Measuring Physical Violence and Rape Against Somali Women Using the Neighborhood Method

Abstract

This study was conducted to estimate prevalence of gender-based violence (GBV) among female Somalis in Ethiopian refugee camps and host communities, compare prevalence in camps and communities, and compare prevalence in flight and in camp. Systematic random sampling was used to select households in Awbare camp (n = 85), Awbare town (n = 76), and Kebribeyah camp (n = 83). GBV was common and overwhelmingly domestic. Prevalence was higher in Awbare town than Kebribeyah camp. Women were at increased risk of GBV in camp compared with in flight. The domestic nature of GBV in humanitarian settings requires attention. Assumptions about violence in humanitarian settings should be further tested.

Keywords

physical violence sexual violence refugee

Introduction

Gender-based violence (GBV) is a critical and preventable cause of injury and death throughout the world (Campbell, 2002; Ellsberg, Jansen, Heise, Watts, & Garcia-Moreno, 2008; Garcia-Moreno & Stockl, 2013). A systematic review estimated that 38.6% of female homicides throughout the world were committed by an intimate partner (Stockl et al., 2013). Physical effects of GBV include traumatic injuries, chronic illness, and death. One review, for example, found prevalence of traumatic brain injury among survivors of intimate partner violence (IPV) to be between 30-74% (Sheridan & Nash, 2007). Other studies have shown GBV survivors to be at increased risk of chronic pain, gastrointestinal disorders, fibromyalgia, and eyesight and hearing problems (Garcia-Moreno & Stockl, 2013; Krug, Dhalberg, Mercy, Zwi, & Lonzono, 2002).

Experiencing GBV has also been shown to compromise women’s mental health. Studies have found that between 17-65% of women who experience sexual assault demonstrate symptoms of posttraumatic stress disorder (PTSD; Campbell, Dworkin, & Cabral, 2009). Similarly, a nationally representative study of women in South Africa found that, compared with women who had not experienced violence, those who had been raped had six times the odds of meeting diagnostic criteria for PTSD (Kaminer, Grimsrud, Myer, Stein, & Williams, 2008). GBV has also been associated with increased risk of depression, substance use disorders, and suicidal ideation and attempts (Ellsberg et al., 2008).

Sexual violence against women and girls during armed conflict has been a strategy of war throughout history (Card, 1996; Stark & Wessells, 2012). Conflicts in Bosnia, Rwanda, and the Democratic Republic of Congo (DRC), among others, provide stark evidence that systematic strategies of sexual violence during conflict persist. In the DRC, for example, it has been estimated that more than 400,000 women were raped in the 12 months prior to participating in the 2007 DRC Demographic and Health Survey and between 1.69 and 1.80 million women in the DRC have ever been raped (Peterman, Palermo, & Brendenkamp, 2011).

Over the past decade, there has been increased attention to the magnitude and consequences of GBV during conflict and war. However, data on the prevalence and nature of such violence remain limited, particularly at a population level, due to factors including, but not limited to, stigma, fear of retribution, and underreporting (Hynes, Robertson, Ward, & Crouse, 2004; Stark & Ager, 2011). A systematic review of prevalence studies of GBV in complex emergencies identified just 10 studies that attempted to measure GBV using population-based methods in humanitarian settings (Stark & Ager, 2011).

Despite the relatively limited empirical knowledge on GBV in complex emergencies, certain assumptions persist. For example, it is often assumed that women and girls are at greatest risk of sexual violence during the acute phase of conflict when refugees are in flight from their home to a host country, and that the risk of sexual violence falls and remains lower once women and girls arrive at a refugee camp.

In addition, it is often assumed that refugees have experienced higher rates of sexual violence than members of host communities that take in refugees. This belief may be informed by assumptions regarding refugees’ risk of sexual violence in flight. To date, no study has been identified that compares the prevalence of sexual violence among refugees with the prevalence of violence among members of the host community.

Data are also lacking on whether the length of time spent in a refugee camp is associated with an increased or decreased prevalence of sexual violence. Refugee camps are intended to be temporary situations. However, the majority of refugees worldwide currently find themselves in protracted refugee situations (Loescher & Milner, 2005). Kebribeyah camp in Ethiopia, for example, has been populated by Somali refugees for more than 20 years as warring parties vie for control of Somalia. According to the United Nations High Commission for Refugees (UNHCR), more than 900,000 Somalis have become refugees as a result of conflict and instability (UNHCR, 2015). Understanding whether the prevalence of GBV differs between recent arrivals and long-term camp residents can inform prevention and treatment services.

It is in this context that the International Rescue Committee (IRC) and Columbia University’s Program on Forced Migration and Health sought to measure the magnitude of violence facing women and girls in the Somali refugee camps of Kebribeyah and Awbare, both located in Ethiopia, as well as the surrounding host community of Awbare. This population-based survey was conducted to quantify the magnitude of violence perpetrated against females in two Somali refugee camps and one surrounding Ethiopian host community to begin to build an evidence base to better understand the experiences of conflict-affected women.

