The visible hand of cluster policy makers: An analysis of Aerospace Valley (2006-2015) using a place-based network methodology

Highlights

• Discusses the rationales of cluster policies based on network failures in innovation processes.

• Builds and disambiguates a dataset of public-funded collaborative projects for Aerospace Valley.

• Develops an original place-based network methodology that overpasses bias and limitations of 1-mode network analysis.

• Identifies the critical determinants of the changing structural properties of the cluster over time.

• Finds that SMEs are the main agents of the cluster structural and technological changes.

Abstract

The paper focuses on cluster policies with particular attention to the role of R&D collaborative incentives in the structuring of knowledge networks in clusters. We disentangle the main network failures in regional innovation systems, and discuss the selection procedures designed by policy makers to foster knowledge collaborations. We draw evidence from the French Aerospace Valley cluster from 2006 to 2015. The case study is based on a dataset of 248 granted research consortia, from which we build 4-cohort knowledge networks that enable us evidencing the evolving structural properties of the cluster over time. We suggest avoiding the bias and limitations of 1 and 2-mode network analysis by developing an original place-based network methodology that emphasizes on structural equivalence and groups’ behaviors. We discuss the results focusing on the convergence degree between the structural properties of the cluster selected by the Program and the policy makers’ objectives. Finally, the methodology allows us to identify the agents of the structural and technological changes observed throughout the period.

Introduction

The development of cluster policies relies on the growing awareness from academics and policy makers that network failures have to be merged with traditional market ones in the design of public innovation incentives (Woolthuis et al., 2005; Vicente, 2017). That is why cluster policies have been implemented in many countries since the end of the 1990s (Uyarra and Ramlogan, 2012; Maffioli et al., 2016). They coexist nowadays with innovation policies based on individual incentives, such as research tax credit and innovation grants sponsored by public agencies (Nishimura and Okamuro, 2011). Cluster policies aim at designing R&D collaborative incentives to strengthen knowledge networks in order to stimulate the expected benefits of local knowledge spillovers (Broekel et al., 2015). Cluster policies basics broadly rely on two related network failures. First, the potentialities of knowledge spillovers from science to industry can be inefficiently exploited due to the cultural divide and the weak absorptive capabilities between the two communities. Considering that positive impacts of knowledge spillovers are geographically bounded (Audretsch and Feldman, 1996), cluster policy guidelines will tend to favor local incentives towards networks mixing public research organizations and companies. Second, entrepreneurship matters in clusters (Rocha and Sternberg, 2005; Delgado et al., 2010). Their effectiveness can be assessed by the rate of SMEs and spinoffs’ birth and entry. The latter is the mark of the level of technological variety and renewal, and therefore represents a significant indicator of the cluster long-run dynamics. Here again, these births and entries are geographically bounded, and contained in the close perimeter of universities and big companies (Audretsch and Lehmann, 2005). But the entry dynamics is not a significant condition of cluster success per se. New entrants sometimes need to benefit from collaborative opportunities, especially in industries in which modularity and interoperability matter (Suire and Vicente, 2014). Then, collaborative incentives between SMEs and big companies are also a regular means used in cluster policies to foster regional performance.

The aim of this research is to have a deeper insight of these policy guidelines, and to find and test adapted network methodologies to deal with (i) the links between the public micro-incentives for knowledge collaboration and the structural properties of the network that emerge from these incentives, and (ii) the identification of the agents at the origin of structural changes. In that respect, the place-based network methodology and the nested cohesive block analysis are developed and offer promising avenues. As a matter of fact, it is common in the literature to assess whether network position increases individual innovative performance (Zaheer and Bell, 2005; Cattani and Ferriani, 2008). Nevertheless, very few contributions have studied the impact of cluster development programs not on the links between the actors’ position and performance but on the links between the structural properties of networks these incentives produce and the patterns of knowledge dynamics at work within the cluster (Crespo et al., 2014a; Giuliani and Pietrobelli, 2016). This question requires going beyond the different centrality degrees developed in the literature to measure individual position in networks. It requires investigating different concepts related to complex structural properties imported from network theories in order to better disentangle the consequences of different properties of network connectivity on the cluster development patterns. Moreover, it also requires overcoming methodological issues that arise when one deals with aggregate relational data at the regional scale. To improve our knowledge on that emerging research topic, we will focus on a single case study: the Aerospace Valley in Toulouse – France from 2006 to 2015, i.e. from the start of the policy to the year from which data are available. Aerospace Valley is one of the leading public-funded clusters granted by the French Cluster Program, and it is also the name of the association nurturing R&D collaborations and managing the international visibility of the cluster. As a consequence, Aerospace Valley can be considered as a particular cluster and knowledge network whose nodes are organizations involved in R&D projects selected by the association, and ties are collaborations having received public incentives. Our goal is not to find causality between the policy and the innovative performance of the organizations affiliated to the cluster program. To do so, systematic analysis on several places should be carried out, and counterfactual analysis required (Giuliani and Pietrobelli, 2016). Our goal is different and just as important for whoever wants to have a better understanding of how policy makers shape the organization of innovation processes in regions. Indeed, our basic starting assumption is related to the fact that, in spite of their control on the selection of R&D collaborations at the micro and dyadic levels, policy makers cannot have a perfect real-time knowledge and control of the structure as a whole. In network theories, this type of micro-macro scales problems is typical (Watts, 2004; Newman et al., 2006): the “macro-behavior” of the network and its structural properties, both resulting from the aggregation of ties, can escape their own intention. Since clusters are foremost networks (Giuliani and Bell, 2005; Vicente et al., 2011), dealing with the links between micro incentives and macro structures can be an alternate means to discuss how innovation policies can shape collaborative patterns and the structure of knowledge networks, as previously documented in the context of European Framework Programs assessment (Breschi and Cusmano, 2004; Vonortas, 2013).

The contribution is divided as follows: Section 2 goes back to the structuring of R&D networks in clusters and the design of public collaborative incentives aiming at repairing network failures. Section 3 aims at exemplifying these incentives and their consequences in the evolving structure of a particular cluster. We start by explaining the historical and technological context in which the Aerospace Valley cluster has been selected by national authorities to be eligible to public-funded incentives for R&D collaborations, before describing the cluster policy guideline developed in order to sustain its development. Section 4 presents the data collection procedure which enables us to build an original and complete dataset of public-funded collaborative R&D projects for this cluster. Then we discuss the methodological issues for building networks over the period. We disentangle the problems that generally arise for the study of networks resulting from the simple aggregation of collaborative and multilateral R&D consortia. To circumvent them, we suggest a place-based network methodology that focuses on structurally-equivalent relational behaviors. Section 5 shows how this methodology helps us identifying the evolving structural properties of knowledge networks in the cluster over time. Section 6 discusses the results under a particular focus related to the convergence degree between the network statistical findings and the objectives stated by the policy makers, with a particular focus on the agents of the structural and technological changes over the period.