Abstract: Assessing whether network position causes organizational outcomes is difficult because networks are usually the result of firm choices (i.e., firm agency). Rather than adopting a statistical approach to address this challenge, we offer a theory-driven solution. Using structural causal modeling, we integrate canonical network theory with the concept of alliancenetwork externalities. This distinguishes self-driven changes in a focal firm’s network position from other-driven changes—the latter of which suppress the agency of the focal firm. Therefore, under certain assumptions that we can evaluate, assessing how other-driven changes affect organizational outcomes can be interpreted as a causal test of network position. Examining the biotechnology industry alliance network (1995-2012), we find that structural holes increase firm innovation only under conditions of self-driven network change, but not under conditions of other-driven network change. We thus do not find support for a causal effect of structural holes per se on innovation. One interpretation is that the effect of network position is spurious. Another is that canonical theory requires updating to account for agency as a factor that activates the benefits of structural position. Our theory and results have profound implications for how scholars theorize and test network effects.