Specifically, we set out to (a) estimate the prevalence of physical violence and rape among female Somali refugees and female members of the neighboring host communities, (b) compare the reported prevalence of physical violence and rape perpetrated against females living in the refugee camps with the reported prevalence of physical violence and rape perpetrated against females living in the host communities, (c) compare reported prevalence of physical violence and rape in flight with reported rates of physical violence and rape in refugee camps, and (d) compare prevalence of physical violence and rape among those who recently arrived at a refugee camp with those who had spent an extended period of time in a refugee camp.

Method

The research team used the “Neighborhood Method,” a methodology developed by Columbia University and previously piloted in collaboration with Childfund in Uganda in 2006 (Stark et al., 2009) and with the IRC in Liberia in 2007 (Stark, Warner, Lehman, Boothby, & Ager, 2013). Study protocols, sampling plans, and interview guides were submitted and approved by Columbia University’s official institutional review board (IRB). Once in country, protocols and interview guides were again reviewed with a group of local experts who served as an in-country IRB check.

The Neighborhood Method is based on the principle that an informant can provide valid and reliable information not only on her experiences and those of her household but also those of others in close proximity—her “neighborhood.” This principle promises a methodology of power and efficiency in humanitarian contexts where timely and efficient data collection is often ethically and logistically necessary.

With this methodology, local interviewers conduct in-depth, semi-structured interviews with respondents. Interviewers ask respondents not only about their own experiences but also about the experiences of all females in their household as well as the experiences of all females in the households of their four closest neighbors. Thus, the Neighborhood Method systematically captures information on a much larger number of females than would be possible if respondents were asked only about their own experiences.

Interviewers were educated female Somali women from either the refugee camps or the surrounding host community who underwent extensive training on concepts of GBV and by the research team on implementation of the interview protocol. Interviewer training involved structured review of the components of the protocol, role-play, and pilot interviewing. Training focused on consistency in the application of the interview protocol and coding and addressed issues of sensitivity in probing on the topics of physical violence and rape. Training also focused on refining case definitions with respect to relevant terms. Field testing on the outer edge of Kebribeyah town followed the training. Once training and field testing were completed, data collection began in the refugee camps of Kebribeyah and Awbare, and in the town of Awbare.

Interviews were semi-structured and included both open- and closed-ended interview questions. The flow of the interviews was conversational, encouraging each respondent to discuss issues of physical violence and rape from her own perspective and emphasize what she deemed to be important. At the same time, specific questions about the experiences of the populations of interest were also interwoven into the discussion.

Sample

Systematic random sampling was used to select households in each of the three survey areas. To determine the minimum number of interviews necessary in each survey area, a number of factors were considered. After discussion with local staff, an average of three females per household was assumed. Based on results from similar previous surveys (Stark et al., 2009; Stark et al., 2013), a conservative GBV incidence of 10% over the recall period was assumed. In addition, a confidence level of 95% and a design effect of 2 were incorporated into the analysis. Based on these factors as well as the size of the population, using EpiInfo, the minimum number of interviews necessary in each survey area was 73. This desired sample size was increased to account for assumed refusals to participate.

Initial population figures were provided by IRC for the town of Awbare and by UNHCR for the refugee camps in Kebribeyah and Awbare. Because current maps were not available for Awbare town, the research team mapped the area, roughly estimating the number of households they observed to confirm population estimates and create a sampling plan. The estimated number of households was divided by the desired sample size to obtain the sampling interval for both Awbare town and the long-established Kebribeyah camp. In Awbare camp, with the collaboration of the Administration for Refugee and Returnee Affairs (ARRA) of the Ethiopian government, administrative blocks were identified within the camp. Each block consisted of 16 households. The interviewers sampled one house from each block. A random number sequence was used within the sampling interval to select the starting household in Awbare town and Kebribeyah camp and to select the household to be interviewed within each block in Awbare camp.

At each selected house, the interviewer asked to speak to the female head of the household. The interviewer would then invite the female head of the household, as identified by the household, to participate in the study. If a woman above age 18 was not available or not interested in participating, the interviewer went to the house next to this one. If there were more than three unavailable households/refusals in a row, the interviewer would begin the counting interval over again at the next house.

Interview

Once the female head of the household was identified, interviewers explained the purpose of the survey and obtained verbal informed consent. Interviewers then worked with respondents to identify a private location in which to conduct the interview. Most interviews occurred either within the respondent’s home when privacy was possible, or in a more secluded space a short distance from the respondent’s home.

The interviewer began by collecting basic demographic information including age and marital status of the respondent, the other females in her household, her four closest neighbors, and all the females in those households. For the purposes of this study, girls were defined as females below the age of 18 and women were defined as females 18 years of age or older. The respondent was invited to share some of the biggest challenges facing women and girls in her community. Following and often building on themes that emerged during these open-ended discussions, the interviewer would question the respondent about her four closest neighbors’ experiences with two categories of violence: physical violence and rape. Physical violence was defined as any act of physical aggression including pushing, shoving, pulling, shaking, slapping, biting, hitting, punching, kicking, strangling, and throwing objects. Rape was defined as the intentional penetration of another person’s vagina or anus without the survivor’s consent.

