A Functional Perspective on Learning and Innovation: Investigating the Organization of Absorptive Capacity

Abstract

We investigate the intra-organizational antecedents of firm-level absorptive capacity (AC). Specifically, we examine how the functional areas of R&D, manufacturing and marketing contribute to the absorption of knowledge coming from different external knowledge sources. The econometric results on a representative sample of Swiss firms show that non-R&D-based AC plays a significantly different role compared to the standard R&D-based one that is typically considered in studies on AC. We also reveal that AC is organized through a specialization of external knowledge absorption across functional areas. In particular, we find: (1) R&D is particularly important as an absorber of knowledge from public research organizations for product innovation; (2) manufacturing is important as an absorber of supplier knowledge for product innovation and of competitor knowledge for process innovation; and (3) marketing helps to absorb customer knowledge for product and process innovation as well as competitor knowledge for product innovation. We further investigate the differences between product and process innovation and find that marketing-based AC is more important for the former, although the overall analysis of these differences is less conclusive. In short, we show how functional areas play a role in the organization of AC and that firms may need an ambidextrous strategy to innovate effectively based on both upstream- and downstream-based AC.

Keywords:

• Absorptive capacity
• functional strategies
• process innovation
• product innovation
• open innovation

Acknowledgements

The authors would like to thank the three anonymous reviewers and the editors of this special issue for their comments and suggestions, which helped improve this paper. The authors are moreover particularly grateful to Allan Afuah, Paul Beije, Georges Bresson, Jan Fagerberg, Dominique Foray, Marc Gruber, Mathilde Maurel, Mette Præst Knudsen, Keld Laursen, Christopher Tucci and Anu Wadhwa for their constructive comments on earlier drafts of this paper. The authors also thank the KOF at ETH Zurich for data access. The support of the College of Management of Technology (CDM) at EPFL as well as the financial assistance of the Swiss National Science Foundation (grant no. 105512-106932) are gratefully acknowledged.

Notes

¹ Many conceptual differences have been introduced in order to describe the stages in the absorption process (see Lane et al., 2001; Zahra and George, 2002). We consider here that absorption is broader than learning: the identification and access to external knowledge by a firm are upstream phases that do not systematically imply subsequent learning stages covering the assimilation and exploitation of external knowledge. For example, accessing alliances (Grant and Baden-Fuller, 2004) or scale alliances (Hennart, 1988) are inter-organizational practices requiring internal AC.

² The findings in the literature are however mixed (e.g. Frishammar and Hörte, 2001; Un et al., 2010).

³ Gatekeepers are boundary-spanning individuals who manage knowledge flows with the external environment.

⁴ Thanks to the layout of the questionnaire which included items both on product and process innovation, firms only doing product innovation during the investigated period also answered the questions on the roles of incoming knowledge and internal functions in process innovation even though they did not introduce or start such innovation during that period. This feature is interesting since it minimizes the potential biases due to the censoring of explained or explanatory variables.

⁵ Product innovations are defined as being new to the market (meeting new demands) and/or being technologically substantially different from earlier products. Process innovations are defined as new or significantly improved production techniques and processes.

⁶ Dichotomization is required in order to limit the number of explanatory variables and to be able to compare the role of different knowledge sources. Doing so, we reduce the number of variables to be included in our model: therefore, instead of 4 dummies we have only one variable per external source measuring every level of contribution. From a general point of view we therefore have 6 explanatory variables instead of a set of 24 (6 × 4) dichotomic variables for knowledge sources. A second solution would be to consider the declared scale as a continuous variable (as in Cohen and Levinthal, 1989, for example) but this would induce difficulties in the interpretation of marginal effects.

⁷ The list of appropriation means includes patents, models or other intellectual property rights, secrecy, product complexity, leadership, long-term relationship with skilled employees, and complementary services.

⁸ Note that with three separately ordered probit equations, the estimator may lead to inefficient results.

⁹ “Equation” refers to the “Model Equation” in Table 2 and Table 3.

¹⁰ The results found with the multivariate-ordered probit model are similar to a seemingly unrelated regression (SUR) model which considers our explained variables as linear (results available upon request).