Sexual violence along ethnic lines? Revisiting rebel-civilian ethnic ties and wartime sexual violence

Abstract

Previous research suggests that wartime sexual violence by rebel groups should generally be committed between rather than within ethnic groups. Since rebels can mobilize through and draw support from coethnic civilian networks, they should be less prone to commit sexual violence against their ethnic brethren. Moreover, ethnic divisions between groups are argued to spur inter-ethnic sexual violence as a strategy of war. Yet, much remains to be tested empirically. A major hindrance has been the lack of data on sexual violence that captures the ethnic identities of victims. This issue is circumvented by geocoding occurrences of sexual violence from the SVAC dataset and intersecting these with geographic patterns of ethnic settlement. Interestingly, the results show no indication of restraint in coethnic areas. They also indicate that mixed areas with both coethnic and non-coethnic civilians are more likely to experience sexual violence than entirely non-coethnic areas.

Investigaciones anteriores sugieren que la violencia sexual en tiempos de guerra por parte de grupos rebeldes generalmente debería cometerse entre grupos étnicos y no dentro de ellos. Dado que los rebeldes pueden movilizarse a través de redes civiles coétnicas y obtener el apoyo de estas, deberían ser menos propensos a cometer actos de violencia sexual contra sus hermanos étnicos. Además, se argumenta que las divisiones étnicas entre grupos estimulan la violencia sexual interétnica como estrategia de guerra. Sin embargo, aún queda mucho por comprobar empíricamente. La falta de datos sobre la violencia sexual que capten las identidades étnicas de las víctimas ha sido un obstáculo importante. Este problema se evita mediante la geocodificación de los casos de violencia sexual del conjunto de datos sobre la violencia sexual en los conflictos armados y mediante su intersección con los patrones geográficos de asentamiento étnico. Curiosamente, los resultados no muestran ningún indicio de moderación en las zonas coétnicas. También indican que las zonas mixtas con civiles coétnicos y no coétnicos tienen más probabilidades de sufrir violencia sexual que las zonas totalmente no coétnicas.

Une recherche antérieure laisse à penser que les violences sexuelles perpétrées en temps de guerre par des groupes rebelles devraient généralement être commises entre groupes ethniques plutôt qu’au sein d’un même groupe ethnique. Les rebelles, pouvant se mobiliser grâce aux réseaux civils de leur ethnie et obtenir le soutien de ces derniers, devraient être moins enclins à commettre des violences sexuelles à l’encontre de leurs frères ethniques. De plus, on estime que les divisions ethniques entre les groupes stimulent les violences sexuelles entre les ethnies en tant que stratégie militaire. Toutefois, beaucoup reste à tester empiriquement. L’absence de données sur les violences sexuelles enregistrant l’identité ethnique des victimes constituait un obstacle majeur. Le géocodage des occurrences de violence sexuelle émanant de l’ensemble des données SVAC et leur recoupage avec les modèles géographiques de peuplement ethnique ont permis de contourner ce problème. Curieusement, les résultats ne montrent aucune indication de retenue dans les zones coethniques. Ils indiquent également que les zones mixtes, peuplées à la fois de civils de la même ethnie ou non, ont davantage de probabilité de connaître des violences sexuelles que celles qui sont entièrement peuplées d’ethnies différentes.

Keywords:

• Armed conflict
• ethnicity
• sexual violence

Data Availability Statement

The geocoded SVAC data used for this study is openly available in Mendeley Data, at: http://doi.org/10.17632/fyyp694myt.1.

Acknowledgments

I owe thanks to Karin Johansson, Maxine Leis, and Angela Muvumba Sellström as well as three anonymous reviewers for their excellent feedback and comments on earlier versions of this paper. I am also immensely thankful to Lisa Hultman for giving me advice and guidance throughout the process.

Notes

1 This paper is limited to sexual violence committed by rebel actors due to the scope of the restraint mechanism.

2 Aggregating individual-level phenomena to geographic areas introduces issues of measurement reliability. I mitigate these issues through alternative specifications of what constitutes a coethnic area in the robustness tests.

3 As the entire time-period is set prior to the independence of South Sudan, the collective admin units of Sudan and South Sudan are treated as one country. Instances of sexual violence in Sudan are therefore geocoded using all admin units of modern-day Sudan and South Sudan and rebel groups are expanded across all these units. The same procedure was employed for Eritrea, where all modern-day admin units of Eritrea were treated as part of Ethiopia 1989–1993, the year of independence. The independence of Namibia from South Africa in 1990 is unproblematic as neither country appears as a location in SVAC.

4 The proportion of sexual violence observations that could be geocoded is higher in Africa compared to all other continents.

5 Neither SVAC 2.1 nor Geo-SVAC can be used for the geocoding. The former dropped the location variable from the update and the latter offers geolocation of conflict events involving actors that use sexual violence, not actual events of sexual violence.

6 The main findings are also robust to alternative thresholds, see the robustness tests.

7 Model 1 is chosen in lieu of model 2 to increase model performance. There is not enough variation within the coethnic category of ethnic ties to include the interaction term with a reduced sample size. Even in the full set, there is just one event of sexual violence in a coethnic area by a rebel group with a weak command.

8 See Supplementary Appendix A for the full results of the robustness tests.

Additional information

Funding

This work was funded by the Knut and Alice Wallenberg Foundation under Grant number 2018.0455.