The paper investigates the effect of child malnutrition on the risk of mortality in Burundi, a very poor country heavily affected by civil war. We use anthropometric data from a longitudinal survey (1998-2007). We find that undernourished children, as measured by the height-for-age z-scores (HAZ) in 1998 had a higher probability to die during subsequent years. In order to address the problem of omitted variables correlated with both nutritional status and the risk of mortality, we use the length of exposure to civil war prior to 1998 as a source of exogenous variation in a child’s nutritional status. Children exposed to civil war in their area of residence have worse nutritional status. The paper finds that one year of exposure translates into a 0.15 decrease in the HAZ, resulting in a 10% increase in the probability to die for the whole sample as well as a 0.34 decrease in HAZ per year of exposure for boys only, resulting in 25% increase in the probability to die. We show the robustness of our results. Food and income transfer programs during civil war should be put in place to avoid the long- term effects of malnutrition.