Abstract
Women participated as combatants in almost 40% of civil conflicts that occurred between 1979 and 2009. We offer a novel argument about the effect of female combatants upon the outcomes of the civil conflicts that they join. Groups that recruit female combatants are more likely to achieve victory in their conflicts than are groups that do not recruit female combatants. However, when rebel groups rely upon forced recruitment, they risk undermining the benefits associated with female combatants, lowering their likelihood of victory relative to that of the government. We test this conditional argument using multinomial logistic regression models on a sample of 194 rebel groups globally from 1979 to 2009. We find that female participation appears to decrease the likelihood of government victory in civil wars; this effect holds primarily in instances in which female participation could plausibly be thought of as voluntary. Forced female participation, by contrast, appears to increase the likelihood of a government victory.
Introduction
Almost 40% of civil conflicts that occurred between 1979 and 2009 included women combatants. An emerging literature quantifies and explains their participation in rebel groups (Thomas and Bond, 2015; Henshaw 2016a, 2016b; Wood and Thomas, 2017). We offer two contributions to this literature. First, we argue that groups that recruit female combatants are more likely to achieve victory in their conflicts as compared to groups that do not recruit female combatants. Second, we argue that this positive effect may be conditioned by those groups’ methods of recruitment. If groups rely upon forced recruitment, they may undermine the benefits associated with female participation.
There are a number of reasons why rebel recruitment of female combatants might be associated with a higher likelihood of achieving victory. Female combatants may reflect and enhance a rebel group’s horizontal ties to the local society. Horizontal ties may endow groups with additional recruits and resources and make them more resilient in the face of government crackdown. Female combatants may also enhance a group’s international profile, increasing the potential for the group to receive much-needed external support via so-called ‘vertical ties.’ This support may help them to better sustain their challenge against the government or make them optimistic regarding the prospects for an external intervention on their behalf.
We complicate this argument by suggesting that the recruitment of female combatants is not a panacea for all rebel groups. Each of the logical positions that lead to a general expectation of benefits from female recruitment arguably rely upon the premise of voluntary participation. We contend that when groups employ methods of forced recruitment, they likely undermine many of the strategic advantages of female participation.
We test these expectations by specifying a multinomial logistic regression that compares the influence of female combatant participation in rebel groups upon the likelihood of victory for both rebels and governments. This model draws upon a sample of almost 200 rebel groups globally from 1979 to 2009. Our large-N results partially reflect the validity of our conditional hypothesis. In general, female participation in rebellion appears to decrease the likelihood of victory for the governments that they target; however, this effect holds primarily in instances in which female participation could plausibly be thought of as voluntary. Forced female participation, by contrast, appears to enhance the government’s likelihood of victory.
The benefits of female combatants in rebellion
Female combatants have been important in many rebel groups (Kampwirth, 2002; Sjoberg and Gentry, 2007; MacKenzie, 2012; Cohen, 2013; Henshaw, 2016a, 2016b). Explanations for female combatant participation have stressed supply- and/or demand-side dynamics. On the supply-side, women are shown across a range of conflicts as having taken up arms to assuage economic or religious grievances that might affect their communities rather than serving personal interests (Mason, 1992; Coulter et al., 2008; Henshaw, 2016b).
From the demand-side perspective, which speaks more directly to questions of efficacy, organizational needs appear to drive recruitment of female participants (Wood, 2017). Rebel groups want to increase their membership numbers in order to maintain their war-fighting efforts (Thompson, 2006; Coulter et al., 2008; Thomas and Bond, 2015). Female combatants may also contribute legitimacy to the groups that they join, convincing observers of the relative unity of the group and its communities (Barth, 2002), as well as enhancing their popular support domestically and internationally (Parkinson, 2013). Their participation may signal the group’s resolve and commitment to achieving a favorable outcome (Mason, 1992; Kampwirth, 2002; Sjoberg and Gentry, 2007). Each of these effects is likely to operate above and beyond that observed in instances in which groups rely exclusively upon male combatants. In other words, female participants may help to solidify the group’s horizontal ties to others within society and vertical ties to potential sources of support external to the state. Accordingly, we anticipate that:
With respect to horizontal ties, the Eritrean People’s Liberation Front (EPLF) benefited from female participation in being able to sustain supply chains to local society (Thomas and Bond, 2015: 489). The South West Africa People’s Organization faced initial concern that highlighting their considerable gender diversity might undermine group cohesion (Olsson, 2005; Thomas and Bond, 2015). Ultimately, however, their inclusive membership helped them to generate vertical ties, attracting recognition, aid, and safe-haven from the UN General Assembly (from Norway and Angola, respectively). In both instances, rebel access to these horizontal and vertical ties also served to limit the government’s likelihood of achieving victory.
