Page - (000351) - in Knowledge and Networks

349 cooperative research process (Singh, 2005). By definition, cooperative patents com- prise inventive success in this context. Although patent data come with certain limi- tations (see Griliches, 1990; Ter Wal & Boschma, 2009), they offer a rich and comprehensive database on inventive activities. While working with patents, one must carefully define the scope of analysis in order to avoid the bias stemming from unobserved heterogeneity in patenting behavior (across industries and nations, for example). To reduce this bias arising from intercountry and interindustry differ- ences, we narrowed our analysis to patents that were filed by German applicants in the field of biotechnology between 1978 and 2010. The biotech industry is charac- terized by a high propensity to patent and a high frequency of joint research (Griliches, 1990; Powell & Grodal, 2006; Ter Wal, 2014). We gathered the data from the OECD REGPAT database5 (January 2012 ed.), which covers patent appli- cations to the European Patent Office (EPO) and the United States Patent and Trademark Office (USPTO). To match the collaborative actors to their respective other patents, we used the OECD Harmonised Applicants’ Names (HAN) database, “which provides a dictionary of applicants’ names which have been elaborated with business register data, so that it can easily be matched by all users” (retrieved July 15, 2015, from http://www.oecd.org/sti/inno/oecdpatentdatabases.htm). The use of patent data in our analysis requires some qualifications. First, our pool of potential collaborators encompassed all applicants with at least one patent appli- cation between 1978 and 2010. The influx of entries meant that this pool was not fixed over time; it grew from year to year, so we had to deal with an unbalanced panel. Second, a link between actors occurred when actors appeared together as applicants on one patent document (coapplication). The probability of false posi- tives in detecting collaborations was assumed to be very small because a coapplica- tion reduces the applicants’ claim to the patent. Third, it was debatable whether continuous cooperation was evident in patent data. If two applicants were persis- tently copatenting, we assumed that they were still conducting joint research. In this sense, we were able to identify long-lasting relationships but may have underesti- mated the number of ongoing partnerships that did not result in patents. Fourth, patents have been established as a measure of technological capabilities (Mowery et al., 1996). The suitability of patent data as a proxy for firms’ knowledge stock derives from the disaggregate information they convey. The International Patent Classification (IPC) offers a standardized and detailed technological classification system that enables one to assign the protected invention to a certain field of tech- nology and to characterize the firms’ research activities by constructing firm- specific technology portfolios (Griliches, 1990; Jaffe, 1986; Benner & Waldfogel, 2008). 5 “The OECD REGPAT database presents patent data that have been linked to regions according to the addresses of the applicants and inventors. The data have been ‘regionalised’ at a very detailed level so that more than 2 000 regions are covered across OECD countries. REGPAT allows patent data to be used in connection with other regional data such as GDP or labour force statistics, and other patent-based information such as citations, technical fields and patent holders’ characteristics (industry, university, etc.), thus providing researchers with the means to develop a rich set of new indicators and undertake a broad range of analyses to address issues relating to the regional dimen- sion of innovation.” (Maraut, Dernis, Webb, Spiezia, & Guellec, 2008, p. 3). 16 Coevolution of Innovative Ties, Proximity, and Competencies