“…Accordingly, "various kinds of data have been used to indicate knowledge networks" (Balland, 2014), including knowledge sharing relations (Giuliani and Bell, 2005;Giuliani, 2007;Morrison, 2008;Broekel and Boschma 2012), patent citations (Agrawal et al, 2006;Breschi and Lissoni, 2009), joint patents (Cantner and Graf, 2006;Hoekman et al 2009), joint publications (Ponds et al 2007(Ponds et al , 2010Frenken et al 2009;Scherngell and Hu, 2011;Hardeman et al 2012) and joint participation in R&D projects (Hagedoorn, 2002, Autant-Bernard et al 2007Maggioni et al 2007;Scherngell and Barber 2009;Balland, 2012). Nevertheless, authors sometimes mention limitations associated with their data (Ter Wal and Boschma, 2011), pointing out that although interview data provide the most information, because interviews themselves are so time-consuming, it is impossible to obtain enough data extending over time and space (for more details on the limitations of data used to analyze a network, see Bernela and Levy (2015) discussing the virtually systematic hypothesis of complete graphs). The proliferation of empirical studies applying the same model and the same methodology to different databases, yet which were accepted by peer-review, was a factor of enrichment but also one of saturation.…”