Page - (000361) - in Knowledge and Networks

359 patents (not copatents) that both partners owned in the 5 years prior to their collabo- ration. We regarded that sum as a proxy for their accumulated innovative capabili- ties (DyadSinglePAT5). To delimit the domain of the variable, we took the logarithm of these values. We limited the observation period to the 5 years preceding the col- laboration of the two firms, assuming the knowledge to be almost obsolete thereaf- ter and accounting for the depreciation of innovative capabilities. Studies on the depreciations of R&D activities (Czarnitzki, Hall, & Oriani, 2006; Edworthy & Wallis, 2009; Hall, 2007) have indicated that R&D investment is completely depre- ciated after 3–5 years. General Collaboration Experience Analogously, to capture the attractiveness of the collaboration opportunity in terms of management ease, we took the sum of the shared patents (copatents) that both actors held in the 5 years prior to the collabora- tion as a proxy for their accumulated collaboration experience (DyadCoopPAT5). Because we wanted to detect the general collaboration experience, we used this measure to add up all collaborations except the one in question. The greater the col- laborative experience is, the higher the likelihood of further collaborations. We also assumed average capability depreciation after 5 years and applied the logarithmic transformation to delimit the range of the variable. Popularity Taking reciprocal incentives into account, Giuliani (2007) has argued that central actors who are popular (as measured by their number of other linkages) tend to connect to similarly embedded actors. We believe that the potential for knowledge spillovers might be greater when partners are equally popular and pos- sess a similar pool of potential knowledge sources (links). To test this relation (hypothesis 3c), we followed Dahlander and McFarland (2013) in using the abso- lute difference between the two partners’ degree of centrality (the number of links) in the year before actual or potential collaboration. We called this variable DCentrality. Theoretically, this measure is closely related to the general collabora- tion experience. In our analysis, however, it captures the reciprocity of popularity in collaboration activity rather than the pure amount of previous collaboration activity. Control Variables Apart from technological, social, and competence aspects, we also wanted to con- trol for additional effects stemming from organizational and age similarity. Both variables might increase the likelihood of collaboration due to ease of communica- tion when the cooperating partners are exposed to the same institutional factors and environments (organizational similarity) or when they have had the same amount of time to operate in these environments and to accumulate experience and resources (age similarity). Organizational dissimilarity—DStatus—is a binary variable taking the value 1 when the two actors differ in organizational nature and zero when they are of the same organizational type (interfirm collaboration). DPatAge is the 16 Coevolution of Innovative Ties, Proximity, and Competencies