We argue, however, that not all rebel groups that recruit female combatants are equally likely to reap the benefits associated with their participation. We expect that there is an important exception when rebel groups depend upon methods of forced recruitment. As can also be true for child and male participants, female participation in rebel groups is often involuntary (Brett, 2002; Mazurana et al., 2002; Henshaw, 2016b). Although women may continue to possess some agency (McKay, 2005; Wood, 2017), methods of forced recruitment are highly likely to undermine the benefits their participation brings to the group. Women that have been forcibly recruited are less likely to bring to bear their horizontal ties and networks to local communities; and forced recruitment might be expected to increase the likelihood of vertical ties providing support against rebel groups rather than in their favor. Accordingly, we anticipate that:
The case of the West Side Boys in Sierra Leone demonstrates this. This small band of former soldiers of the Sierra Leone Army emerged in 1999 largely off the back of the forced recruitment of combatants (McGreal, 2000). Their reliance upon forced methods of recruitment and employment of tactics of sexual violence to command control over their recruits and local populations likely undermined their ties to local communities (as demonstrated elsewhere; see, e.g. Cohen, 2013; Cohen and Nordås, 2015). These coercive actions also provoked a decisive response from international actors against rather than in support of their interests. Intervening British forces were deployed in defense of British citizens and UN peacekeeping forces subjected to violence by the West Side Boys; however, the protection of vulnerable Sierra Leonean populations was also an important motivation (McGreal, 2000).
The effects of female combatants and their recruitment on rebel outcomes
Our primary task is to assess the effect of female combatants upon the likelihood of rebel or government victory in civil conflicts. To measure our dependent variable, the outcome of the civil conflict, we employ the Uppsala Conflict Data Program’s Conflict Termination dataset, which details the outcome of all conflicts resulting in a minimum of 25 battle-related fatalities (Kreutz, 2010). Specifically, we code a multinomial variable indicating whether the conflict ended in ‘Government Victory,’ ‘Rebel Victory,’ or ‘other’ (the ‘other’ category is our own reference, and includes conflicts that remain ongoing, drop to low levels of activity, or result in negotiated agreement). 25 of the 194 conflicts in our sample end in victory for either the government (13) or the rebel group (12).
To test our hypotheses, we rely upon two independent variables drawn from the Women in Armed Rebellion Dataset (WARD; Wood and Thomas, 2017). First, we use a binary indicator of the presence or absence of female combatants on the side of the rebel group at any stage during their conflict. Second, we use a binary variable that is coded as a ‘1’ for rebel groups that rely upon forced methods of recruitment. We then introduce an interaction term that is the product of these two binary variables. This interaction term enables us to test our conditional hypothesis. These data are available for a majority of rebel organizations active between 1979 and 2009.
There are two important limitations to these WARD data. First, our data on female combatant participation are time-invariant. Accordingly, it is not possible to directly explore when female combatants join a group nor whether their participation affects groups’ horizontal or vertical social ties. Second, the ‘Forced Recruitment’ variable does not specify that the female combatants themselves were necessarily recruited forcibly but, rather, that the group has forcibly recruited some proportion of its combatants. Accordingly, in generating our independent variable, we are assuming that groups that depend upon forced methods of recruitment are likely to use these methods for male and female combatants alike.
We conduct a multinomial logistic regression model to uncover the general correlates of decisive civil conflict outcomes. Our model specification includes four conflict-level characteristics: the number of years of conflict (‘Conflict Duration’), the number of rebel groups concurrently fighting against the government (‘Number Rebel Dyads’), and binary indicators of whether the rebel group is at parity with or stronger than the government in terms of military strength (‘Rebel Fighting Advantage’) or capacity to mobilize participants (‘Rebel Mobilization Advantage’). Each of these variables have been commonly employed elsewhere in models of conflict outcomes (Cunningham et al., 2009; Chu and Braithwaite, 2017). We also include three country-level control variables: the regime’s Polity score (Vreeland, 2008), logged GDP per capita, and logged population from the final full year of the conflict. Summary statistics for all variables are reported in Table 1.
Summary statistics.
DV: dependent variable; Ln: log transformation.
