Women in Armed Rebellion Dataset (WARD)

The Women in Armed Rebellion Dataset (WARD) currently includes information on the prevalence of female fighters for a sample of 211 rebel organizations active between 1979 and 2009. We base our sample on the groups included in the Uppsala Conflict Data Program’s (UCDP) Dyadic Dataset (Harbom, Melander & Wallensteen, 2007). Relevant information on women’s participation in rebel groups was collected through an extensive search of news reports, academic accounts, and international and non-governmental organization reports. In order to make a determination about women’s inclusion and roles within an organization, we required the confirmation of three independent sources. Furthermore, because we are interested in women’s participation as combatants and fighters, we required evidence that a group formally incorporated women in the organization and deployed them specifically in combat roles.

Female Combatants, 1979-2009


In conceptualizing and defining female fighters, we draw on definitions commonly used in disarmament, demobilization, and reintegration (DDR) programs sponsored by the United Nations and related international organizations. These programs often differentiate ‘female combatants’ from ‘females associated with armed groups’ (see UN Women, 2012: 22-23). The first category, upon which we focus, refers to the subset of female group members who are armed and participate in organized combat activities on behalf of a rebel organization. This includes women employed in frontline combat, female suicide bombers, or assassins, and female auxiliaries or members of civil defense forces who receive military training, carry combat weapons, and could be called upon to participate in combat when necessary. It excludes women who served in other capacities such as fundraisers, recruiters, couriers, or informants but do not engage directly in organized violence on behalf of an organization.

Distribution of Female Combatants in Sample


We construct two indicators reflecting the presence of female fighters in armed groups. The first is a dichotomous variable indicating the presence of any number of female fighters. We consider female fighters absent from the group where reports explicitly state that women did not participate in combat, where women’s roles were described as exclusively supportive (e.g. caregivers, fundraisers, couriers, etc.), or where we were unable to locate any evidence of women participating in combat despite locating substantial information regarding other group characteristics.

We also construct  a categorical indicator accounting for the estimated proportion of a group’s combat force comprised of women. We construct a categorical indicator rather than a direct estimate of the proportion of female combatants in an armed group because different sources sometimes provide varying estimates of the numbers of women serving as combatants and occasionally provide only qualitative descriptions of the extent of women’s participation. Therefore, we utilize a blunter coding scheme to increase our confidence that we accurately capture the prevalence of female combatants within rebel groups, although doing so reduces the precision of the resulting measure.

Finally, owing to differences in the numbers of female combatants reported in different sources and occasional difficulties in differentiating between women in combat and non-combat roles , we include both a “best” and “high” estimate for each of the above measures.

The WARD was first published in the Journal of Peace Research.  Please include the following citation when using these data: Wood, Reed M. and Jakana L. Thomas. 2017. “Women on the Frontline: Rebel Group Ideology and Women’s Participation in Violent Rebellion“, Journal of Peace Research 54(1): 31-46.

Replication data for this article can be downloaded here.

Help us improve our data collection

Locating accurate information on women’s participation in armed movements is extremely challenging, and we recognize the possibility that we have overlooked women’s contributions in some cases and exaggerated their contributions in others. Our categorical estimates reflect only the information we have successfully uncovered in a multi-year, multi-researcher data collection effort. If you have believe you have found an error in our coding or have additional information you believe we might have overlooked, please feel free to email the project manager.