Seite - (000359) - in Knowledge and Networks

357 assigned to single patents, then the knowledge had been successfully integrated and was applicable afterward without further collaboration. Used in conjunction with this procedure, the binary variable TransKnowledge indicates whether knowledge has been exchanged in prior collaborations. This variable takes the value 1 if either partner has gained new knowledge; otherwise it takes the value zero. That is, the variable captures both symmetric and asymmetric learning. Our three measures of cognitive proximity—RelOverlap, ReciPot, and TransKnowledge—do not develop independently of each other. Their changes over time go hand in hand. Figure 16.1 illustrates the dynamics of these three variables. Two actors, I and II, hold specific knowledge portfolios before cooperating with each other (precollaboration). Actor I’s portfolio comprises ABCDEF; actor II’s, ABGH. The knowledge overlap in t-1 is given by AB and amounts to .2, relative to the overall knowledge. The reciprocal potential equals .5 because actor II possesses two knowledge units that actor I can gain as opposed to four knowledge units that actor II might be able to acquire from actor I. In other words, actor I can gain at most only half the amount of knowledge that actor II, the partner, stands to gain. Formulated differently, actor II can earn twice the amount of new knowledge that is being offered to actor I. In this example, the potential gains are unequal. Assume that collaboration then leads to symmetric learning in that C and G are exchanged. Actor I’s postcollaboration portfolio is thereby enlarged to ABCDEFG; actor II’s, to ABCGH. As a result, the overlap has increased to ABCEG and amounts now to .3 in relation to the overall knowledge possessed by the two firms. In turn, the ratio between the potential knowledge gains has decreased to .3 because actor II now offers only one new knowledge unit to actor I, whereas actor I now offers three knowledge units to actor II. The potential for knowledge flows has thus decreased and become more uneven. The attractiveness of this fictive alliance and the likeli- hood that it will continue have therefore declined. This example illustrates the case of knowledge having been efficiently exchanged. When actors collaborate but are unable to integrate new knowledge into their stock, then knowledge has only been shared and the collaboration is more likely to continue than if they are able to inte- grate the new knowledge. In this sense, a continuation of collaboration can be inter- preted as a failure to learn (Hamel, 1991). Social Proximity Between the Cooperation Partners To test whether the probability for the creation or re-creation of a link increases with the social proximity between the partners (hypothesis 2), we included a variable for common experience, CoopExp, as a proxy for social proximity. CoopExp measures how often the pair was cooperating prior to the cooperation in question. The number of prior research projects with the partner is commonly used as a measure of the strength of the tie and is assumed to capture the trust and ease of communication between the partners (Cantner & Meder, 2007). 16 Coevolution of Innovative Ties, Proximity, and Competencies