Page - (000352) - in Knowledge and Networks

350 Jaffe (1986) was one of the first researchers to use patent data as a proxy for tech- nological competencies of firms. He constructed the knowledge portfolios as a vec- tor of patent classes in which firms patented, and he computed the distances between technology vectors of firms to obtain a measure of proximity among them. Researchers subsequently adopted Jaffe’s approach in using patent classes to show a firm’s technology portfolio, technological distances among firms, or potential pools of knowledge spillover in the firm’s environment (Benner & Waldfogel, 2008; Boschma & Frenken, 2010; Cantner & Graf, 2006; Cantner & Meder, 2007). We, too, made use of this rich information by constructing the knowledge portfolios of the actors and tracing their changes over time. Because it is unfeasible to approxi- mate knowledge portfolios of the individual inventor by means of patent informa- tion, we focused our analysis on the organizational level. Sample The basic characteristics of the sample are presented in Table 16.1. The sample consisted of 197 firms that applied for patents with partners between 1983 and 2010, the period for which we sought to explain links between partners. Because our objective was to explain a link between actors by examining their prior patent- ing activities, we consulted patent information on the 5 years before the actors’ first link as of 1983. Our calculation of the variables is therefore based on all patents the two actors applied for between 1978 and 2010. To analyze the dynamics of coopera- tion choice, we considered only the 91 firms that had cooperated at least twice between 1983 and 2010, and we observed their collaborative behavior over the years that followed the firms’ first appearance in the dataset. When a firm was coop- erating in 1 year, we paired it with each of the potential cooperation partners that were active in the pool at the same time. The pool of a firm’s potential cooperation partners consisted of all patenting actors who were active in the focal year or had entered the sample before that point (Cantner & Meder, 2007). For all possible combinations, we assigned a 1 for each realized cooperation and a zero otherwise. Double pairs were excluded. The size of the pool of potential partners was nonde- creasing from year to year. It amounted to a maximum of 2369 potential partners. By definition, the collaborations we looked at included the subject firm and, from the pool, one potential partner that could be of any type (e.g., firm, university), implying that the observations were not symmetric. All told, the 27-year span cov- ered by our analysis encompassed 321,683 possibilities to form dyads, of which 293 were ultimately realized. When we grouped actors according to their overall collaboration activity over the whole period or over their all-time partner portfolio (Wuyts et al., 2005), we identified 106 firms that had collaborated only once (one-shot), 27 that had collabo- rated at least twice but with different partners (hop-on, hop-off), 24 that had col- laborated persistently with the same partner (persistent), and 40 that had pursued a mixed strategy (mixed-type). For the purpose of our analysis, we focused on the U. Cantner et al.