The effects of conflict and displacement on violence against adolescent girls in South Sudan: the case of adolescent girls in the Protection of Civilian sites in Juba.

There is a paucity of data on violence against women and girls (VAWG) during times of conflict in general and even less information specifically on violence against adolescent girls. Based on secondary analysis of a larger study on VAWG in South Sudan, this article highlights the specific experience of conflict-affected adolescent girls resident in the Juba Protection of Civilian sites. Quantitative data from a cross-sectional household survey shows that the prevalence of non-partner sexual violence (NPSV) (26.5%) and intimate partner violence (IPV) (43.1% of partnered respondents) was high among a cohort of girls who were of adolescent age during the 2013 crisis. Direct exposure to armed conflict increased the odds of respondents experiencing NPSV (AOR: 7.21; 95%CI: 3.94-13.17) and IPV (AOR: 2.37; 95%CI: 1.07-5.29). Quantitative and qualitative data also showed that patriarchal practices, compounded by poverty and unequal power relationships within the home, remain some of the primary drivers of VAWG even in conflict-affected settings. Prevention activities need to consider these wider underlying drivers of VAWG during times of armed conflict, as they remain key factors affecting violence against adolescent girls.

Introduction

South Sudan has experienced almost continuous ethnic and political conflict since gaining independence from Sudan in 2011. In December 2013, large-scale violence broke out in Juba between the government, led by the Sudanese People’s Liberation Movement (SPLM), and the Sudanese People’s Liberation Movement/Army in Opposition (SPLM–IO).[1] While the conflict began as a political dispute, it quickly took on an ethnic component with the SPLM–IO and its leader, former South Sudan vice president, Riek Machar, supported by many of the Nuer population. This initial outbreak of violence caused more than one million people to flee their homes and tens of thousands to seek shelter on United Nations (UN) bases – termed Protection of Civilian (PoC) sites – throughout the country. While a peace agreement was signed in August 2015, conflict re-erupted in July 2016 and spread to previously peaceful areas of the country. This caused more people to flee to the PoC sites. In 2018 there were almost 200,000 people seeking shelter in PoC sites across five states, including almost 40,000 people in Juba.[2]

The ongoing conflict has had devastating impacts on the population and has led to an estimated 383,000 excess deaths from the outbreak of the crisis through April 2018.[3] Women and girls, in particular, have borne the brunt of the consequences of the conflict, including considerable amounts of violence against women and girls (VAWG). While globally it is estimated that one in three women experience violence during their lifetimes, documented rates of violence in conflict-affected areas of South Sudan are almost double the global average.[4,5]

Despite the clear theoretical connections between armed conflict and VAWG, there have been few rigorous research studies that use population-based data to explicitly explore the impact of conflict on rates of VAWG. Meta-analysis of available data on sexual violence has estimated overall prevalence rates that are three times the rates of sexual violence in non-conflict settings, further supporting the association of armed conflict with increased rates of sexual violence.[4,6] Studies that have examined the effect of conflict on other forms of VAWG, such as intimate partner violence (IPV), are more limited. Available data shows that rates of IPV among conflict-affected populations are high.[7-10] However, there is conflicting evidence regarding whether this violence is increased, reduced or remains the same during times of conflict.[11]

While there is a paucity of data on VAWG during times of conflict in general, there is even less information specifically on violence against adolescent girls. The authors are not aware of any population-based data that look at how violence affects this sub-group. However, existing data from impact evaluations in conflict-affected countries suggest that the rates of violence experienced by adolescent girls are very high. For example, more than half of adolescent girls assessed in a baseline survey from sites in the Democratic Republic of Congo and Ethiopia had experienced some form of violence in the year prior to the survey.[12] While there is no specific data on the prevalence of IPV amongst adolescent girls during times of conflict, there is some evidence that rates of early marriage increase during conflict, which may increase girls’ exposure to violence from a husband.[5,13] Given this limited evidence base, this paper seeks to further document the unique experiences of adolescent girls by specifically examining the types of violence affecting this group as well as drivers of this violence.

