We analyse the effect of violence and rebellion on the evolution of household welfare. We collected new panel data for Burundi (1999-2007) in which we re- interviewed original as well as newly formed households (split-offs). We use several definitions of the household as unit of analysis and test for resource pooling between parental and split-off households. Focusing on the effect of civil war, we find that village- level violence, measured as the number of battle-related deaths or wounded reduces consumption growth by 9% for every 25 casualties. Joining an armed rebel group was a lucrative livelihood strategy: households of which at least one member joined an armed group experienced 41% higher growth in welfare over the study period. Results are robust to alternative variables of civil war shocks and model specifications, including household fixed effects and initial household fixed effects.