimonious solutions are core, and are denoted in output tables by a large full circle when it is their presence that matters (⬤) or a large crossed out circle when their absence is key (⊗). Conditions that are only part of the intermediate solution are peripheral, and are indicated by smaller circles (● and ⊗). Unimportant conditions appear as blank spaces in fsQCA output tables, and neither their presence nor absence matters for the outcome in question.

In our main results reported in the next section we prioritize cluster intermediation practices as these are the focal point of our paper, examining how they cohere with the presence or absence of core cluster conditions. We report both the coverage and consistency values of our configurations, which indicate the proportion of cases that are included and the extent to which the data matches the relationships presented (Schneider & Wagemann, 2012). For coverage, two statistics are displayed: a raw and unique value. The raw value relates to the total share, and the unique value only to the share covered exclusively by a configuration. Finally, we report the overall solution coverage and consistency which are based on combinations of configurations, and the total number of cases covered.

4. RESULTS

The descriptive statistics of our three selections of cluster members, that is, those with weak, moderately strong, and strong internal resources, are presented in Table 3 (the supplementary full-sample analysis is reported in Appendix D, and correlations are presented in Appendix F, both in the supplemental data online). This table reveals substantial variation over the conditions and the outcome under study within the selections of cluster members, and limited variation across the groups of members as evidenced by the non-significant F-statistics in the reported analysis of variance (ANOVA) test. For instance, it does not seem to be the case that, on average, organizations with stronger internal resources are located in clusters with different geographical characteristics or even significantly engage in more networking. Instead of suggesting endogeneity, our results below therefore point to differentials in how particular sets of circumstances are experienced by the different cluster members and are shaped by the intermediation they receive. Regarding the latter, it is notable that most cluster members are ‘somewhat’ aided in the achievement of their individual goals by their respective intermediary (as denoted by a score of three on our four-point Likert measure).

Table 3. Descriptive statistics.

	Weak				Moderately strong				Strong				ANOVA
Variables	Mean	SD	Maximum	Minimum	Mean	SD	Maximum	Minimum	Mean	SD	Maximum	Minimum	F-value (significance)
1. Successful intermediation	3.00	0.84	4.00	1.00	2.95	0.86	4.00	1.00	2.93	0.68	4.00	1.00	0.05 (0.95)
2. Geography: urbanization	87.38	14.78	97.03	59.82	83.87	14.38	97.03	59.82	81.70	16.31	97.03	59.82	0.82 (0.45)
3. Geography: localization	146.48	81.98	306.00	64.00	145.86	83.70	306.00	64.00	176.67	87.92	306.00	85.00	1.06 (0.35)
4. Networks: local	3.19	3.68	10.00	0.00	4.33	3.77	10.00	0.00	3.93	3.86	10.00	0.00	0.50 (0.61)
5. Networks: non-local	1.57	2.29	7.00	0.00	2.43	3.09	10.00	0.00	1.81	2.95	10.00	0.00	0.52 (0.60)
6. Institutions: formal	67.19	5.46	73.25	56.57	63.16	13.52	73.25	7.65	62.20	16.08	73.25	7.65	0.95 (0.39)
7. Institutions: informal	3.05	0.74	4.00	1.00	3.29	0.56	4.00	2.00	3.30	0.78	4.00	1.00	0.88 (0.42)
8. Intermediation: external network focus	0.52	0.15	0.67	0.33	0.48	0.15	0.67	0.33	0.46	0.13	0.67	0.33	1.46 (0.24)
9. Intermediation: top-down nature	0.41	0.41	1.00	0.00	0.43	0.40	1.00	0.00	0.57	0.39	1.00	0.00	1.14 (0.33)
	N = 21				N = 21				N = 27

Table 4 presents our fsQCA results. This table unpacks the varied ways in which different types of members are facilitated by intermediaries to achieve their goals, from intermediaries being top-down and externally oriented, to bottom-up and prioritizing clusters’ internal networking. We found a total of nine pathways, of which five are related, which we unpack below. We focus on the different intermediation practices that characterize them, and how these combine with cluster conditions and member organizations’ own networks.