Results
Table 2 details the results of our main analysis. The two columns reflect our decisive conflict outcomes: government victory and rebel victory. Given that all variables are binary with meaningful zero values, we can offer some direct interpretation by comparing parameter estimates to the ‘other’ reference category. Our findings align with one implication of our hypotheses. In the first column, we observe a negative and statistically significant coefficient estimate for ‘Female Combatants’ and a positive and statistically significant coefficient estimate for the interaction term between female combatants and forced recruitment. These estimates tell us that when rebel groups recruit female combatants there is a reduced likelihood of government victory. However, the Government is more likely to achieve victory when rebel groups forcibly recruited their female combatants.
Multinomial logistic regression results.
Note. Base Outcome includes ‘Negotiated Agreement,’ ‘Ongoing Conflict,’ and ‘Low-Level Activity.’ Robust SEs clustered in the country are in parentheses.
Ln: log transformation.
p < 0.10.
p < 0.05.
p < 0.01.
We do not, however, observe statistical significance for any of our key explanatory variables with respect to the likelihood of rebel groups achieving victory. The recruitment of female combatants, by whatever method, does not appear to have a direct bearing upon the likelihood of rebel victory. Thus, we conclude that female combatants can help to undermine a government’s likelihood of victory even if they do not directly boost the rebels’ likelihood of victory.
A comparison of predicted probabilities offers us a means of identifying the substantive effects of these key explanatory factors. Table 3 summarizes the predicted probabilities associated with our core regression results. Here we compare probabilities associated with the four conditions described by the interaction between our two binary variables: the presence/absence of female combatants and the use/non-use of methods of forced recruitment. In line with the general trends identified in the multinomial model, we can see that, ceteris paribus, when rebel groups recruit female combatants through voluntary methods, there is a 100% decrease in the likelihood of their opponent achieving victory, as compared to groups that do not recruit female combatants nor rely upon forced recruitment. We observe, furthermore, that this beneficial effect of female recruitment is undermined somewhat in instances in which groups also rely upon forced recruitment. Under such conditions, there is just a drop of 68%. Table 3 also details predicted probabilities associated with rebel victory. As noted earlier, these results were not statistically significant. Nonetheless, it is interesting to note that methods of forced recruitment are associated with sharp reductions (58%) in rebels’ likelihood of victory.
Predicted probabilities of government and rebel victories.
Elsewhere, our models appear to return results largely consistent with other studies exploring conflict outcomes (Cunningham et al., 2009; Chu and Braithwaite, 2017). Government victory is shown to be less likely as the conflict endures and when they face rebel groups that enjoy a fighting advantage. At the same time, rebel victory is more likely when rebel groups enjoy either a fighting or mobilization advantage, as well as in larger, poorer, less democratic countries.
Decisive outcomes in civil conflicts are relatively rare. We therefore complement our large-N analysis with a closer look at some decisive cases. We have three main goals in examining these cases. First, do the cases appear, in general, to fit with the expectations of our argument? Second, do we indeed see groups employing forcible recruitment targeting female combatants using this method? Third, what bearing does the forcible recruitment of female combatants appear to have upon the prospects of these groups? Table 4 provides an overview of all 25 cases of decisive victories in civil conflicts in our sample.
The population of government and rebel victories.
AFDL: Alliance for the Democratic Liberation of Congo-Zaire; ATNMC: Alliance Touaregue Nord Mali pour le Changement; BDK: Bundu dia Kongo; DRC: Democratic Republic of Congo; EPRDF: Ethiopian People’s Revolutionary Democratic Front; FAN: Forces Armeés du Nord; GAM, Gerakan Aceh Merdeka; JVP: Janatha Vimukthi Peramuna; LTTE: Liberation Tigers of Tamil Eelam; MPS: Patriotic Salvation Movement; NRA: National Resistance Army; PULF: People’s United Liberation Front; RPF: Rwandan Patriotic Front; UIFSA, United Islamic Front for Salvation of Afghanistan; UNLF: Uganda National Liberation Front; UPA: Uganda People’s Army; USC/SSA: United Somali Congress/Somali Salvation Alliance; UTO: United Tajik Opposition; UWSA: United Wa State Army; WMBD: Wahhabi Movement of Buinaksk District; WSB: West Side Boys; WSLF: Western Somali Liberation Front.
