Seite - (000021) - in Knowledge and Networks

9 regions and that those regions with the strongest spillovers in technological devel- opment were the ones where job mobility was restricted mostly to intraregional job moves. Using network analysis, even more sophisticated research designs have illustrated that patent citations tend to be local because inventors tend to change jobs locally and stay within their labor-market region (Breschi & Lissoni, 2009). This evidence suggests that the most fundamental reason why geography matters for localized knowledge creation is the relative immobility of researchers. The job mobility of inventors increases the transfer of technological knowledge (measured as patent citations) between firms independently of geography (Rosenkopf & Almeida, 2003). In support of this conclusion, Breschi, Lenzi, Lissoni, and Vezzulli (2010) offered evidence that proximate and remote job moves occur in equal pro- portions, and Song et al. (2003) demonstrated that both proximate and distant hiring of inventors lead to effective transfer in technological knowledge as measured by patent citations. A second example of how networks mediate the relation between space and knowledge is provided in the context of information search. Borgatti and Cross (2003) found that when one knows an informant and can access that source, physical proximity is no longer associated with information transfer.5 In summary, this line of research suggests that the association between geography and knowl- edge is not a direct effect, that it is mediated through inventor mobility, the acces- sibility of other partners, and prior knowing. Agency as a Moderator of Relational “Places” in Multilevel Relationships on Collective Learning Social scientists, especially sociologists and geographers, have arguably been build- ing a strong alliance in the social sciences to measure, model, and understand the multilevel dynamics of places, positions, and the effects of such dynamics on all the generic social processes we are interested in, notably knowledge-building and col- lective learning. In particular, when complex position in an organized system of places allows actors to try and change that formal structure (albeit with varying success), scholars interested in spatial and organizational movement can help com- bine institutional locations, position in relational infrastructures (e.g., status and niches), and geographical place (e.g., Bathelt & Glückler, 2011; Glückler & Hammer, 2012). Social processes such as collective learning and knowledge creation are also contingent on multilevel interdependencies and require unprecedented amounts of coordination among actors at and across given levels. Actors think in multilevel terms (“this person is a big fish in a big pond”) and are required to manage these 5 Mediation implies that the mediated variable (proximity) predicts the mediating variables as well as the dependent variable (e.g., innovation, information exchange) and that the coefficient for the mediated variable becomes insignificant when the mediators are included in the model (Baron & Kenny, 1986). 1 Introduction