Table 4. Configurations for successful intermediation.

		Weak			Moderately strong		Strong
		1	2	3	4a	4b	5a	5b	5c	6
Causal conditions		High-trust, externally oriented	Top-down, unspecialized, externally oriented	Bottom-up, high-trust, externally oriented	Top-down, externally oriented	Bottom-up, internally oriented	High-trust, top-down, externally oriented	High-trust, top-down, internally oriented	High-trust, top-down, internally oriented	Bottom-up, internally oriented
Geography	Urbanization	●	●	●	●	●	●	⊗	⊗	●
	Localization		⊗		⊗	⊗	⊗	●	●	⊗
Networks	Local	⊗	●	●	⊗	⊗		●		⬤
	Non-local	⊗	●	●	⊗	⊗	⊗		●	●
Institutions	Formal			⬤		●		●	●	●
	Informal	⬤	●	⬤	●	●	⬤	⬤	⬤	●
Intermediation	External network focus	●	●	●	●	⊗	●	⊗	⊗	⊗
	Top-down nature		⬤	⊗	●	⊗	⬤	⬤	⬤	⊗
Raw coverage		0.35	0.13	0.22	0.10	0.20	0.23	0.21	0.18	0.16
Unique coverage		0.31	0.05	0.12	0.03	0.12	0.19	0.10	0.07	0.12
Consistency		0.99	1.00	1.00	0.97	0.98	0.96	1.00	1.00	1.00
Organizations in solution		7	2	2	2	2	6	5	5	2
Overall solution coverage			0.58			0.22		0.63
Overall solution consistency			1.00^a			0.98		0.98

Note: ^aRounded from 0.996.

⬤, Core condition (present); ⊗, core condition (absent); ●, peripheral condition (present); ⊗, peripheral condition (absent); blank space = causal conditions may be present or absent.

4.1. Weak members

There are three pathways in which weak cluster members’ individual goals are successfully facilitated. These share that their respective intermediaries prioritize extra- over intra-cluster networking. Nonetheless, there are marked contrasts between the paths. This concerns their empirical relevance, with pathway 1 capturing the greatest share of members, followed by pathways 3 and 2, respectively. More importantly, however, the combinations of conditions leading to the outcome under study differ substantially.

Pathways 1 and 2 differ in the extent to which cluster members have strong inter-organizational networks. At the core of the first configuration is the absence of strong local networks and the presence of well-developed informal institutions. Urbanization is a peripheral, present condition – a role it also takes on in pathways 2 and 3. In such high-trust environments, there is likely an abundance of diverse spillovers, and thus localized learning without the use of networks. This may open up possibilities for weak members to focus on building networks with organizations outside the cluster when their intermediaries facilitate this. Although organizations in pathway 2 also appear to benefit from being in externally oriented clusters, they already have comprehensive local and non-local networks. Their presence is peripheral to this second configuration. At its core is the top-down nature of the intermediary, and the absence of strong localization. This indicates that there may be few relevant co-located partners in such unspecialized clusters, and that these experienced networkers are looking to interface with non-local organizations instead. By enhancing their reputation, being member of a top-down cluster may significantly improve their ability to do so.

In contrast to the first two configurations, the clusters in pathway 3 are bottom-up. This is core to the solution in conjunction with the presence of strong formal and informal institutions. This configuration indicates that bottom-up intermediation may only be effective with extensive localized trust and formal safeguards. This type of intermediation is likely responsive to members’ needs under these conditions. Strong local and non-local networks are peripheral, indicating that the cluster members in this pathway are leveraging their internal and external embeddedness to most benefit from the cluster.