Methods

Data and study population

This article presents an analysis of cross-sectional quantitative and qualitative data collected as part of the research programme of the What Works to Prevent Violence Against Women and Girls in Conflict and Humanitarian Crises Consortium (“What Works”) in South Sudan, and focuses on data collected in the Juba PoCs. The What Works project is an international multi-disciplinary partnership led by the International Rescue Committee (IRC) with George Washington University’s Global Women’s Institute (GWI) and CARE International UK (CIUK).

The overall household survey was administered to a representative sample of women and girls aged 15–64 in two PoC sites in Juba. A multi-stage cluster sampling design was used to select individual women and girls for inclusion in the cross-sectional survey. The research team first divided each PoC site into roughly equivalent blocks of households. Using a random number selector, final blocks for inclusion in the study were randomly chosen. Within each selected block, a systematic sampling strategy was used to select the final household, with a standard interval of five households applied to select each subsequent house. Finally, within each household, all eligible women and girls were listed and a random number generator was used to select the final respondent. Eligible women and girls were aged 15–64 years old and were resident in the Juba PoCs in June 2016. They had to live usually in the selected household, be visiting the household and sleeping there for at least four weeks or be employed as a domestic servant and sleeping in the household at least five nights a week. Potential respondents were excluded from the study if they were unable to speak one of the languages of survey administration or were seriously ill, mentally incompetent or deaf.

Data collection began in June 2016 but was paused in mid-July due to a new outbreak of violence in Juba. Data collection resumed and was completed in November-December 2016. For this second round of data collection, any potential respondent who had entered the PoC after the July crisis was excluded from the sample. In total, 963 women and girls (aged 15–64) were interviewed, with a response rate of 84%. All interviews were conducted in the local language, primarily Nuer or Juba Arabic. Verbal consent was obtained from all participants prior to data collection.

Qualitative data were collected with community members, key informants (for example, non-governmental organisations staff, service providers, government representatives, local leaders, etc) and survivors of VAWG, utilising semi-structured in-depth interviews and focus group discussions (for community members and key informants). Both men and women were included in these samples including gender-segregated focus groups with men and women from the community. Purposive sampling was used to select all respondents in collaboration with local IRC staff. All participants in qualitative data collection were aged 18 or older. Qualitative data collection was undertaken during three separate periods, initially during the formative research stage in 2015, in parallel to the initial quantitative data collection in June 2016, and then when quantitative data collection resumed in late 2016. The data collection tools were structured to collect data on common types of VAWG, the effect of the conflicts on VAWG, consequences of this violence and access to support services.

To ensure the safety of respondents, the study followed the World Health Organization’s (WHO) ethical and safety recommendations for researching, documenting and monitoring sexual violence in emergencies. This included developing specific procedures to ensure privacy and confidentiality, provision of referral cards for relevant survivor services to all participants, gender-matched and highly trained data collectors and routine monitoring for adverse events.[14] The research protocol was approved by the Institutional Review Board (IRB) of the George Washington University as well as the Technical Advisory Group (TAG) in South Sudan, which is an independent body of experts in VAWG research and programming in South Sudan made up of local, national and international practitioners representing both NGO and government structures.

Measures

The survey was based on the WHO Multi-country Study on Women’s Health and Domestic Violence Against Women and was adapted for the unique context of South Sudan by adding specifics on the multiple conflicts experienced in the country and relevant forms of VAWG and patriarchal practices that emerged during formative data collection.[15] The analysis in this paper focuses primarily on experiences of non-partner sexual violence (NPSV) and physical or sexual intimate partner violence (IPV). A series of questions were asked to assess each construct by first asking if it ever happened in the respondent’s lifetime and then if it happened in the past 12 months. For NPSV, respondents were also asked if the incident occurred since the 2013 crisis began.

