This paper exploits exogenous variation in Burundi’s latest episode of civil war violence to estimate the causal impact of exposure to violence and the probability of growing coffee, some four years after the war ended. I constructed a unique panel dataset for a nationally representative sample of 242 households in 35 communities. Household fixed effects estimates show that exposure to individual violence decreases the probability of growing coffee in the post-war period by some 16%. This effect seems to be driven by a household member’s experience of physical violence rather than theft of crops or assets. The results are consistent with the idea that physical violence may decrease household labour availability if such violence caused people to die or become physically or mentally disabled. On the other hand, I cannot rule out that changing risk and time preferences in response to violence may also play a role.