Abstract
This article analyses the motives of violence against civilians during internal wars. It is suggested that soldiers may terrorize civilians because they need the loot to augment their resources while the rest of the time is engaged in fighting proper. An alternative hypothesis suggests that terrorizing the civilian population plays a direct military role. The displacement of large fractions of the civilian population reduces the fighting efficiency of the enemy, as they cannot hide as easily and obtain less support. These two alternative hypotheses are investigated in a simple two-stage game-theoretic model. At stage 1, the government and the rebels simultaneously decide on the level of forces engaged in violence against civilians before they choose the level of forces that they engage in the fighting proper at stage 2. There are two types of sub-game perfect equilibria in this model: there is a pure fighting equilibrium in which no violence against civilians takes place and a pure terror equilibrium. In the latter equilibrium, it is shown that terror substitutes for fighting if the government can afford it. Predictions of the model are tested using African refugee data. In accordance with the theoretical model, the refugee population displays strong positive serial correlation, and after controlling for war, overseas development assistance has a positive impact on the outflow of refugees. Thus, the results support the hypothesis that violence against civilians is motivated by military objectives and suggest that donor funding to governments at war should be cut if the protection of civilians is regarded as more important than the fate of the fighters.