For NPSV, respondents were asked if anyone who was not their husband, partner or boyfriend had: (1) forced her to undress or stripped off her clothing; (2) forced her into having sex when she did not want it, for example, by threatening her, holding her down, or putting her in a situation where she could not say no; (3) attempted to force her into sex (which did not take place); or, (4) touched her sexually or did anything else sexually that she did not want to do.

For physical IPV, respondents were asked if their current, or any previous, husband, partner or boyfriend had: (1) slapped her or thrown something at her that could hurt her; (2) pushed or shoved her or pulled her hair; (3) hit her; (4) kicked her or dragged her; (5) choked or burnt her on purpose; or (6) threatened her with or actually used a gun, knife or other weapon against her.

For sexual IPV, respondents were asked if their current or any previous husband, partner or boyfriend had: (1) forced her to have sex when she did not want to; (2) forced her to have sex when she didn’t want to because she was afraid of what her husband/partner might do if she refused; or (3) forced her to do anything else sexual that she did not want or that she found degrading or humiliating.

Exposure to conflict was measured through two separate variables. In one, entitled “other direct conflict experience”, respondents were asked if during any time of conflict they were seriously injured, physically disfigured, or abducted. In a second, the respondent was asked if she ever experienced an attack on her village/community of residence.

Agreement with inequitable gender and violence attitudes was assessed through a series of attitude questions (e.g. agreement that a woman should tolerate violence to keep her family together). These variables were combined to create a scale that measures overall agreement with these attitudes.

Data analysis

This analysis focused on the experiences of girls who were of adolescent age since the 2013 crisis began – aged 15–22. This cohort was chosen to specifically explore the relationships between the most recent armed conflict and VAWG. In total 527 adolescent girls and young women met these criteria and were included in the final analysis. Quantitative data analysis was conducted using STATA Version 15.1. Descriptive statistics as well as bivariate (Pearson chi-square) and multivariate (logistic regression) statistical methods were used for analysis. For the analysis of factors related to experiences of violence (both IPV and NPSV), two multivariate models are presented. The first includes as independent variables a range of socio-demographic and conflict factors, and the second presents the final statistically significant independent variables that are risk factors for VAWG, adjusted for any other factors that were significant in the previous model.

For qualitative data, analysis from the Juba and Juba PoC sites was used to inform this report, as many key informants working in the Juba PoCs also worked in Juba City. Grounded theory was used to examine issues specifically related to adolescent girls. Analysis and coding took place in three stages, after each period of qualitative data collection was completed. The research team looked for saturation in responses at each stage of analysis. Further rounds of qualitative data collection were targeted to fill in any gaps in understanding after the research team coded and debriefed following each wave of data collection. A combination of a priori and inductive coding was used to identify emerging themes and patterns from the transcribed notes. All data were coded using the qualitative software, OpenCode.[16] An initial list of potential codes, focused on types and forms of VAWG and relevant conflicts affecting the study population in South Sudan, was developed prior to the data analysis process. Further codes emerged during the coding process, adding specificity to the experiences and circumstances of VAWG and conflict in South Sudan (e.g. cattle raiding, abduction, revenge killing, etc.). Through this process a codebook with definitions of each code was developed and updated. Two separate researchers initially coded each transcript, noting and discussing each new emerging code. These transcripts were then compared and consolidated for consistency. Using these codes, analytical categories were developed and explored, allowing researchers to capture key features of the different forms of violence affecting adolescent girls and to identify patterns and typologies in the data. The primary analytical categories utilised were: types of VAWG, inequitable gender norms, exposure to conflict and displacement, perpetrators of VAWG, consequences of violence, and service access. Sub-codes were organised under each of these categories and data were reviewed under each to identify common themes and trends.