First, the interviewer asked whether the first neighbor or any females living in that household had experienced physical violence during the recall period. If an incident had occurred, the interviewer probed for additional information about the perpetrator (i.e., relationship to survivor), whether or not the survivor had reported the incident, if so, to whom, and where the incident occurred. This information, including reporting information, was asked of all survivors about whom data were collected, including neighbors, females in the respondents’ household, and respondents. Interviewers recorded additional details in narrative form. Once the interviewer obtained all the necessary information about the first incident, she would then ask whether that first neighbor experienced physical violence from anyone besides that first perpetrator. If so, the interviewer would obtain the same information about this second incident. This process was then repeated, asking about any physical violence experienced by three other neighbors, all the females in those three households, the respondent, and all the females in the respondent’s household. After this was completed, the interviewer would repeat the same process asking about experiences with rape among all the females in the households of her four closest neighbors and among all the females in the respondent’s household, including the respondent.

Recall

The time frame covered in the study was December 2006 through the day of the interview (approximately 18 months). Several memorable events had occurred in the same time span, 18 months before the survey began, and were used to support bounded recall. The first event used to aid recall was “two Arafa’s ago.” Arafa is a Muslim holiday, and at the time of the study, “two Arafa’s ago” corresponded with December 2006. That particular Arafa was memorable because it fell within a week of Mogadishu’s fall to Ethiopian troops, the bombing of the airport in Mogadishu by Ethiopian forces, and the execution of Saddam Hussein. All of these points were used in various interviews to aid recall, though Saddam Hussein’s death turned out to be the most salient event and was used most frequently.

Recall Period in Awbare Camp

The neighborhood methodology relies on a respondent’s ability to accurately report on the experiences of her neighbors. In Awbare town and Kebribeyah camp, the research team utilized an 18-month recall period when interviewing respondents about their experiences and the experiences of their neighbors. Because individuals had relocated either to Awbare town or Kebribeyah camp more than 18 months before the interview, it was deemed reasonable to ask them to report on the experiences of their families and their neighbors over the last 18 months. At the time of this survey, however, Awbare camp had been open for 12 months, and respondents had relocated to the camp anywhere between 1-12 months prior. As such, the research team adjusted their recall period when surveying Awbare camp. In this setting, the research team asked respondents to report on the experiences of their neighbors since the neighbors’ household had moved to the camp. However, when the interviewer asked the respondent about her own experiences and the experiences of females in her own household, she would ask about the last 18 months. In Awbare camp, the interviewer asked the respondent how many months she and each of the female members of her household had been living in Awbare camp. For each incident of violence or rape perpetrated against the respondent or females in the respondent’s household, the interviewer would clarify whether the incident occurred at Awbare camp or before the person arrived at camp. Incidents that occurred prior to respondents or members of the respondents’ households arriving at camp were categorized as occurring in flight from Somalia to the refugee camp. Rates of violence and rape (incidents per person-month) were constructed for time in flight from Somalia to Awbare camp and for time since arrival at Awbare camp. This also allowed the researchers to compare rates of violence for women in flight from Somalia with rates of violence once they had settled in the camp.

Ethics

This study is one of a small number of studies that attempted to systematically measure the prevalence of GBV in humanitarian settings. Such research, however, presents unique ethical challenges and risks. In light of these risks, the World Health Organization (WHO) published ethical and safety recommendations for those conducting research on IPV (WHO, 1999). The research team reviewed the WHO recommendations and worked to address each recommendation with care and deliberation. Informed consent was obtained from all participants in the study. The research teams were careful to make clear to potential respondents that involvement was completely voluntary and that there would be no consequences, positive or negative, based on whether or not someone agreed to participate. The importance of informed consent was stressed to the research team during training and emphasized during role-plays and pilot testing. In addition, the study team ensured that research was only conducted in regions where services were available and accessible. In advance of the study, the research team mapped available services near the research area and compiled a list of relevant referral services. Special psychosocial point people were on call in the research settings in the event that serious allegations of abuse or violence were discovered and additional support required. These individuals were available both during and after the research study. Interview sites were carefully chosen to draw the least possible attention to the interviewees. Extensive training was provided to the research team on how to interview women about sensitive subjects, when to stop an interview, and what to do if an interview had to be stopped prematurely.

Results

Eighty-three interviews were conducted in Kebribeyah refugee camp, 85 in Awbare camp, and 76 in Awbare town (see Table 1). The participation rate was 83% in Kebribeyah camp (16 empty houses and zero refusals), 81% in Awbare camp (19 empty houses and one refusal), and 70% in Awbare town (24 empty houses and nine refusals). The mean age of females in the sample was 22 years. The majority of women were married.

Table 1.

Profile of Sample Population.

	Awbare camp	Awbare town	Kebribeyah camp
Sample size	1,184	1,102	1,264
Respondents	85	76	83
Others in respondents’ households	212	165	220
Neighbors and others in neighbors’ households	887	861	961
Average age	22.2	22.2	21.6
% less than 18	51.5	46.9	45.2
% 18 and older	46.3	49.0	49.8
Age missing	2.2%	4.1%	5.0%
Marital status (18 or older)
Single	18.0%	25.0%	25.1%
Married	56.3%	53.1%	54.6%
Divorced	10.9%	10.7%	7.0%
Widowed	14.8%	11.2%	13.2%

Physical Violence

Experiences of physical violence were common in all sample populations (see Table 2). Taken together, physical violence was reported by at least 39% of women and girls in each population. Experiences of physical violence were more common for women as compared with girls. Across survey areas, women were more likely to have experienced physical violence as compared with girls.