Based on this, we propose the following:

Proposition 1. The individual goal attainment of weak cluster members with few local network partners and supported by strong informal institutions can be facilitated by intermediaries prioritizing external networking.

Proposition 2. The individual goal attainment of weak cluster members that are located in an unspecialized cluster can be facilitated by intermediaries prioritizing external networking and governing top-down.

Proposition 3. The individual goal attainment of weak cluster members supported by both strong formal and informal institutions can be facilitated by intermediaries prioritizing external networking and governing bottom-up.

4.2. Moderately strong members

Two pathways – 4a and 4b – exist for members with moderately strong internal resources. Despite covering relatively few cases, with an overall solution coverage of 0.22, these configurations are theoretically significant as their mode of intermediation is mirrored. Pathway 4a is characterized by a top-down nature and external networking focus, and 4b is bottom-up and internally oriented. In both pathways, members lack extensive non-local networks, indicated by their core absence. They share the peripheral presence of strong urbanization, and informal institutions, and the peripheral absence of significant localization, and local networks. In addition, well-developed formal institutions fulfil a supporting role in pathway 4b.

As these two pathways have commonalities in cluster conditions, they suggest that member organizations with moderate internal resources can benefit from two very different combinations of intermediation practices. Although urbanization is not core to configuration 4b, its peripheral presence hints that an intermediary focus on local networking may only be useful for members with a moderately strong internal resource strength when there is a diverse pool of co-located potential partners. The fact that this coincides with bottom-up intermediation alludes to the importance of cluster intermediaries being responsive to members’ needs. In path 4a, the conjunction between top-down intermediation and external networking suggests complementarity, boosting NAOs’ ability to facilitate the creation of extra-cluster networks among their membership.

Consequently, we propose:

Proposition 4. The individual goal attainment of moderately strong cluster members with few non-local network partners can be facilitated by intermediaries prioritizing (i) external networking and top-down governance or (ii) internal networking and bottom-up governance.

4.3. Strong members

Our analysis uncovered four pathways in which intermediaries successfully support their strong membership, with three of them being related. The three related solutions, that is, 5a–5c, are comparably common as denoted by their raw coverage values – though 5b and 5c display more overlap with other pathways than 5c does with its higher unique coverage. At their core is the top-down nature of the intermediary in conjunction with strong informal institutions. However, these high-trust, top-down clusters differ in many of their peripheral conditions. Whereas solution 5a features extensive urbanization, and the absence of localization, this is mirrored in its two related pathways. Only solution 5b features strong local networks, and where non-local networks are absent in 5a, they are present in 5c. Moreover, formal institutions is a ‘don’t care’ condition in pathway 5a, but present in 5b and 5c. Crucially, whereas intermediaries prioritize external networking in the first of these three pathways, 5b and 5c have intermediaries with a local networking focus. Although these three related pathways are complex, their commonalities are noteworthy. Even for the strongest of members, it appears that institutions are important – and especially the informal institution of trust. This is supported by our supplementary analyses on the necessity of this measure of localized trust, which increases with members’ internal resource strength (see Appendix E in the supplemental data online).

Pathway 6 is distinct from the three described above and less common. It has members maintaining strong local networks at its core, and the intermediaries are bottom-up and have an internal networking focus. In this final pathway, members having extensive non-local networks is a peripheral present condition, as are formal and informal institutions. Given this configuration, it is not surprising that urbanization is a peripheral present condition while localization is absent. Strong members likely benefit more from engaging with local partners that have a dissimilar knowledge base from their own, and their intermediaries’ practices facilitate pursuing such collaborations. Pathway 6 further underlines the multiplicity of ways in which intermediaries can support their cluster members.

Therefore, we propose the following:

Proposition 5. The individual goal attainment of strong cluster members supported by strong informal institutional conditions can be facilitated by top-down intermediaries prioritizing either (i) external or (ii) internal networking.