Socio-demographics of the respondents

Overall, respondents in the survey were primarily from the Nuer tribe and reported that they were literate and had some education, though less than half had reached secondary school. About half of respondents reported that they were students at the time of data collection. More than half said that their household had no income or relied on humanitarian aid. About 40% of respondents reported that they had a partner or were married. These partners were relatively well educated (about half had some university education), however almost half were not working. There were considerable age differentials with 40% of men more than 10 years older than their wives. All respondents had been displaced by the 2013 crisis as they were all resident in the Juba PoCs at the time of the interview. In addition, half had experienced a direct attack on their village or community of residence during the conflict and almost 18% had some other direct conflict experience. See Table 1 for details.

Table 1.

Description of the sample: cohort of girls of adolescent age during the 2013 crisis (n = 527)

	%
Ethnicity
Not Nuer	9.4
Nuer	90.6
Religion
Catholic	46.9
Protestant	46.1
Other religion	7.1
Literacy
Not literate	21.8
Literate	78.2
Education
No education	19.7
Primary education	44.7
Secondary education+	35.6
Work status
Not working	46.2
Domestic work	3.6
Student	45.9
Other work	4.3
Income status
No income	54.1
From husband	9.0
From relative	34.0
Other income status	2.9
Fuel source
Wood, grass, leaves	24.3
Charcoal	75.7
Marital status
Never married	57.4
Nonmarital partner	13.9
Married	28.7
Village ever attacked
No	49.5
Yes	50.5
Other direct conflict experience
No	82.5
Yes	17.5
Total	100.0

Findings

Both quantitative and qualitative data highlighted the extremely high rates of NPSV amongst girls who were of adolescent age during the 2013 crisis. Overall, 26.5% of all respondents reported that they had experienced some form of NPSV at least once during their lifetimes and the largest proportion of respondents had experienced the most serious form of NPSV – rape or an attempted rape (21.9%). In addition, almost as many respondents (20%) reported being touched sexually when they didn’t want it and 12.4% were forced to undress or strip (Table 2)

Table 2.

Prevalence of non-partner sexual violence amongst girls of adolescent age during the 2013 crisis (n = 527)

	%	95% C.I.
Ever experienced
Ever experienced any sexualNPSV	26.5	(21.0–32.9)
Ever forced to undress or strip	12.4	(8.5–17.9)
Ever experienced rape orattempted rape	21.9	(16.9–27.7)
Ever touched sexually when shedidn’t want it	20.0	(15.5–25.3)
Experienced since the 2013 crisis began
Experienced any sexual NPSVsince the 2013 crisis began	14.1	(10.0–19.5)

The data from this analysis also shows that the ongoing conflict in South Sudan has had a considerable impact on incidents of NPSV. Overall, 14.1% of respondents experienced some form of NPSV since the crisis began in 2013. This is a majority (53%) of all reported incidents, showing that most incidents of NPSV experienced have occurred since the beginning of the 2013 crisis.

Qualitative data provides further details on the experiences of NPSV from the perspectives of women and girls. In particular, a finding that emerged during analysis was that heightened periods of sexual violence occurred during acute periods of armed conflict in Juba. In addition, women and girls also highlighted the ongoing risk of sexual violence when they left the relative safety of the PoCs to collect firewood, go to the market, etc. This relationship between violence and conflict was also seen in multivariate analysis. Respondents who had a direct conflict-related experience (were seriously injured, physically disfigured, or abducted) were significantly more likely to have experienced NPSV since the 2013 crisis began (AOR: 7.21; 95%CI: 3.94–13.17). Age was also a significant risk factor for violence with older respondents (within the 15–22 year-old cohort) with increased odds of experiencing an incident of NPSV since the 2013 crisis began (AOR: 1.15; 95%CI: 1.02–1.30) (Table 3).

Table 3.

Drivers of non-partner sexual violence among girls of adolescent age who experienced NPSV during the 2013 Crisis (n = 527)a

	%	Bivariate	Multivariate 1b	Multivariate 2c
		OR (95% CI)	AOR (95% CI)	AOR (95% CI)
Age		1.20**(1.05–1.35)	1.17*(1.02–1.33)	1.15**(1.02–1.30)
Direct conflict exposure
No	8.4	1	1	1
Yes	40.8	7.48***(4.06–13.79)	7.45***(4.21–13.19)	7.21***(3.94–13.17)

aIn this table age was measured as a continuous variable.

bAdjusted for socio-demographic factors (religion, literacy, education, employment, income source, fuel source, marital status) and conflict exposure (village ever attacked).

cAdjusted for socio-demographic factors that were significant in previous models (ethnicity).

