Femicide in Brazil (2017–2025): Temporal Regimes, Structural Dynamics, and Scenario Projections via GAM-BN Modeling

This article examines the dynamics of femicide in Brazil between 2017 and 2025 and projects scenarios for 2026 through the identification of temporal regimes and the modeling of count data. The series exhibits non-linear behavior, characterized by alternating phases of growth, stabilization, and recent inflection. A generalized additive model with a negative binomial distribution (GAM-BN) is employed to account for overdispersion and trend variation. The findings are consistent with the presence of three regimes: initial growth, stabilization at high levels, and a recent resurgence of upward trends. The projection for 2026 suggests a central value around 1,618 cases and is interpreted as a set of scenarios defined by prediction intervals rather than a single point estimate. The continuation of the historical trend concentrates most of the probability mass, while alternative scenarios remain possible. From a substantive perspective, the relative stabilization of femicide at high levels suggests the structural persistence of gender-based violence despite normative and institutional advances. The article contributes by integrating regime identification, flexible modeling, scenario interpretation, and sociological reflection on the temporal dynamics of femicide, offering a non-deterministic perspective on the phenomenon with implications for policymaking under uncertainty.