The Straw that Broke the Camel’s Back: Separating Close Calls from True Onsets in Conflict Prediction

Conflict forecasting seeks to correctly anticipate events of armed violence. While the field has made significant advancements, differentiating between close calls and true onsets has been overlooked despite its potential to improve our ability to predict new outbreaks of war. In this paper, we investigate whether some high-risk situations are more conducive to conflict whereas others will see limited violence only. We argue that the contexts that prevent the transition into more substantial levels of armed conflict display recurring patterns that can be leveraged to enhance conflict prediction. Against this background, we propose an extension of the Shape finder (Schincariol et al. 2025) to separate the underlying processes of true onsets from the close calls of conflict, where low-intensity violence does not surpass a certain level. Drawing on an escalation spiral framework, we theorize mechanisms that drive and prevent conflict onset. Combining published UCDP event data with raw data on low-intensity violence, we apply the Shape finder to match low-intensity conflict patterns to contexts where similar patterns have previously led to conflict outbreak. We use this information to make onset predictions one year ahead. Our preliminary results show that the Shape finder is able to separate contexts that consistently lead to armed conflict from those where intensity remains low, and that this information increases the model's ability to anticipate true onsets. Our findings highlight that the relationship between a predicted high conflict risk and subsequent conflict onset is context specific.