*P ≤ .05; **P ≤ .01; ***P ≤ .001.

“It happens several times during the conflict; people go and attack a certain place or town and after fighting they come and rape women and ladies.” (Women in the Juba PoCs)

“This is happening during the conflict especially by soldiers. They rape girls and women who are innocent. If you refuse, they can kill you.” (Key Informant in Juba)

“If we go outside the PoC to get food, or if you go outside to get firewood, a military soldier may rape me, or they may be physically violent. That is happening to a lot of people.” (Women in the Juba PoCs)

For respondents who experienced NPSV since the 2013 crisis, the most common male perpetrators were police or members of another community/tribe (who typically were involved in fighting around the 2013 crisis or other inter-communal conflicts between tribes). Other armed actors were also reported as perpetrators of NPSV, showing the continued effect of the conflict on incidents of NPSV. Finally, male family members, friends, neighbours and classmates were also common perpetrators of NPSV during the conflict-period (Table 4).

Table 4.

Most commonly reported male perpetrators of NPSV since the 2013 crisis began (n = 74)

Perpetrators	%
Male police	24.2
Male member of other community/tribe	23.6
Father/stepfather	6.7
Male armed actor	6.2
Other male family member	6.1
Male friend, neighbour, or classmate	5.7

Consequences of sexual violence for girls’ resident in the PoC sites were explored through qualitative data analysis. A primary finding that emerged through this process was that girls who were raped were typically forced by their families to marry the perpetrator of this violence. Girls who do not marry their rapists are considered “spoilt” by the community due to the loss of their virginity, which was thought by participants to have long-term consequences for girls by affecting their prospects for marriage.

“If she doesn’t marry the rapist, she can’t marry anyone else. Then her family doesn’t get a dowry.” (Key Informant Juba PoCs)

“In the case that she is forced to marry the man, she will not be happy in her married life because there is no love between them.” (Women in the Juba PoCs)

“It’s like you lose some of your value … it comes down to the issue of virginity.” (Key Informant Juba)

Intimate partner violence

Rates of IPV were also high for girls who were of adolescent age during the 2013 crisis. As can be seen in Table 5, 43.1% of partnered respondents who were adolescents during the 2013 crisis had experienced physical or sexual IPV during their lifetime. Among this population, sexual violence rates (35.2%) were higher than physical violence rates (30.4%).

Table 5.

Lifetime experience of physical or sexual IPV perpetrated against girls who were adolescent age during the 2013 crisis (N = 283)

	%	95% C.I.
Physical and/or sexual IPV	43.1	(35.9–50.7)
Any physical IPV	30.4	(25.0–36.3)
Any sexual IPV	35.2	(27.7–43.5)

Multivariate analysis was also conducted to explore the risk factors associated with IPV perpetrated against girls of adolescent age during the 2013 crisis. Overall, the primary risk factors that came out in this analysis were related to gender-inequitable norms and power differentials, poverty and exposure to conflict. For example, respondents who agreed with more gender-inequitable attitudes had an almost four-fold increase in the odds of having experienced IPV (AOR: 3.89; 95%CI: 1.17–12.97). In addition, respondents whose husbands demonstrated more controlling behaviours in their relationship had an increased likelihood of experiencing IPV (AOR: 1.40; 95%CI: 1.20–1.64). Respondents who used more expensive fuel sources, which can be a proxy for wealth in this context, reported experiencing less IPV (AOR: .49; 95%CI: .24–.98) compared to those who used cheaper/free sources. Finally, respondents whose village or community was attacked during the crisis had more than double the odds of experiencing IPV in their lifetime (AOR: 2.37; 95%CI: 1.07–5.29) (Table 6).

