The paper proposes a general conceptual framework for state fragility that aims to disentangle the identification of fragility from its underlying drivers. It reduces the identification state fragility to the responsiveness of its economic or political systems to shocks. In an extension of Taleb’s work, fragility is thus defined as a disproportionate (supra-linear) response to negative shocks and an underwhelming (sub-linear) response to positive shocks or to time. Consequently, the framework distinguishes between fragility to stress as breakdowns in political or economic systems in response to negative shocks and chronic fragility as the inabilty of economies to generate growth over time. The framework can be applied to both manifested fragility (where the list of fragile states it produces closely aligns with classifications by international organizations) and to latent fragility (where underlying drivers can offer a probabilistic proximity to a fragile response). We illustrate how latent fragility can be identified by focusing on coups d’état as reflective of political fragility: we use machine learning to examine coup drivers and their nonlinear interactions, and to derive implied coup probabilities across all countries. The paper parses out lessons for engagement with fragile states, especially on the importance of prioritizing the elimination of sources of fragility through strengthened structural fundamentals and the mitigation of stressors through stronger policies, both of which have higher returns in chronically fragile environments.