Table 2.

Incidence Rates December 2006-June 2008.

	Awbare camp^a (n = 297)	Awbare town (n = 1,102)	Kebribeyah camp (n = 1,264)	p value^b
Physical violence	48.1%, 95% CI [42.2, 53.5]	43.5%, 95% CI [40.0, 45.9]	39.4%, 95% CI [36.6, 41.2]	.0433
Below 18	33.9%	28.8%	24.6%	.0210
18 and older	65.9%	57.5%	52.7%	.0192
Rape	18.8%, 95% CI [13.0, 23.0]	25.0%, 95% CI [20.5, 25.5]	19.6%, 95% CI [17.8, 22.2]	.0016
Below 18	2.4%	7.1%	3.3%	<.0001
18 and older	39.5%	42.3%	34.5%	.0001

Note. CI = confidence interval.

Data for respondents and members of respondents’ households only.

p values derived from t tests comparing the sample from Awbare town with the sample from Kebribeyah camp.

Experiences of physical violence were overwhelmingly domestic in nature, with the majority of violence occurring within the home and perpetrated by an intimate partner or other family member (see Table 3). Across sample populations, two thirds of physical violence was perpetrated by an intimate partner or other family member. As seen in Table 4, approximately 75% of violence was reported to have occurred in the home, with just 5% of violence having been reported to occur while women and girls were in the forest or collecting firewood.

Table 3.

Perpetrators of Physical Violence and Rape.

Perpetrators	Physical violence against girls			Physical violence against women			Rape of girls			Rape of women
Perpetrators	Awbare camp (n = 54)	Awbare town (n = 155)	Kebribeyah camp (n = 147)	Awbare camp (n = 322)	Awbare town (n = 120)	Kebribeyah camp (n = 348)	Awbare camp (n = 5)	Awbare town (n = 38)	Kebribeyah camp (n = 147)	Awbare camp (n = 120)	Awbare town (n = 322)	Kebribeyah camp (n = 348)
Husband/partner	1.9%	3.9%	0.7%	63.7%	67.5%	56.0%	0%	15.8%	10.0%	76.2%	71.2%	72.9%
Other family member	46.3%	63.9%	49.7%	18.3%	14.2%	16.1%	40.0%	13.2%	5.0%	3.9%	5.3%	3.6%
Neighbor/community member	7.4%	8.4%	14.3%	7.8%	6.7%	14.4%	0%	13.2%	35.0%	4.3%	5.7%	7.6%
Child/student	31.6%	8.4%	23.1%	2.2%	1.7%	2.3%	0%	0%	5.0%	0%	0%	0%
Stranger	5.6%	11.0%	8.8%	7.8%	10.0%	8.9%	60.0%	52.6%	45.0%	13.4%	15.7%	13.9%
Teacher	7.4%	3.9%	3.4%	0%	0%	2.0%	0%	0%	0%	0%	0%	0%
Other	0%	0%	0%	0%	0%	0%	0%	5.3%	0%	2.2%	2.2%	2.0%

Table 4.

Location of Violence.

	Physical violence			Rape
	Awbare camp	Awbare town	Kebribeyah camp	Awbare camp	Awbare town	Kebribeyah camp
Home	74%	81%	74%	86%	84%	85%
Street/market	10%	9%	13%	6%	8%	9%
School	7%	5%	7%	0%	0%	0%
Forest/collecting firewood	2%	4%	4%	3%	7%	5%
Other	4%	0.6%	0.6%	6%	1%	1%
Collecting water	3%	0%	2%	0%	0%	0%

As seen in Table 5, disclosure of physical violence was quite common. Between 98-99% of those who experienced physical violence told someone about the incident. Informal disclosure to family and friends was far more common than disclosure to more formal sources (e.g., police, hospital). Among those who experienced physical violence and told someone about the incident, more than two thirds told a friend or neighbor whereas fewer than 10% reported the incident to the police, court, hospital, international organizations, or government representatives.

Table 5.

Individuals or Organizations to Whom Violence Was Reported.

	Physical violence			Rape
	Awbare camp	Awbare town	Kebribeyah camp	Awbare camp	Awbare town	Kebribeyah camp
Family member	34.7%	55.3%	45.1%	51.5%	55.5%	49.5%
Friend/neighbor	77.5%	69.1%	70.8%	33.7%	25.8%	37.1%
Police	5.6%	6.8%	6.8%	14.9%	32.0%	22.7%
Court	1.2%	0%	0.9%	0%	0%	0%
Hospital	2.5%	4.3%	3.1%	7.9%	28.1%	18.9%
UNHCR	0.2%	0%	1.1%	3.0%	0%	6.1%
Teacher	3.1%	2.0%	3.1%	0%	0%	0%
ARRA	2.5%	0%	4.4%	13.9%	0%	5.3%

Note. UNHCR = United Nations High Commission for Refugees; ARRA = Administration for Refugee and Returnee Affairs.

Rape

Experiences of rape were also reported to be common among women across survey areas and, similar to physical violence, were more common among women as compared with girls. Taken together, rape was reported by at least 18% of women and girls in each of the sample populations.