Table 6.

Drivers of lifetime experience of IPV, perpetrated against girls who were adolescent age during the 2013 crisis (n = 283)a

	%	Bivariate	Multivariate 1b	Multivariate 2
		OR (95% CI)	AOR (95% CI)	AOR (95% CI)
Literacy
Not literate	41.5	1	1	1
Literate	43.8	1.10(0.59–2.04)	5.37**(1.61–17.94)	3.65(0.98–13.62)
Highest level of education
No education	44.2	1	1	1
Primary school	42.6	0.94(0.47–1.86)	0.21*(0.05–0.83)	0.25(0.06–1.06)
Secondary or higher	42.9	0.95(0.44–2.04)	0.23(0.04–1.18)	0.32(0.07–1.46)
Fuel source
Wood, leaves,grass	61.0	1	1	1
Charcoal	37.4	0.38**(0.19–0.77)	0.33**(0.14–0.76)	0.49*(0.24–0.98)
Marital status
Non-maritalpartner	53.8	1	1	1
Married	43.7	1.02(0.56–1.88)	0.28*(0.09–0.91)	0.48(0.21–1.10)
Number of centrolling behaviours experienced
	–	1.34***(1.15–1.58)	1.49***(1.27–1.75)	1.40***(1.20–1.64)
Agreement with inequitable gender and violence attitudes
	–	2.53(0.90–7.11)	6.02*(1.55–23.43)	3.89*(1.17–12.97)
Village ever attacked
No	26.7	1	1	1
Yes	56.6	3.57***(1.82–7.00)	2.79*(1.03–7.52)	2.37*(1.07–5.29)

aIn this table, number of controlling behaviours experienced and agreement with inequitable gender and violence attitudes were measured as continuous variables.

bAdjusting for socio-demographic factors (religion, literacy, education, employment, income source, fuel source, marital status) and conflict exposure.

*P ≤ .05; **P ≤ .01; ***P ≤ .001.

Conflict and displacement also emerged in qualitative data analysis as drivers affecting the rates and severity of IPV for women and girls resident in the PoCs. Women and girls described this increase in the severity of violence as a by-product of the increasing brutality of the armed conflict, which had escalated during the 2013 crisis.

“If a woman beats her husband, he can either break her arm, remove her teeth, injure or even kill his wife.” (Women in the Juba PoC)

“Women are mostly affected with domestic violence and rape. Domestic violence mostly affects women because they are inferior in strength to men and at all times beaten by their husbands and raped.” (Men in the Juba PoCs)

Discussion

Rates of violence are extremely high amongst women and girls in South Sudan. Even while examining a small proportion of the overall female population (15–22-year-olds), over 26% had already experienced an incident of non-partner sexual violence. This statistic is even higher than global estimates of sexual violence rates in complex emergencies, which have been estimated at 21% of women and girls of any age.[6] In addition, almost half of the incidents of NPSV reported occurred before 2013, suggesting that not only is sexual violence rampant in South Sudan but it is beginning very early in the lives of girls. While the exact reasons for these high rates of VAWG in South Sudan are not entirely clear, the qualitative and quantitative data in this study suggest that the embedded patriarchal norms and practices prevalent throughout South Sudan are compounded and re-enforced by protracted armed conflict which leads to heighted rates of VAWG.

The 2013 crisis in South Sudan was notable for its brutality and the use of sexual violence as a tactic of war. This is reflected in the data produced in this study, which show that a majority of the respondents who had experienced an incident of sexual violence reported that this incident occurred after the 2013 crisis began. Qualitative data further echo this finding and suggest that sexual violence was particularly prevalent during the acute phases of fighting (notably December 2013 and July 2016) in Juba. However, adolescents and young women also reported ongoing threats of sexual violence even after the acute conflict passed. This included sexual violence perpetrated by soldiers or police officers when young women leave the relative security of the PoC sites to collect firewood or to go to the market. In addition, perpetrators of NPSV were also commonly reported to be non-conflict actors, such as family members, friends, neighbours and classmates. This demonstrates the continued vulnerability of women and girls to sexual violence even when they are residing in the PoC sites and moving about their daily lives. While acute conflict may be a significant driver of this sexual violence during conflict, ongoing risk mitigation and violence prevention strategies need to continue to be prioritised in humanitarian settings even once the acute phase of the emergency has passed.

