Zusammenfassung:
This paper develops a novel measure of state fragility: the probability of state failure. We define state failure as the inability to perform core functions and operationalize it through observable breakdowns- conflict, territorial control, institutional deficiency, and public service deterioration. Using a machine-learning approach, we estimate failure probabilities for 160 countries and assess their predictive performance and key predictors. The proposed measure is forward-looking, continuous, comparable across countries and over time, and exhibits strong predictive power. Political and institutional factors are the main predictors of failure, whereas aggregate macroeconomic conditions contribute little. State failure is highly persistent, different breakdowns tend to accumulate, and conflict emerges as the predominant mode of failure. This framework provides a clear and theoretically grounded approach to analyse fragility that is useful for both academic research and policy analysis.