As with physical violence, experiences of rape were overwhelmingly domestic in nature, with the majority of rape reportedly having occurred in the home, perpetrated by husbands and other intimate partners. In all sample populations, more than 70% of rape of women was reportedly perpetrated by husbands or other intimate partners, and approximately 85% of all incidents of rape occurred at home.

Disclosure of rape was relatively common, although less common than disclosure of physical violence. Between 44-52% of those who experienced rape reportedly told someone about the incident. Of those who had been raped and told someone about the incident, approximately half told a family member. Between 15-32% of those who had been raped and told someone reported the incident to the police.

Rates of Violence in Refugee Camps as Compared With Host Community

The assumption of higher rates in the camp as compared to the host community was not supported by the data. The p values derived from t tests comparing the sample from Awbare town with the sample from Kebribeyah camp indicate that rates of violence in Awbare town were significantly higher than those in Kebribeyah camp. Females in Awbare town had 1.10 times the odds of experiencing physical violence compared with females in Kebribeyah camp (p = .0433). These same females had 1.25 times the odds of experiencing a rape event compared with females in the camps (p = .0016). Adult women in town were especially vulnerable, with 2.15 times the odds of experiencing rape compared with women in the camp (p < .001).

Rates of Violence in Newer Camp as Compared With Older Camp

This study also allowed for a comparison of rates of GBV in the 1-year old camp, Awbare, with rates of GBV against females in the 18-year old camp, Kebribeyah. Because different recall periods were used in these two settings, formal t tests were not performed. However, visual examination of the findings in Table 2 suggest no significant difference in rates of rape reported by those who were living in the newer camp of Awbare as compared with those living in the older camp of Kebribeyah.

The rate of physical violence in the more recently settled Awbare camp (48.1%, 95% CI [42.2, 53.5]), however, appears somewhat higher than the rate of violence in Kebribeyah camp (39.4%, 95% CI [36.6, 41.2]). This difference may reflect a true difference in rates between the two camps or may be an effect of the methodology used to calculate and compare these rates due to the fact that most residents had lived in Awbare camp for fewer than 18 months. In the Uganda and Liberia Neighborhood Method studies (Stark et al., 2009; Stark et al., 2013), respondents consistently reported slightly higher rates of violence for themselves than for their neighboring households. Given the length of time Awbare camp had been in existence at the time of the survey, experiences of respondents and females in respondents’ households were reported over an 18-month recall period, whereas the experiences of neighbors were reported only since their arrival in the camp. To maintain consistency of length of recall periods across survey areas, the rates of physical violence and rape for Awbare camp were calculated by using only data from respondents and females in respondents’ households, whereas rates of physical violence and rape in Kebribeyah camp were calculated using data from the entire sample. When rates of physical violence were calculated in Kebribeyah camp using only respondents and members of respondents’ households, the incidence rate of physical violence increased to 46.5%, more closely approximating the rate of physical violence captured in Awbare camp (48.1%, 95% CI [42.2, 53.5]).

Violence During Flight as Compared With at Camp

The recall period in Awbare camp for respondents and females in the respondents’ household included time before the respondent and her family relocated to Awbare camp, and, thus, included the respondent’s and her family’s journey from Somalia to Ethiopia. Rates of violence and rape during women’s journeys from Somalia to Ethiopia were compared with rates of violence and rape in the camp setting. Findings indicate a relative risk of 1.8 (95% CI [1.6, 2.0]) of experiencing physical violence in the camp setting as compared with in flight from Somalia to Ethiopia. Similarly, findings indicate a relative risk of 1.3 (95% CI [1.2, 1.5]) of experiencing rape in the camp setting as compared with in flight from Somalia to Ethiopia.

These findings suggest that females in Awbare camp are at greater risk of experiencing physical violence and rape in the camp setting than in flight from Somalia to Ethiopia. The supplemental qualitative notes from the interviews supported these findings. When asked about their experiences in flight, women would often explain that the soldiers they encountered during their journey from Somalia to Ethiopia were “well-behaved” and served as protectors rather than perpetrators of violence.

Discussion

The findings of this study indicate that women and girls across sample areas are experiencing physical violence and rape at alarming rates. The overwhelming majority of physical violence and rape was domestic in nature, perpetrated primarily by husbands and other family members and occurring primarily at home. The findings of this study are consistent with a systematic review on the prevalence of GBV in humanitarian settings that found high prevalence of IPV across countries as well as with previous studies using the neighborhood methodology (Stark & Ager, 2011; Stark et al., 2009; Stark et al., 2013). A study that used the Neighborhood Method to investigate incidence of physical violence and rape against females in Liberia found that more than half (56%) of the study sample had experienced physical violence in the previous 18 months and three quarters of adult women had experienced domestic violence in the same time period (Stark et al., 2013). Similarly, a study of females in four internally displaced persons camps (IDP) in Uganda found that half (51%) of respondents reported IPV in the previous year (Stark et al., 2009)).

Given the current findings, it is critical for programming in refugee settings to address violence within the household. Services to support women in violent relationships and to help women improve their ability to keep themselves safe within the context of their families and marriages are essential. GBV programmers should adapt the messaging of their services for perpetrators and victims as well as the location and types of services to encourage women who are experiencing violence in the home to reach out for help and support. In addition to placing GBV response units in police stations or hospitals, GBV response units should be established in safe, non-stigmatizing locations where suspicions about a woman’s presence might not be raised.

