Is there a fallacy of composition of external R&D? An empirical assessment of the impact of quasi-internal, external and offshored R&D

Abstract

Empirical studies at the individual firm level often find a positive impact of external R&D on innovation. However, external R&D sourcing might produce an impoverishment of the local knowledge base and thus damage innovation production at the regional level. To address this possible fallacy of composition, we first review the various forms of ‘external’ R&D. We then use the French R&D survey to assess the impact of four different ways of transacting or collaborating on R&D: onshore affiliate external R&D, offshore affiliate external R&D, onshore non-affiliate and offshore non-affiliate external R&D. We then estimate knowledge production functions on the 94 metropolitan French NUTS3 regions observed between 1997 and 2008, differentiating internal R&D and these categories of external R&D. We obtain that the impact of onshore non-affiliate external R&D is significantly negative. The other external R&D categories are non-significant and we do not detect any complementarity between internal and external R&D.

Keywords:

• Innovation
• knowledge production
• external R&D
• offshoring
• outsourcing
• subcontracting
• R&D partnerships

JEL codes:

• R12
• R15

Acknowledgement

We acknowledge financial support of the ANR programme ‘T-RES’: ‘Territories and technologies in an unstable knowledge economy: An evolutionary framework of regional resilience’.

Notes

¹ The fDi Markets database that is used in some recent studies at the aggregate level reports foreign direct R&D investments but does not provide the amount invested (Castellani and Pieri 2013; D’Agostino, Laursen, and Santangelo 2013). If we had used this data source, we could only have computed a count of the number of offshore R&D investments made by each region. With the French R&D survey, we have access to total R&D expense (including investment expenses, but also wages, current expenses and so on) and to all types of external R&D expenses (including, e.g., onshore R&D expenses and quasi-internal R&D expenses).

² Hagedoorn (2002, 478) uses a similar argument regarding equity joint ventures.

³ Narula and Martinez-Noya (2014) also propose a detailed typology of strategic technology partnerships.

⁴ They use a survey implemented in 1997 by the SESSI (a research department of the French Ministry of Industry) providing information on the set of competences that French firms possess. They classify these competences to differentiate internal and external ones in relation to the innovation process. Their results show the importance of external competences in regional innovation processes, which we consider as valuable empirical evidence of the importance of absorptive capacity. Nevertheless, the complementarity between these factors contributing to build a good absorptive capacity and external R&D is not tested.

⁵ Because the INPI only provided us with the patent counts and not the detailed data, we could not implement the solutions that are sometimes used in the literature. Nevertheless, we can argue that these solutions are not necessarily satisfying: weighting the patents by their number of citations has disadvantages since many citations are imposed by patent examiners according to criteria that do not really reflect economic value. Moreover, very new ideas may not be cited for quite a while. In addition, the other solution that would consist in running the regressions with only the patents of one single office would imply a large measurement error in the dependent variable, a solution we consider worse than the disease.

⁶ This is also due to the NUTS3 aggregation level.

⁷ More precisely, all the 243 French firms classified as ‘large’ are surveyed each year, representing 87% of the French internal R&D expenses, and the rest of the firms are partially surveyed each year. The parent population is made of all the firms that implement R&D, which represents approximately 23,000 companies on a total of 3,14 million businesses in France (of which 3 million are classified as ‘micro-firms’, 138,000 are ‘SMEs’, 5000 are ‘intermediary size firms’ and 243 are ‘large enterprises’). The drawing is exhaustive for two categories of firms (6000 entities): those that make more than 750 K€ of internal R&D expense and those that recently entered the parent population because they started doing R&D. 5000 entities are drawn among the 17,000 remaining ones.

⁸ We thank the referee who suggested this method.