Heterogeneous Scaling of Research Institutions

Research institutions provide the infrastructure for scientific discovery, yet their role in the production of knowledge is not well understood. To address this gap, we analyze activity of researchers and their institutions from millions of scientific papers. Our analysis reveals statistical regularities in the growth of institutions, and how collaborations, research output and its impact scale with institution size. We find that scaling is heterogeneous and time-independent. Significantly, this result is missed by cross-sectional analysis, which measures complex systems at a point in time. To help explain the findings, we create a simple statistical model and show that it reproduces the scaling patterns of collaboration and institution growth, including heterogeneous densification within a collaboration network. Our work provides policy insights to facilitate innovation and methods to infer statistical patterns of complex systems.