Can Mediation Bend the Conflict Curve?

Three documented peace agreements anchor a hopeful read of the last decade — and a sobering one. Each demonstrably eased fighting in its own theatre, yet the global welfare signals for conflict have kept deteriorating. That tension is the question this analysis explores.

The evidence that mediation works. The Colombia and FARC-EP Final Peace Agreement, signed 24 November 2016, formally ended more than five decades of internal armed conflict — Latin America's longest insurgency. It was driven by Juan Manuel Santos, then President of Colombia, on the government side. The Eritrea-Ethiopia Joint Declaration of Peace and Friendship of 9 July 2018, signed by Abiy Ahmed and Isaias Afwerki, declared a 20-year state of war over. And the AU-brokered Pretoria Agreement of 2 November 2022, mediated by the African Union, produced a permanent cessation of hostilities that ended one of the deadliest recent wars.

In Factrail's graph all three strengthen the Peacebuilding and Mediation Efforts driver, and two of them — Colombia and Pretoria — also directly weaken Armed-Conflict Intensity. The model links stronger mediation to fewer deaths in Global Deaths in Armed Conflicts (UCDP) at a roughly 90-day lag. The clearest real-world signature is the 2023 fall in global conflict deaths to about 149,900, down from the 2022 peak near 401,500, after the Tigray ceasefire took hold.

Why the global curve still bends the wrong way. The same indicator shows why optimism must be hedged. Conflict deaths spiked to 145,600 in 2014, eased to about 74,600 by 2018, then climbed steeply through 2020-2022. The driver rationale notes a record number of active state-based conflicts — the most since WWII. Mediation is winning battles theatre by theatre while losing the global count, because new and intensifying wars outpace settled ones. The Ukraine displacement crisis is the starkest example of intensity surging even as other theatres cooled; the dossier marks that fact needs_review and high-sensitivity, so its attribution is stated cautiously as documented by UNHCR.

What is verified versus uncertain. The three agreements and their signing dates are high-confidence and source-backed, including a UN Security Council press statement on the Eritrea-Ethiopia declaration and the African Union's own text for Pretoria. The death-series turning points are real UCDP totals. What is analytical rather than measured is the mediation-intensity series itself: the driver's data points are the Factrail model's estimate of how active and effective mediation was globally, carrying medium confidence. So the direction of the relationship is well supported; its precise magnitude is an estimate.

Alternative explanations. A fall in deaths after a ceasefire need not be wholly the ceasefire's doing — conflicts can also burn out, shift to lower-intensity phases, or be undercounted. And local settlements can coincide with displacement that stays elevated, as the companion analysis on diverging signals shows. The model treats mediation as a genuine counter-pressure, not a guarantee: its weakening effect on deaths is rated medium, and its effect on displacement is weaker still and slower.

What may happen next. If the cadence of mediated settlements continues and no new great-power war erupts, Factrail's structure suggests the global death curve can be held below its 2022 peak — but bending it back toward the ~50,000 normal line would require mediation to scale faster than new conflicts emerge. That is the conditional projection developed in the domain forecast, and it is explicitly a scenario.

Sources. The agreements rest on the African Union Peace and Security Department, a UN Security Council press statement, UNHCR's Refworld archive of the Colombia accord, and corroborating encyclopedic summaries; the death series rests on UCDP totals via Our World in Data and the World Bank. Together they support a measured conclusion: mediation can bend local curves reliably, but bending the global curve is a question of scale and speed, not of whether peacebuilding works.