In addition, country-level processes and accountability mechanisms related to issues of household violence and intimate partner rape need to be strengthened. Intimate partners should be included in the legal definition of rape, and policy-oriented advocacy coalitions, including media outlets, civil society groups, and research institutions, should work to raise awareness of the domestic nature of GBV.

Patterns of Reporting

The majority of reporting of incidents of GBV was of an informal nature to family, friends, or neighbors. Given the high rates of informal reporting, a separate analysis was conducted to assess whether rates of informal reporting for neighbors were significantly higher than rates of informal reporting for respondents (which could suggest a possible bias in the kinds of violent events reported for neighbors). Rates of informal reports varied by no more than 10% across sites except in the case of rape in Kebribeyah camp, suggesting that, overall, violence was shared widely through informal channels in this setting.

Rates of reporting to more formal mechanisms such as the police or courts remained low, particularly when the perpetrator was known to the survivor. Oftentimes, however, GBV programming and policy in emergencies rely on incident reports to the police, hospital, or other formal reporting mechanisms to estimate the magnitude and nature of GBV. Such case reporting mechanisms rarely provide an accurate representation of the magnitude and nature of GBV affecting a community. These systems capture only individuals who reported their experience to formal authorities, a small, and perhaps biased, minority of those who have experienced violence. Conclusions about the magnitude or nature of GBV based exclusively on data reported through formal mechanisms should be interpreted with caution as they are unlikely to be representative of the experiences of the majority of survivors and can lead to inaccurate understandings of patterns of violence in emergencies. Given that survivors of GBV are much more likely to report incidents informally to friends and family, it is important to consider the types of incidents that are most frequently reported through formal mechanisms and the implications of these reporting trends. Formal systems of reporting must be strengthened and safe, supportive, and confidential reporting mechanisms developed within the police, courts, and governmental and non-governmental structures. In addition, retaliation against survivors who formally report violence must be addressed to eliminate this barrier to formal reporting and increase survivors’ safety. Further research to explore ways to collaborate with the informal networks of reporting that are already being utilized is warranted.

Experiences of Violence in Flight and at Camp

Women were significantly less vulnerable to physical violence and rape during their time in flight from Somalia as compared with their time in the refugee camps. This finding contradicts assumptions that a displaced woman’s risk of violence and rape is greater in flight than post-flight. In addition, research on sexual violence among conflict-affected women in East Timor found that having been displaced to a camp setting was associated with increased risk of sexual violence post-conflict as compared with not having been displaced to a camp setting (Hynes et al., 2004). Little is known about the constellation of factors that may have contributed to experiences of relative safety in flight for this sample. However, these findings are also supported by the overwhelmingly domestic nature of both physical violence and rape across sites. In addition, flight from Somalia to the Ethiopia Somali Region may not be representative of typical emergency flight situations as flight from Somalia generally occurred more slowly and over a longer period of time than is typical when refugees are fleeing immediate danger. Further research is needed to determine whether the experience of Somali refugees in Ethiopia is unusual or whether some of the fundamental assumptions underpinning GBV programming in emergencies are flawed. This finding provides a unique opportunity to explore more fully the factors that contribute to relatively safe experiences of flight in which GBV was not many women’s primary concern. Such findings also have the potential to inform GBV-related programming and interventions in refugee camps post-flight. Understanding the factors that kept women relatively safe in flight can provide critical knowledge to inform future programming in displacement situations. Similarly, the factors that put females at increased risk of violence and rape in the camp setting must be explored, and policies and procedures established to increase the safety and security of refugee women and girls.

Limitations

The Neighborhood Method in general, and this study in particular, has several limitations worth noting. The Neighborhood Method relies on the interconnectedness and relative stability of communities during the length of the recall period. Residents need to have been situated long enough to know about the experiences of one another. The method will not work well for a newly settled refugee population where individuals do not yet know one another. In addition, the Neighborhood Method is best suited for clearly defined events and, thus, would be difficult to use with less clearly defined or more subjective outcomes such as psychological or emotional violence. The Neighborhood Method relies on respondents’ willingness to report on their neighbors’ experiences of violence. Respondents may be unwilling to report on neighbors’ experiences of violence if they do not feel safe or that their report will remain anonymous. Unwillingness to report on neighbors’ experiences of violence would lead to underestimation of GBV among women other than the respondent. To investigate this bias, rates between self- and secondary reports were compared in Kebribeyah camp and Awbare town. The recall period in Awbare camp made such a comparison impossible for this location. In both sites, women reported statistically significantly higher rates of physical violence for themselves and females in their households than for their neighbors and females in their neighbors’ households. However, reported rates of rape did not differ significantly between respondents and females in their households as compared with neighbors and females in neighbors’ households. Similar results were found when Stark et al. (2013) investigated this potential bias in a Neighborhood Method study of GBV among women and girls in Liberia. A non-statistically significant trend toward higher rates of domestic violence was found among respondents as compared with female heads of neighboring households in Liberia; there was no statistically significant difference in reporting of rape between respondents and female heads of neighboring households (Stark et al., 2013). As a result, it is likely that this study underestimates the true prevalence of physical violence occurring in these communities. The Neighborhood Method requires well-trained interviewers to prevent large variations between interviewers in prevalence of violence reported. The interviewing team for this study was previously inexperienced with issues of GBV and data collection. The research team invested a significant amount of time in training the interviewers throughout this project in both GBV and data collection skills. Issues such as consistency in coding, effective interviewing techniques, and general knowledge regarding GBV were paramount during training. In addition, no primary data were obtained on the experiences of children. As only women above the age of 18 were interviewed, all data on children are secondary reports. As such, the experiences of children captured in this study need to be further verified.

