Knowledge networks are important to understand learning in industry clusters but surprisingly little is known about what drives the formation, persistence and dissolution of ties. Applying stochastic actor-oriented models on longitudinal relational data from a mature cluster in a medium-tech industry, we show that triadic closure and geographical proximity increase the probability of tie creation but does not influence tie persistence. Cognitive proximity is positively correlated to tie persistence but firms create ties to cognitively proximate firms only if they are loosely connected through common third partners. We propose a micro perspective to understand how endogenous network effects, cognitive proximity of actors and their interplay influence the evolutionary process of network formation in clusters.

No abstract