High rates of sexual violence create a narrative that girls need “protection” which, in the context of South Sudan, often manifests as controlling behaviours and gender inequitable practices. For example, in order to “protect” girls and young women from rape, girls may be kept home from school, not allowed to leave the house and forced into early marriage. As such, conflict-related sexual violence also compounds pre-existing patriarchal norms and practices. This analysis demonstrates the importance of programming to address conflict-related sexual violence that acknowledges and addresses the power imbalances that underlie this violence, even during times of conflict, where even the threat of sexual violence can have considerable consequences on the lives of adolescent girls and young women.

Conflict-related violence often specifically focuses on sexual violence. However, this research demonstrates that other forms of violence also may be impacted by exposure to armed conflict and displacement. Specific data were not collected on the prevalence of IPV since 2013. However, the median age of marriage for respondents in the Juba PoC was 16 years old and we therefore expect that the vast majority of married respondents in our age group would have been married since the outbreak of the 2013 crisis. Due to this we used variables measuring any experience of physical or sexual IPV as a proxy indicator of violence that occurred since the 2013 crisis. Using this measure, almost half of the respondents in this sample had already experienced physical or sexual IPV. Unusually, reported rates of sexual violence were higher than physical violence, perhaps reflecting the patriarchal view that a wife is expected to submit to sex whenever mandated by her husband, which was also seen in the qualitative data.

For adolescent girls who reside in the PoC sites, exposure to conflict was also a major risk factor for IPV in multivariate analysis. This study is one of the first to demonstrate an empirical link between exposure to conflict and an increase in IPV among adolescent girls. It points to the need for programmers and policymakers to consider IPV as a form of violence affecting adolescent girls during times of conflict and to ensure programming specifically targeting this age group is incorporated into wider VAWG programming strategies in these settings.

In addition to conflict, agreement with inequitable gender attitudes was also shown to be a considerable risk factor for adolescents experiencing IPV. This finding demonstrates how adolescents are internalising wider patriarchal norms and how this may affect their odds of experiencing IPV. Importantly, these findings show that the drivers of violence against adolescent girls in a conflict-affected country are not completely divorced from drivers of violence in other, more stable settings. Unequal and patriarchal practices, which are compounded by poverty and unequal power relationships within the home, remain some of the primary drivers of violence in conflict-affected settings. In order to truly impact rates of VAWG in conflict settings, it is important that programmers and policymakers address these wider drivers of violence as well as conflict-specific factors.

There were several limitations to the study. For one, the research focused on secondary analysis of a larger data set collected for the What Works programme. As such, the sample size of girls of adolescent age since the 2013 crisis began was relatively small, making it difficult to draw firm conclusions about the experiences of adolescent girls overall in this setting. In addition, while we were specifically interested in the impact of armed conflict on adolescent girls, our measures were not precise enough to specifically draw out this relationship. For example, we did not have data specifically on intimate partner violence since the 2013 crisis began and as such had to use lifetime experience of violence for adolescent girls and young women as a proxy. Finally, all our measures were based on self-reported data, which could have led to under-reporting of violence, particularly amongst this young age group where stigma may be high.

Conclusions

Adolescent girls in South Sudan are experiencing high rates of gendered violence. Programmers and policy-makers need to ensure that the unique needs and experiences of this age group are considered when implementing both child protection and VAWG programmes in conflict and humanitarian settings. Prevention activities also need to consider the wider underlying causes of VAWG – patriarchal practices and norms – even during times of armed conflict, as they remain key factors driving violence against adolescent girls.