Seite - (000366) - in Knowledge and Networks

364 that the squared term of the relative overlap (RelOverlap2) between the knowledge bases of the two partners had a highly significant positive correlation to the probability of collaboration. However, we found no evidence of a moderate overlap and, hence, no support for hypothesis 1a. When controlling for combined effects of experience and overlap (see the column labeled “Interactions”), we found only a pure positive correlation between overlap and the likelihood of collaboration. Thus, the degree of mutual understanding seems to increase the likelihood that linkages will be recreated. Second, our impression of the search for diversity as illustrated in Fig. 16.2 was confirmed by the results of our estimation. We found that firms were more likely to reconnect with actors who differed from them in the amount of potentially new knowledge than with actors who were the same or similar in that respect. The nega- tive relation between reciprocal potential (ReciPot) and the likelihood of collabora- tion indicates that reciprocity in knowledge gains is not a necessary precondition for the continuity of collaborations. Our result was opposite to the assumed relation stated in hypothesis 1b. Third, concerning hypothesis 1c, we did not find a significantly positive correla- tion between collaboration and previous knowledge transfer (TransKnowledge). Our results seem to contradict our hypotheses on the relevance of knowledge diver- sity in the evolution of cooperation. Concerning cognitive proximity, the need for mutual understanding seems to predominate over need for reciprocity in potential knowledge gains. Regarding social proximity, we found no empirical connection between the chances for cooperation and prior common experience (CoopExp), a result that does not support our suggestion in hypothesis 2 that the propensity of collaboration increases with prior common experience. Even though common experience did not play a significant role in partner choice among the firms in our sample, the combined overall cooperation experience (DyadCoopPAT5) was positively and significantly correlated with the re-creation of linkages. That is, choices to collaborate were preferred when at least one actor exhibited great capability in managing cooperation. This finding is consistent with the results reported by Gulati (1999), who observed the same supportive effect that an actor’s general experience with collaboration has on that actor’s chances of forming linkages. The importance of cumulative advantages is also reflected in the negative relation between collaboration propensity and the difference in the degree of popularity (DCentrality). Firms tended to seek reciprocal incentives when it came to accumulating experience and building their cooperation capability but not when they were interested in gaining knowledge benefits. Our results indicate that firms prefer to link up with actors who offer an equal amount of accumulated resources. Dahlander and McFarland (2013) found the same negative correlation between the difference between the “cumulative advantage” (p. 72) of both partners and the persistence of collaboration between researchers at Stanford University. Conversely, the common cumulative innovative potential as measured by the total number of single patents held by both actors (DyadSinglePAT5) seems rather irrel- evant when it comes to partner choice. Therefore, we find support for our hypothe- ses 3b and 3c but not for hypothesis 3a. U. Cantner et al.