The knowledge-base evolution in biotechnology: a social network analysis

Abstract

This paper applies the methodological tools typical of social network analysis (SNA) within an evolutionary framework, to investigate the knowledge-base (KB) dynamics of the biotechnology sector. Knowledge is here considered a collective good represented as a co-relational and a retrieval-interpretative structure. The internal structure of knowledge is described as a network, the nodes of which are small units within traces of knowledge, such as patent documents, connected by links determined by their joint utilization. We used measures referring to the network (like density) and to its nodes (like degree, closeness and betweenness centrality) to provide a synthetic description of the structure of the KB and of its evolution over time. Eventually, we compared such measures with more established properties of the KB calculated on the basis of co-occurrences of technological classes within patent documents. Empirical results show the existence of interesting and meaningful relationships across the different measures, providing support for the use of SNA to study the evolution of the KBs of industrial sectors and their lifecycles.

Keywords:

• knowledge base
• social network analysis
• variety
• coherence
• industry lifecycles
• exploration/exploitation

JEL Classification :

• O33

Acknowledgements

This work is part of a research project funded by the Agence Nationale de la Recherche (contract number: ANR JCJC06_141306, ‘Knowledge Intensive Sectors: Models and Evidence’), the Provence Alpes Côte d'Azur (PACA) Region, and the University of Nice (BQR). The authors acknowledge the funding of Collegio Carlo Alberto through the BRICK research centre, as well as the support of CNRS through the GREDEG research centre.

Notes

In spite of this clearly established trend knowledge produced according to more traditional methods is still used alongside the one produced in specialized institutions. Thus, the operational definition of knowledge that we propose in this paper cannot establish a demarcation between scientific and non-scientific knowledge but needs to be generally applicable to all the types of knowledge which can be combined in human activities.

We consider thus patent applications as the best indicator of firms KBs, though the usual caveats mentioned in the literature may apply. We use these data to map the frequency of co-occurrences of technological classes within patents and to calculate a number of indexes, i.e. information entropy used to measure related and unrelated variety, knowledge coherence and cognitive distance.

Though the use of IPC classes to define sectors’ boundaries may present some drawbacks, as they are function-oriented (Corrocher, Malerba, and Montobbio 2007), the merging of two classifications allows our study to be much more inclusive than many other studies, and reduce the risk of neglecting important classes. It is worth noting that these classes include quite different technologies and processes, which might be placed at different stages of an ideal filière of the knowledge production process. This is a potential source of misunderstandings or misinterpretation of our results, due to the fact that one could claim that classes in certain stages of such filière are more likely to be central than classes impinging upon other stages. However, given the interactive nature of the knowledge creation process, this may help more the discussion of empirical results than the ex-ante formulation of expectations.

It is fair to note that a similar approach has been attempted at the firm level by Yayavaram and Ahuja (2008).

This section builds upon Scott (2000) and Wasserman and Faust (2007).