Do Firms Know the Scope of their R&D Network? An Empirical Investigation of the Determinants of Network Awareness on French Survey Data

Abstract

Although research and development (R&D) networks influence the innovation performance of their members, firms may not be fully aware of the scope of their network. In particular, due to cost reasons, they may not be fully informed of their “indirect ties”, that is, of the ties between their partners and other firms. To investigate this issue, the paper uses a survey inquiring about whether firms are aware of the ties that their main direct R&D partners may (or may not) have between themselves. Our results show that responding firms are more informed about their partners' other collaboration projects when the partnership is more directly linked to intangible R&D capital, when at least one partner is a public research organization or when the partnership is needed to access a new market. Network awareness is also higher when both R&D partners are from the same type (e.g. public research organization, companies, technical centers). Firms with a high R&D intensity or with a large size, as well as those affiliated to a group, are less likely to know their indirect ties. Finally, network awareness is lower in high-technology industries.

Keywords:

• R&D cooperation
• incomplete information
• social network

Acknowledgements

The authors would like to thank the Ministry of Industry (SESSI) and the Ministry of Higher Education and Research (DGESIP/DGRI/SIES) for data availability. The authors gratefully acknowledge Anu Wadhwa, Emmanuelle Fauchart, Marc Gruber, Christopher Tucci and the attendees of the EPFL MTEI “ERIG” seminar (April 2009), the participants at the REPERES seminar (6 July 2009, Paris, Ministry of Higher Education and Research), the attendees of the DIME workshop on “The Structure and Dynamics of Knowledge Networks” (12–14 May 2009, Eindhoven University of Technology) and two anonymous referees for their comments. Any remaining errors are the authors’ and they remain solely responsible for the views expressed in this paper, which are not necessarily endorsed by their respective institutions

Notes

¹ Oppositely, in Granovetter's terminology, “strong” ties are used to exploit and deepen existing knowledge (Uzzi, 1997; Rowley et al., 2000; Hagedoorn and Duysters, 2002).

² For n partners, the number of potential ties between these partners is n(n − 1)/2.

³ The matching was made possible thanks to the use of a unique identification number of firms (the SIREN code) in both data sets.

⁴ ERIE stands for “Enquête sur les Relations Inter-Entreprises”.

⁵ Thus, firms with less than 20 employees can be included in the sample provided that their turnover exceeds the threshold of 5 million euros.

⁶ This R&D survey is described in greater detail below.

⁷ The random selection process was calibrated to reflect the structure of the French industries, as taken from annual statistical surveys (EAE: enquêtes annuelles d'entreprises).

⁸ Thirty-seven firms declared at least three direct R&D links but detailed only two direct R&D ties out of three.

⁹ The censored variable is set to 0 when the number of declared direct R&D partners is exactly two or three since we can observe all the potential indirect ties. An alternative count would be the total number of ties. However, using the total number of ties would imply that no difference is made between firms who describe all their partnerships in the survey and the other firms in our sample. Furthermore, it would introduce a ranking between the three direct R&D partnerships, which is not necessarily present in the answers of the responding firms and which is already controlled for by three dummy variables relating to each of the described partnerships (see the subsection on the econometric method).

¹⁰ Firms who declared only intra-group R&D partnerships were deleted from our sample. However, some firms described both an external R&D partnership and an intra-group R&D partnership. These firms were kept in the sample.

¹¹ Hence, the dummy variables describing the location of the partner are not significant, reflecting the ambiguous influence that this characteristic of partners could have on network awareness.