Ethics

Even though the research team felt confident that the WHO recommendations had been fully implemented throughout the research, the research team incorporated an additional ethical check following the interviews. To assess the well-being of respondents following their involvement with this study, the research team returned to Awbare camp 4 weeks after conducting their original interviews. During these second interviews, interviewers asked the original respondents whether they experienced any positive or negative consequences as a result of being interviewed for this study. Sixteen of 79 women re-interviewed reported experiencing violence from a family member that they attributed to participation in the interview. In addition, six of 79 women reported violent disputes with neighbors, and six of 79 reported non-violent disputes with neighbors that they attributed to participation in this study.

Sadly, this is not the first study to discover negative unintended consequences of research in humanitarian settings. There are numerous examples from the field where programs and research with good intentions caused unintended harm. In 1996 in Rwanda, for example, centers were established to house and protect unaccompanied and separated children after the genocide. An unfortunate and unanticipated consequence was that these centers contributed to family separation, as mothers desperate to support their babies abandoned their babies on the orphanages’ doorsteps (Wessells, 2009). In another example, a randomized control trial of Mozambique refugees in Malawi, attempting to determine whether a bucket with a cover and a spout prevented diarrheal diseases compared with the standard open bucket, had the unintended consequence of encouraging those who received the covered bucket to stay in the camp. At the end of 6 months, 15% of the control group had returned home, whereas only 1% of the intervention group had repatriated (Roberts et al., 2001).

Furthermore, in the case of GBV especially, certain negative unintended consequences are not only unsurprising, but may be expected. Briefings on these findings triggered three informal reports of similar experiences in the past that had not been documented or publicized. Although the research team was devastated by the results of this ethical check, the local team was largely unsurprised, expecting some level of negative response from the community. Based on conversations with the GBV staff, it is not unusual for communities to engage in harmful practices when a GBV program is first established. Another example of unintended negative consequences can be found with the Grameen Bank model. This much-lauded model disperses micro-credit loans to women. Household violence against women receiving loans has been documented (Dowla, 2001).

Although similar findings of unintended negative consequences were actively sought out and not found in the Central African Republic (Potts, Myer, & Roberts, 2011), Sri Lanka (Rogers, Anderson, Stark, & Roberts, 2009), or any subsequent study that the research team has since undertaken, these unintended results provide important lessons for all researchers intending to study such sensitive topics moving forward. Follow-up ethical checks such as the one conducted in this study are rarely conducted. These data highlight the critical importance and ethical obligation that researchers face to investigate and understand the unintended consequences of their research. The full implementation of existing ethical guidelines, although necessary, remains insufficient. Researchers must also incorporate mechanisms within their research to understand the consequences of participation for study participants. Equally important is the need to create a culture of learning where researchers and humanitarian responders feel able to identify and disclose the unintended negative consequences of their work.

Conclusion

This population-based survey was conducted to quantify the magnitude of physical violence and rape perpetrated against females in two Somali refugee camps and one of the surrounding Ethiopian host communities. The findings of this study demonstrate that women and girls across all sample areas are experiencing high rates of physical violence and rape. The overwhelming majority of physical violence and rape was perpetrated by husbands and other family members and occurred within the home. The domestic nature of physical violence and rape across survey areas must be addressed, and programming and services must be further developed for survivors of GBV. Services to support women in violent relationships and to strengthen women’s ability to keep themselves safe within the context of their families and marriages are essential. More must be done to strengthen systems of formal reporting and to develop safe and supportive reporting mechanisms within the police, courts, and governmental and intergovernmental structures.

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) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: The Program on Forced Migration and Health acknowledges the United States Agency for International Development (Displaced Children and Orphans Fund) and the Oak Foundation for the financial support and the United States Institute for Peace for its work on child protection.

Author Biographies

Angela Parcesepe is a postdoctoral research fellow at the HIV Center for Clinical and Behavioral Studies at Columbia University and the New York State Psychiatric Institute. . Her research focuses on gender-based violence generally and, more specifically, on gender-based violence in situations of war, disaster, or displacement. Her research also explores the intersection of gender-based violence, substance use, and HIV risk. She received her PhD from the University of North Carolina at Chapel Hill.

Lindsay Stark is an assistant professor of clinical population and family health in Columbia University’s Program on Forced Migration and Health. She helped pioneer the development of methodologies such as the Neighborhood Method to assess incidence of human rights violations, a participatory ranking method included in a recent World Health Organization (WHO) assessment toolkit, and the Child Protection Rapid Assessment in Emergencies Toolkit developed by the global Child Protection Working Group. She serves as principal investigator and executive director of the Child Protection in Crisis (CPC) Learning Network, a consortium working on global learning associated with children in disaster and war.