Proposition 6. The individual goal attainment of strong cluster members with extensive local networks can be facilitated by intermediaries prioritizing internal networking and bottom-up governance.

5. DISCUSSION AND CONCLUSIONS

Our study has investigated the conditions under which intermediaries support cluster members with varying levels of internal resources in achieving their individual goals. We have considered the geographical, networked, and institutional conditions in clusters, in conjunction with whether intermediaries prioritize external or internal networking, and whether they have a top-down or bottom-up governing nature. By doing so, our study connects insights from the broader discourse on clusters (Porter, 1998) to the diverse scholarship on the effectiveness of intermediation (Howells, 2006; Provan & Kenis, 2007).

Our key contribution to both literatures is establishing the varied ways in which successful intermediation outcomes emerge in clusters, and that different types of members have dissimilar intermediary needs that are shaped by their cluster context. While the intermediation literature increasingly acknowledges the interaction between intermediary practices and conditioning factors such as goal consensus (e.g., Agogué et al., 2017), it has ignored cluster-specific conditions and only hints at heterogeneous member-level outcomes (e.g., Vernay et al., 2018; Villani et al., 2017). Further, although the cluster literature is sensitive to the latter two matters (Frenken et al., 2015; Giuliani, 2007), it has not comprehensively considered intermediation. Through a configurational analysis of six European aerospace clusters and a sample of their membership, we were able to consider all these pertinent factors in conjunction. We unpacked nine pathways in which cluster members with weak, moderately strong, and strong internal resources are supported in the achievement of their individual goals – all characterized by distinct configurations of cluster conditions and intermediation practices. Crucially, the attributes whose presence is central to one pathway are often unimportant or absent in another. For example, one pathway for moderately strong members is characterized by top-down, externally oriented intermediation, whereas another for strong members has intermediaries with a bottom-up governing nature that prioritize intra-cluster networking. Overlaps exist as well, with these two pathways being enabled by the peripheral presence of urbanization. This coincides with access to a diverse pool of co-located, potential partners that is especially critical for resource-rich organizations that may otherwise seek to build non-local networks (Fitjar & Rodriguez-Pose, 2011; Menzel & Fornahl, 2010).

As the presence of nine pathways indicates, there are not only asymmetries between how types of cluster members are facilitated in their goal attainment but also a multiplicity of ways in which this can occur for members with a similar level of internal resources. This complexity has thus far largely been ignored in the cluster literature (Molina-Morales et al., 2019; Speldekamp et al., 2020b). We found three successful pathways for weak, two for moderately strong, and four for organizations with strong internal resources. Those for moderately strong members are most illustrative of highlighting this equifinality, with the first configuration being characterized by top-down, externally oriented intermediation whereas this is bottom-up, and internal in the second. When such complexity is not accounted for, the multifaceted ways in which intermediaries can support their membership under varying combinations of cluster conditions may go undetected.

In addition to this core contribution, our study adds to what is known about clusters and their intermediation by challenging three common assumptions made in previous scholarly work. First, we found that trust among cluster participants is necessary, though not sufficient, for effective intermediation. This primacy is surprising given that our clusters are all NAOs. Although previous studies have found trust to be a facilitating factor (e.g., Cristofoli et al., 2017; Trang et al., 2015), scholars typically argue that the effectiveness of such brokered networks is not dependent on it being particularly strong or widespread (Kenis et al., 2009; Mitterlechner, 2019). Our contrasting result is likely attributable to the competitive pressures in clusters (Sorenson & Audia, 2000), amplifying the role of trust.

Second, we did not find a consistent pathway where weak members’ goal attainment was facilitated by intermediaries prioritizing networking within the cluster. This is unexpected, as much of the literature stresses the primacy of intra-cluster networks for the operations of such organizations (Huggins & Thompson, 2014; Morrison et al., 2013; Vernay et al., 2018). Nevertheless, there have been previous studies suggesting that maintaining extra-cluster collaborative partners could be beneficial even for organizations with few internal resources (Speldekamp et al., 2020a). Given the high costs associated with creating such networks, there is an especially strong case for assistance from intermediaries.

