Abstract
Abstract
Between 2023 and 2025 the Colombian government decreed a sequence of bilateral ceasefires with the National Liberation Army (ELN), the Estado Mayor Central (EMC) and its splinter Estado Mayor de Bloques y Frente (EMBF), the Segunda Marquetalia, the Autodefensas Gaitanistas de Colombia (EGC, "Clan del Golfo") and the Autodefensas Conquistadoras de la Sierra Nevada. A widespread interpretation holds that these ceasefires were exploited: armed groups stopped fighting the state while expanding against rivals and civilians. This article tests that displacement hypothesis with event-level data. I combine the UCDP Georeferenced Event Dataset (version 26.1) and the UCDP Candidate Events Dataset (January–June 2026) into an actor × department × month panel of six armed groups in 31 departments (2018–2025; 7,860 observations), code ceasefire exposure from the fourteen presidential decrees that opened, extended, suspended or ended each truce, and estimate Poisson pseudo-maximum-likelihood difference-in-differences models separately for state-based, non-state and one-sided violence. Ceasefire exposure is associated with an 86–94% reduction in state-based events involving the covered group (incidence-rate ratio 0.06–0.14 in the preferred national-time specifications), robust to fatalities as outcome, continuous exposure, exclusion of the EMC and extension into 2026, and absent in a placebo test with windows shifted 24 months earlier. I find no evidence that violence was displaced into inter-group fighting: non-state estimates are below or near one in every specification, and non-state fatalities fall. Effects on violence against civilians are imprecise and sign-unstable. The large rise in inter-group violence after 2023 is shared by the uncovered Clan del Golfo, so naive before–after comparisons overstate displacement. Ceasefires froze the state front without detectably reallocating lethal violence; their failure lay elsewhere.
JEL codes: D74, H56, K42, O54