This paper explores the importance of the risk of violence on the decision making of rural households, using a unique panel data set for Colombian coffee-growers. We identify two channels. First, we examine the direct impact of conflict on agricultural production through the change in the percentage of the farm allocated to coffee. Second, we explore how conflict generates incentives to substitute from legal agricultural production to illegal crops. Following Dercon and Christiaensen (2011), we develop a dynamic consumption model where economic risk and the risk of violence are explicitly included. Theoretical results are tested using a parametric and semi-parametric approach. We find a significant negative effect of the risk of violence and the presence of illegal crops on the decision to continue coffee production and on the percentage of the farm allocated to coffee. Results are robust after controlling for endogeneity bias and after relaxing the normality assumption.