Leslie Roberts holds a PhD in environmental engineering and did a post-doctorate fellowship in epidemiology at the Centers for Disease Control and Prevention and afterward worked for WHO in Rwanda during the 1994 civil war and then for the International Rescue Committee. Presently, he is director of the Program on Forced Migration and Health at Columbia University. His present research focuses on developing methods to document human rights violations and on statistically representative surveillance methodologies.

Neil Boothby is the Allan Rosenfield professor of Forced Migration and Health at Columbia University’s Mailman School of Public Health where he founded the CPC Learning Network, a constellation of 80 agencies working in a dozen countries to develop an evidence base for efficacious child development policies and programs. He has published extensively on the subject of children and adversity and also received a number of awards for his field work, including the Red Cross Humanitarian of the Year Award, the Mickey Leland Award, and the United Nation’s Global Achievement Award for Excellence in the Social Sector.

References

Campbell

(2002). The health consequences of intimate partner violence. Lancet, 359, 1509-1514.

Campbell

Dworkin

Cabral

(2009). An ecological model of the impact of sexual assault on women’s mental health. Trauma, Violence, & Abuse, 10, 225-246.

Card

(1996). Rape as a weapon of war. Hypatia, 11, 5-18.

Dowla

(2001). Grameen bank and women’s empowerment in Bangladesh: A review essay. Journal of Microfinance, 3, 159-169.

Ellsberg

Jansen

H. A.

Heise

Watts

C. H.

Garcia-Moreno

(2008). Intimate partner violence and women’s physical and mental health in the WHO multi-country study on women’s health and domestic violence: An observational study. Lancet, 371, 1165-1172.

Garcia-Moreno

Stockl

(2013). Violence against women, its prevalence and health consequences. In Riecher-Rossler

Garcia-Moreno

(Eds.), Violence against women and mental health: Key issues in mental health (Vol. 178, pp. 1-11). Basel, Switzerland: Karger.

Hynes

Robertson

Ward

Crouse

(2004). A determination of the prevalence of gender-based violence among conflict-affected populations in East Timor. Disasters, 28, 294-321.

Kaminer

Grimsrud

Myer

Stein

Williams

(2008). Risk for post-traumatic stress disorder associated with different forms of interpersonal violence. Social Science & Medicine, 67, 1589-1595.

Krug

E. G.

Dhalberg

L. L.

Mercy

J. A.

Zwi

Lozono

(Eds.). (2002). World report on violence and health. Geneva, Switzerland: World Health Organization.

10.

Loescher

Milner

(2005). Protracted refugee situations: Domestic and international security implications. New York: Routledge.

11.

Peterman

Palermo

Brendenkamp

(2011). Estimates and determinants of sexual violence against women in the Democratic Republic of Congo. American Journal of Public Health, 101, 1060-1067.

12.

Potts

Myer

Roberts

(2011). Measuring human rights violations in a conflict-affected country: Results from a nationwide cluster survey in Central African Republic. Conflict and Health, 5, Article 4.

13.

Roberts

Chartier

Malenga

Toole

Rodka

(2001). Keeping clean water clean in a Malawi refugee camp: A randomized intervention trial. Bulletin of the World Health Organization, 79, 280-287.

14.

Rogers

Anderson

Stark

Roberts

(2009). Estimating the incidence of physical and sexual violence against children and women in Trincomalee District, Sri Lanka: The neighborhood method. Retrieved from http://www.cpcnetwork.org/resource/estimating-the-incidence-of-physical-and-sexual-violence-against-children-and-women-in-trincomalee-district-sri-lanka/

15.

Sheridan

D. J.

Nash

K. R.

(2007). Acute injury patterns of intimate partner violence victims. Trauma, Violence, & Abuse, 12, 115-126.

16.

Stark

Ager

(2011). A systematic review of prevalence studies of gender-based violence in complex emergencies. Trauma, Violence, & Abuse, 12, 127-134.

17.

Stark

Roberts

Wheaton

Acham

Boothby

Ager

(2009). Measuring violence against women amidst war and displacement in northern Uganda using the “neighborhood method.” Journal of Epidemiology & Community Health, 64, 1056-1061.

18.

Stark

Warner

Lehman

Boothby

Ager

(2013). Measuring the incidence and reporting of violence against women and girls in Liberia using the “neighborhood method.” Conflict and Health, 7, Article 20.

19.

Stark

Wessells

(2012). Sexual violence as a weapon of war. Journal of the American Medical Association, 308, 677-678.

20.

Stockl

Devries

Rotstein

Abrahams

Campbell

Watts

Moreno

C. G.

(2013). The global prevalence of intimate partner homicide: A systematic review. Lancet, 382, 859-865.

21.

United Nations High Commission for Refugees. (2015). Refugees in the Horn of Africa: Somali displacement crisis. Retrieved from http://www.unhcr.org/pages/49e483ad6.html#

22.

Wessells

(2009). Do no harm: Toward contextually appropriate psychosocial support in international emergencies. American Psychologist, 64, 842-854.

23.

World Health Organization. (1999). Putting women’s safety first: Ethical and safety recommendations for research on domestic violence against women. Geneva, Switzerland: Global Program on Evidence for Health Policy, World Health Organization.