This dissertation explores sources of variation that increase the innovation output of firms. There are three main premises of this dissertation. First, I argue that co-authorship relationships among scientists constitute an interorganizational network and the position of a firm in this network affects its innovation output. Second, a firm has to access diverse knowledge sources and also has to effectively transfer and utilize scientific knowledge from these sources in order to create innovations. These needs exert conflicting pressures on the optimum social structure of the firm. The need for diversity increases the benefits of structural holes, whereas the need for the transfer of tacit knowledge increases the benefits of closure and tie strength. Firms need to find a balance between structural holes, closure and tie strength in their scientific network to enhance their innovation output. Third, firms in the biotechnology industry are embedded in multiple networks—specifically the scientific and alliance networks—and examining the relationship between the structural positions of the firm in these networks increases our understanding of the determinants of innovation and the relationship between these networks. The networks can be independent, complementary, or substitutes. I develop hypotheses regarding the effects of the position of a firm in the scientific network and its relational characteristics in this network on its innovation output. I also develop hypotheses regarding the effects of the position of a firm in the scientific network on the relationship between its position in the alliance network and its innovation output.

I test these hypotheses in the US biotechnology industry between 1990 and 2002 using a fixed-effects Poisson regression. The results suggest that the position of the firm in the scientific network significantly affects its innovation output. I find a positive relationship between the number of alters of a firm and its innovation output as well as an inverted-u shaped relationship between both its structural holes and its average tie strength and its innovation output. The results also suggest that the scientific and alliance networks have a complex relationship.