Our primary interest lies in the first four cells of this table — each of the cases involving female participants. In terms of the general fit of these cases to our argument, the first noteworthy point is that there are zero cases of government victory when competing against a rebel group that enjoyed exclusively voluntary participation by female combatants. Moreover, there are two cases in which women are voluntarily recruited into rebel groups and those groups subsequently achieve victory. The National Resistance Army’s defeat of the Government of Uganda in 1986 depended heavily upon their recruitment of female volunteers who helped to reinforce their popular support through horizontal ties to local communities (Kasfir, 2005). In a similar fashion, the Rwandan Patriotic Front’s voluntary recruitment of female combatants bolstered their credentials as a community-oriented and inclusive group and, thus, enhanced their reputation as a umuryango – an association that spanned Hutu, Tutsi, and Two populations. This credibility was central to them being able to defeat both military and civilian perpetrators of the genocide in 1994 (Purdeková, 2011). On their face, the cases listed in these first two cells would appear to provide strong evidence in line with our test hypotheses.
With respect to our second and third concerns, there are four cases that involve female combatants and forced recruitment. In all four cases, it is clear that female combatants were indeed forcibly recruited. In the first two cases, consistent with the expectation of our second hypothesis, the rebel groups were subsequently defeated. The first is the case of the aforementioned West Side Boys of Sierra Leone. The British raid of 10 September 2000 freed 22 Sierra Leoneans, including five women who were abducted months earlier (McGreal, 2000). These abductions triggered opposition among external actors. Certainly, it was unlikely that the West Side Boys would ever defeat the Sierra Leonean government; however, their demise certainly appears to have been hastened by their forcible recruitment of women.
In the second case, the Liberation Tigers of Tamil Eelam (LTTE) in Sri Lanka recruited large numbers of women into their ranks. They famously urged all Tamil families to contribute at least one woman and although many did so voluntarily, their recruitment campaign was always maligned by rumors of forcible recruitment (De Silva, 1994). Alison’s (2003) interviews with 14 female former LTTE combatants demonstrates that many joined the LTTE because they feared and wanted revenge for the common raping of Tamil women by Indian and government forces. It is in this context that the LTTE’s occasional reliance upon forced recruitment may have backfired in undermining their support from women. Of course, by the time of their eventual defeat in 2009, there are a number of reasons why the government achieves victory; and this record of forced recruitment can only have been a relatively small contributing factor.
Finally, we observe two cases in which women are forcibly recruited into rebel groups and yet they still achieve victory. These two cases clearly contradict the expectations of our conditional hypothesis. Nonetheless, both can perhaps be understood upon closer examination. The Khmer Rouge defeated the Government of Cambodia in 1975. They recruited many women to their ranks using both voluntary and forcible methods. It is possible that they were nonetheless able to benefit from female participation by virtue of having promoted women to prominent roles within the movement. Long Thirith, for example, served as the movement’s Minister for Social Affairs after they achieved their victory (Smith, 1994). In a second case, the Ethiopian People’s Revolutionary Democratic Front (EPRDF) enjoyed victory over the Government of Ethiopia in 1991 after a 2-year struggle even though they had relied upon forced methods of recruitment. We speculate that the EPRDF’s victory may have benefited from their ability to piggy-back off of the efforts of the EPLF to establish horizontal ties to local social groups. This seems plausible, given that the EPLF fought a 30-year struggle for Women’s liberation.
Discussion
We contribute to the emerging cross-national empirical literature on female combatants in rebel groups. We focus upon the effects that female combatants have upon the likelihood of victory for the rebel groups that they join. We argued that although female combatants ought to have a generally positive effect, this would depend upon the mode of recruitment employed by the rebel group. When female combatants are likely to have been forcibly recruited, we argue that their benefits to the rebel group are undermined. Multinomial logistic regressions provide support for our hypotheses.
Our findings are preliminary, but highlight a topic that warrants further investigation. In particular, we believe that two limitations ought to be addressed. First, the WARD dataset has provided a welcome boon to the study of female combatants in civil conflicts. In moving forward, it is necessary to generate time-varying data on when female combatants enter conflicts. This would enable more direct evaluation of the causal mechanisms underlying their subsequent effect upon the conflicts that they join.
Second, the present study has depended upon an assumption that groups that employ forced recruitment in general are employing it specifically to recruit their female participants. Our closer look at six cases appears to verify the validity of this assumption in a limited setting. Improved data could help researchers to validate this assumption further and explore the specific interactions between rebel groups and their female recruits. Future research could also consider whether or not groups are also engaged in child soldiering, which ought plausibly to have a similar negative effect on group prospects by undermining horizontal and vertical ties.
Footnotes
Declaration of conflicting interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) received no financial support for the research, authorship, and/or publication of this article.
Carnegie Corporation of New York Grant
This publication was made possible (in part) by a grant from Carnegie Corporation of New York. The statements made and views expressed are solely the responsibility of the author.