Open, Semi-Open and Closed Innovators: Towards an Explanation of Degree of Openness

Abstract

There is much controversy in the literature over the relationship between the openness of firms' innovation strategies and firm characteristics such as size, R&D intensity and sector. We argue that the controversy arises because, both theoretically and empirically, only a binary, open vs. closed, strategy has been considered. In this paper, we distinguish among three firm strategies: open, semi-open and closed, drawing upon a panel of Spanish firms (2004–2006) using data from Community Innovation Survey (CIS)-type surveys, and two different indicators of openness. Our results show that open innovators are smaller and less R&D intensive than semi-open ones, although larger and more R&D intensive than closed innovators. These results reduce some of the controversies, and show that two conflicting forces, absorptive capacity and a “need” effect, are at stake in open innovation strategies.

Keywords:

• Open innovation strategies
• collaboration
• size
• R&D
• sector
• technology policy

Acknowledgements

The author wants to acknowledge financial support from project 08SEC008201PR (Xunta de Galicia) and from UAM-Accenture Economics and Innovation Management Award. A previous version of the paper was presented in Zvi Griliches 2009 Research Summer Seminar on the Economics of Innovation and the DRUID 2010 Summer Conference. Comments received from participants are acknowledged. The usual disclaimers apply.

Notes

¹ Note that the questions related to OI refer to all its modes. Here we focus only on inbound OI; thus, in this paper, OI strategies refer only to this mode.

² The method and type of questions are described in the Oslo Manual (OECD, 1997, 2005).

³ We refer to non-linearities related to degree not likelihood of openness.

⁴ Size could be considered a proxy for some dimensions of absorptive capacity as it is usually an indicator of better developed R&D management practices (Huergo, 2006), which facilitate the recognition, assimiliation and application of external knowledge for commercial ends.

⁵ This framework does not include sector for two reasons. First, the relationship between “need effect” and sector is unclear, second, and more important, we explore the mediating role of sector on the framework proposed (Section 5.3).

⁶ Although not all are official surveys, they employ similar questions to those in the CIS.

⁷ We focus here on studies that analyse cooperation in general, and not with a specific type of partner.

⁸ In the French case, and only after instrumenting the variable (Abramovsky et al., 2009).

⁹ The reason is probably that Barge-Gil (2010) measures the importance of cooperation in relative terms (to the internal processes) while de Faria et al. (2010) measure it in absolute terms.

¹⁰ More information on the database and its anonymization can be found at http://sise.fecyt.es/Estudios/PITEC.asp (in English).

¹¹ The question was asked separately for product and process innovations. Here, we consider that a firm is an open innovator if product or process innovation were developed mainly through collaboration with other entities or mainly by other entities. (See Tables A2(a) and (b) in the Appendix for separate estimations for product and process innovators.)

¹² Using OPENNESS_IMP improves the dummy variable of cooperation used in existing studies as it distinguishes between those firms that innovate mainly through cooperation from those that, although cooperating, innovate mainly through their own efforts (Barge-Gil, 2010) and allows to measure not only the existence of ties but also their strength, which is an issue deserving more attention (Tomlinson, 2010).

¹³ To confirm the adequacy of this method, we tested our results by dropping those firms assigning no or low importance to every source. The results are almost identical (see Tables A3(a) and (b) in the Appendix).

¹⁴ We want to thank an anonymous referee for raising this important point. For analyses focused on breadth, see, for example, Laursen and Salter (2006), Nieto and Santamaría (2007) or Vega-Jurado et al. (2009).

¹⁵ Independent variables for the period covered by dependent variables. We tried them also with 1 lag and the results (unreported, available upon request) were qualitatively unchanged.

¹⁶ We also tried binary logit regressions eliminating every alternative. Results (not reported here, but available upon request) were qualitatively similar, indicating that the assumption of independence of irrelevant alternatives (IIA) holds. We decided to employ this method since recent studies tend to conclude that IIA tests are not suited to applied work as they often reject the assumption when alternatives seem distinct, and often fail to reject it when alternatives can reasonably be viewed as close substitutes (Chen and Long, 2007).

¹⁷ Only coefficients for the comparison between open and closed innovators are shown, as the comparison between closed and semi-open is already provided in Table 6(a).

¹⁸ The same robustness checks were performed for the sectoral regressions, being results qualitatively unchanged. They are not provided here for reasons of space but are available from the authors upon request.

¹⁹ In this case, no systematic differences were found between open and closed innovators.

²⁰ Note that this means that firms in HT sectors would be outbound open innovators.

²¹ Other studies that are not precisely focused on this issue, derive similar conclusions. For example, de Jong and Marsili (2006) find that innovation subsidies are much more frequent in clusters of “science-based” firms.