The Effect of Conflict on the Risk of Experiencing Sexual Violence in Kivu

The aim of this thesis is to explore a new way of estimating to what degree the conflicts in eastern Congo, more specifically the Kivu regions, have altered the risk of experiencing sexual violence. I estimate this conflict-effect by combining two methods. These are event history analysis and the synthetic control group method. The first method has earlier been used to study the effect of conflict on age at sexual debut in a case study concerning the genocide in Rwanda (Elveborg Lindskog, 2014). The latter method has previously been used to study the conflict-effect on economic growth, in the first paper that presents this method (Abadie & Gardeazabal, 2003). Both methods have proven to be useful tools for estimating effects of conflict, but have never been combined before. The result of this two-stage analysis shows that the conflict in Kivu increased the risk of experiencing sexual violence by 58 percent at the most, before decreasing as a result of heightened risk in the control-Kivu by the second half of the conflict. The estimated probability of experiencing sexual violence by age 35 is 0.33, which is 38 percent higher than for the synthetic Kivu. These conflict-effects are large, but could still be downward biased, as it is probable that there have been some spillover-effects from the conflict in Kivu to other regions. As there are only nine regions that have experienced little or no conflict during the period observed here, it is not possible to draw inferences of the results that are statistically significant on a level lower than 12.5 percent. On the other hand, the measured development of the conflict effect is a more precise measure than earlier works that study sexual violence in Kivu. Peterman et al. found that Kivu has a positive differential likelihood of having experienced sexual violence, but were not able to estimate any development of this likelihood or identify causal effects (2011). Other studies have investigated patient-journals at hospitals in Kivu that treat victims, and find conflict-related trends in the type of SV and scope, but these are not representative for the whole region, and do not give information on other regions (Bartels et al., 2011; Bartels et al., 2010). Thus I believe the combination of the two methods done in this analysis opens up for new possibilities concerning measuring how the effect of conflict alters different aspects of people’s lives over time.