The third and final counter-intuitive result was that weak members could be aided by both top-down and bottom-up governing intermediaries. It has previously been suggested that bottom-up intermediation is more suitable for weak cluster members as they lack the necessary resources to steer top-down arrangements to address their needs (Fromhold-Eisebith & Eisebith, 2005; Vernay et al., 2018). There may be conditions under which this holds true, but this was not the case in our empirical context where they can both lead to successful outcomes.

5.1. Practical implications

The multiple, equally viable configurations that we uncovered can be considered as ‘recipes’ for successful intermediation. They inform cluster intermediaries on how to best service their diverse membership, and help to understand the accompanying trade-offs. For example, under certain cluster conditions it is beneficial for moderately strong and strong members to have intermediaries focus on intra-cluster networking. However, as summarized in the above, our results indicate that weak members do not benefit from this. Moreover, the necessity of trust for successful intermediation suggests that where it is lacking intermediaries need to foster it. However, trust is not sufficient by itself and therefore not the only consideration important to the operation of clusters.

For policymakers, our finding that both government-initiated and bottom-up clusters can be valuable to the three types of cluster incumbents we studied is important. Depending on the particular sets of conditions in clusters, either intermediation type can successfully combine with practices focusing on building external or internal networks. Finally, although the strength of formal institutions is not necessary for good intermediation it does frequently take on a facilitating role. Policymakers may therefore seek to strengthen national rules and regulations, and their regional enforcement, in the pursuit of successful cluster policies.

5.2. Limitations

Our paper is a first exploration of how cluster intermediation heterogeneously affects its membership, and can do so successfully through multiple, equally viable sets of practices. It suffers from several limitations. Most importantly, the depth and breadth of our measurements can be expanded upon, and other empirical contexts should be considered. One caveat of our study is its coarse measurement of internal resources through the percentage of sales invested in internal R&D. Despite this indicator being in line with the literature, it can be expanded upon through the consideration of, for instance, employee and managerial skill sets (Cohen & Levinthal, 1990; Lefebvre & Lefebvre, 1998). Including structural characteristics of networks is also promising, as it is a determinant of how knowledge and capabilities are disseminated and accessed (Giuliani, 2013b; Ter Wal & Boschma, 2009). Regarding our empirical context, including different industries may drastically alter the types of intermediation that are most successful under particular sets of conditions. For example, top-down arrangements may be more effective in industries such as biotech, where subsidies likely play an even more integral role in the competitiveness of organizations than in aerospace (Kang & Park, 2012). This link to the government may improve organizations’ access to such funds. Moreover, including other types of clusters than the NAOs we studied, such as those intermediated by a lead organization, or those that are entirely unbrokered, seems equally promising (Provan & Kenis, 2007). A greater diversity in cluster intermediaries would allow future researchers to study a more varied range of intermediation practices.

There are two limitations relating to our methodological approach that are worth highlighting. First, we could not maintain a time lag between each of our conditions and the outcome. Our measures of urbanization and formal institutions pertain to 2013, and internal resource strength to the situation between 2013 and 2015. However, localization economies, organizations’ networks, and the strength of informal institutions related to circumstances at the time of our survey, that is, 2016–2017. The latter is also the time at which the outcome condition of successful intermediation was recorded, whereas intermediation practices relate to 2015–16. The cluster intermediaries we interviewed, and the member organizations that filled in our survey could not provide reliable estimates of these indicators further in the past.

