The impact of violence on child health has long-lasting consequences that increase the overall cost of conflict. Beyond the damage caused to direct victims of violence, behavioral responses to insecurity can lead to major health setbacks for young children. The fear of ex- posure to conflict events often triggers such responses even before/without any manifestation of violence in a given area. This generates a treatment status (exposure to conflict risk) that goes beyond violence incidence. In this paper, I develop new metrics that capture perceived insecurity at the local level through a statistical model of violence in order to investigate the impact of conflict on child health. Violence is modeled as a space-time process with an unknown underlying distribution that drives the expectations of agents on the ground. Each observed event is interpreted as a random realization of this process, and its underlying dis- tribution is estimated using adaptive kernel density estimation methods. The new measure of violence risk is then used to document the effects of conflicts in Ivory Coast and Uganda on child health. The empirical evidence suggests that conflict is a local public bad, with cohorts of children exposed to high risk of violence equally suffering major health setbacks even when the risk does not materialize in violent events around them.