This article asks why armed groups form high-end cooperation—training or intelligence sharing—even though such ties raise downstream political risks. The author develops a principal–agent theoretical logic: principals (providers of high-end support) mitigate risk by choosing agents who are skill-complementary but politically non-substitutable, while agents accept support only when it does not politically overshadow them. The study uses original network data on 53 groups in Northeast India from 1981–2021 and applies temporal exponential random graph models (TERGMs) to test the theory, producing evidence about organizational capacity differences, complementary attack portfolios, and constituency distinctiveness as correlates of high-end cooperation.
