Which Peers Matter? The Relative Impacts of Collaborators, Colleagues, and Competitors

Many economists believe that knowledge production generates positive spillovers among knowledge producers. We argue that spillovers may exist on at least three dimensions (in idea, geographic, and collaboration space), and that the strength or weakness of the spillover effects can vary across these dimensions. Using a unique data set that contains information on all publications of active mathematicians in the former Soviet Union, we examine the impact of a large post-­1992 exodus of Soviet mathematicians on the output of the non-­emigres. We find that a supply shock in the space of ideas (i.e., where the exodus consists of peers who work in similar topics) increases the output of the non-­emigres. However, neither a supply shock in geographic space (i.e., where the émigrés are members of the same university department) nor in collaboration space (i.e., where the emigres consist of coauthors) has a statistically significant effect on the output of a typical mathematician. We find evidence of human capital externalities only if the supply shock in collaboration space involves the loss of high-­quality coauthors. Spillovers, therefore, are most likely to be empirically important and dominate the competitive forces unleashed by a supply shock when high-­quality researchers are directly engaged in the joint production of new knowledge.