
Repair, Capacity, and Collapse: A Mechanistic Early-Warning Model of Political Rupture
Abstract
Governance ruptures, coups, revolutions, and state collapses, occur when the state’s capacity for repair fails to match accumulating violations of legitimacy, security, or welfare. This article terms this a Contradiction Debt (CD), a model that formalizes this imbalance as the ratio of effective repair to total violation (R/V). Building on theories of grievance, capacity, and legitimacy, the CD framework quantifies how ‘repair deficits’ erode trust and predict the timing of rupture. Using quarterly data from 52 country cases (2005–2025), the model distinguishes high-risk from low-risk cases about as well as leading fragility indices do (measured by AUROC, a standard score of how well a model separates the two, where 1.0 is perfect and 0.5 is no better than chance; for the CD model, AUROC ≈ 0.54). However, the CD model has superior temporal precision, and is able to forecast most ruptures within ± one quarter. The framework was recently applied to the case of Hungary (not in the sample) and was able to specify not only the timing but also the form of the institutional rupture – constitutional confrontation at oversight institutions. This was confirmed by the April 2026 electoral rupture at exactly those institutions. Statistical regression models confirm that the R/V ratio adds predictive value beyond existing risk indicators. The decision-curve analysis – a method for evaluating whether acting on a model’s alerts yields better outcomes than either alerting on every case or none – confirms the practical usefulness of this model for taking preventive action. By linking grievances and state capacity through a measurable repair process, the CD model transforms opaque early-warning signals into interpretable diagnostics of governance strain. It thus complements event-based systems such as ACLED and ViEWS.
© 2026 David Koepsell, published by Department of Peace Studies and International Development, University of Bradford
This work is licensed under the Creative Commons Attribution 4.0 License.