We do not expect the above to have negatively impacted the reliability of our results, given that most of these factors remain stable over time (Briscoe & Tsai, 2011; Levinthal, 2017). Both our geographical conditions, localization and urbanization, are not subject to great change. Clusters tend to grow slowly, and the regions they are part of do not often experience rapid population change. Further, although collaborative networks are not necessarily stable, we focus on innovation networks. Previous studies suggest that these are relatively undynamic and long-term (e.g., Frenken, 2000). Similarly, formal and informal institutions evolve slowly and incrementally (Rodriguez-Pose & Di Cataldo, 2015). Although this is not true for R&D expenditure, which forms the basis of our cluster member selections, we minimized the impact of yearly variations by using a three-year window. This approach is shared by the Community Innovation Survey (Eurostat, 2019a; see also Grimpe & Sofka, 2009), and related popular datasets (Knoben & Oerlemans, 2012), and is preferable to a single-year measurement further into the past (e.g., year 1 of the three-year window) which is sensitive to incidental higher or lower R&D investments.

The second limitation that is important to highlight is that because fsQCA is set-based, different calibrations can alter the results. To limit this second concern, we did our utmost to clearly communicate the reasoning behind our calibrations (see Appendix C in the supplemental data online), and based them on external and theoretical criteria when possible.

5.3. Future research

In addition to the aforementioned limitations, our work leaves several important topics relating to clusters and their intermediation unexplored. Key among these is the role of time. Foremost, after being established cluster intermediaries will need time to build their capabilities (Clarke & Ramirez, 2014). Moreover, cluster networks are subject to change (Giuliani, 2013b; Lindqvist et al., 2013), as are the concentrations of economic activity underlying clusters – albeit at a slower pace. Regarding the latter, clusters go through stages of emergence, growth, and decline between which their characteristics differ (Menzel & Fornahl, 2010). For instance, the emergence stage is characterized by both high uncertainty and economic diversification (Menzel & Fornahl, 2010), and may bring challenges in terms of establishing legitimacy (Human & Provan, 2000) and building trust between members (Boschma, 2005; Villani et al., 2017), while elevating the value of local connections due to the heterogeneous knowledge to which they offer access (Huggins & Thompson, 2014). Hence, a cluster’s life stage likely impacts which intermediation practices are possible and most effective for its membership. The configurational approach undertaken in this paper could be extended to study such change by employing recent methodological advances in QCA (Gerrits & Pagliarin, 2021; Verweij & Vis, 2021). A trajectory-based approach that captures within-case development and compares configurations over time seems especially promising (Pagliarin & Gerrits, 2020). This could give policymakers and cluster intermediaries valuable information on how to facilitate different types of organizations in their operation over time.

5.4. Conclusions

Our study demonstrates that the effects of cluster intermediation are shaped by cluster conditions, and vary for different types of cluster members. We have explored this complexity through a configurational lens, revealing that there are multiple pathways to successful intermediation and not one ‘best practice’ that works over all contexts and for all members of a cluster. There are important trade-offs that intermediaries and policymakers need to consider. We hope that these will continue to be unpacked in future research, enabling better intermediation.

DISCLOSURE STATEMENT

No potential conflict of interest was reported by the authors.

Correction Statement

This article has been corrected with minor changes. These changes do not impact the academic content of the article.

Notes

1. A four-point Likert measurement avoids set ambiguity and increases the ease of survey completion compared with more complex scales, boosting (complete) responses. Our survey pilot suggested that our respondents were adequately able to express their opinion with only four anchors, thus providing the best balance for our purposes (Chyung et al., 2017). We thank one of the anonymous reviewers for raising this matter.

2. We thank one of the anonymous reviewers for bringing this to our attention.

3. We ensured no configurations sufficient for the outcome also predicted its absence. For these analyses we lowered the minimum case count to 1, consistency to 0.80 and PRI to 0.60, broadening the range of configurations captured. Further, we ran two alternative iterations of analyses without (1) the necessary condition of informal institutions and (2) the two clusters with the fewest cases (Madrid AeroSpace and DAC) (see Appendix E in the supplemental data online). We found very similar results. All checks are available from the authors upon request.