The association between perceived neighborhood social cohesion and intimate partner violence in a refugee camp in Dollo Ado, Ethiopia

Abstract

Background

Intimate partner violence (IPV) is the most common form of gender-based violence affecting women and girls worldwide and is exacerbated in humanitarian settings. There is evidence that neighborhood social processes influence IPV. Perceived neighborhood social cohesion (P-NSC)—a measure of community trust, attachment, safety, and reciprocity—may be protective against women’s experience of and men’s perpetration of IPV and controlling behaviors.

Methods

A quantitative social network study, comprised of individual verbally-administered surveys, was conducted in Bokolmayo refugee camp in Dollo Ado, Ethiopia in 2019. In total, 302 Somali refugees (147 women and 155 men), sampled using snowball sampling, participated in the data collection. Logistic regression was used to examine P-NSC and its association with IPV to inform an IPV and HIV prevention intervention.

Results

Low P-NSC and men’s perpetration of physical IPV in the past month were strongly associated (adjusted AOR = 23.6, 95% CI: 6.2–89.9). Low P-NSC, conversely, was associated with decreased odds of women’s experiences of controlling behaviors by an intimate partner in the past year (AOR = 0.1, 95% CI: 0.0–0.5). Women’s experiences of other forms of IPV, including physical, sexual, and emotional IPV within the past year, were not associated with P-NSC in adjusted models; P-NSC was significantly associated with all forms of IPV in unadjusted models.

Conclusion

Social cohesion programs and other neighborhood approaches to improve P-NSC should be explored as potential avenues to prevent and reduce IPV, with a focus on male IPV and controlling behavior perpetration.

Introduction

Intimate partner violence (IPV) has adverse health, social, and economic implications for women and families [1, 2]. Intimate partner violence is defined as “behavior by an intimate partner that causes physical, sexual, or psychological harm, including physical aggression, sexual coercion, psychological abuse, and controlling behaviors”[3]. Globally, 27% of women have experienced lifetime physical and/or sexual IPV [4]. In humanitarian settings, IPV prevalence is higher [5] as displacement has been shown to disrupt gender norms – a key driver of IPV risk, along with family and community cohesion and support systems [6, 7].

Heise’s socio-ecologic model, alongside adapted versions, conceptualizes multiple individual, relationship, community, and society level factors interacting to influence IPV risk [8, 9]. While many studies have focused on individual- and relationship-level factors, there has been increasing attention on community-level factors such as residential environments and their role in IPV [10]. Scholars have posited that high neighborhood social cohesion can prevent IPV due to a sense of trust, fostering bystander action through the exchange of support, information and resources, social capital, and collective efficacy [11,12,13]. However, neighborhood norms can be pluralistic—meaning multiple norms exist and their influences vary [14]. For example, within the same community, some neighbors may be more likely to seek support from each other or access resources to deter IPV perpetration, while others may not feel obligated to intervene, or may condone and amplify IPV through victim-blaming [12, 13, 15]. Accordingly, measuring perceptions of neighbourhood processes is important as it defines neighbourhoods through the perspectives of residents and improves comparability across residents, which is especially valuable in the context of cross-cultural research [5, 10, 16].

Social disorganization theory offers a valuable framework for understanding how neighborhood factors can influence IPV [10, 17]. The theory posits that communities with weaker social organization—characterized by limited social cohesion, economic instability, and a lack of collective efficacy—struggle to maintain informal social control, creating an environment where violence is more likely to occur. These factors can undermine community support networks and social norms that discourage violence, contributing to conditions that enable IPV.

Supporting this theory, research has shown that perceived neighborhood social cohesion (P-NSC) and other neighborhood processes are important predictors of IPV, other forms of violence, and health behaviors and outcomes [10, 18,19,20,21,22,23,24,25]. Definitions of P-NSC often comprise the level of trust (e.g. in people including members of the neighborhood who are not personally known), attachment (e.g. feeling part of the community), safety, and reciprocity [26,27,28,29]. High P-NSC has been shown to be associated with reduced male-perpetrated IPV [10, 30], and attenuated adverse mental health outcomes associated with IPV [10]. However, no studies have examined links between P-NSC and male controlling behaviors. Additionally, some interventions have been shown to improve social cohesion, solidarity and reciprocity within communities, suggesting these may be modifiable protective factors that could be addressed as part of IPV prevention strategies [31, 32].

Fewer studies have examined indicators of social cohesion such as P-NSC in humanitarian contexts, where social cohesion and community connectedness may be reduced due to displacement. Existing research from conflict-affected, refugee-hosting areas [13, 33], refugee camps, or other settlements [6, 34, 35] suggests neighborhood factors may be influential on residents’ health and emotional well-being in these contexts. In addition, low social cohesion can contribute to increased social tensions and inter- and intra-community conflict and violence [28, 36], and delivery of humanitarian aid can have unintended effects on the quality of relations between displaced populations and the communities they settle in [35]. Many humanitarian agencies have, therefore, integrated social cohesion as an important policy and programming objective in order to build and sustain peace in displacement contexts – typically with a focus on strengthening relations between refugees and host communities [28, 37,38,39]. However, there is a lack of clear definitions and validated indicators for inter- and intra-community social cohesion for these settings [28, 40], as well as limited data on factors that enhance or reduce social cohesion and on the effectiveness of social cohesion interventions. In addition, prior studies have focused on the social nexus between displaced and hosting communities; few have assessed intra-community social cohesion within the refugee or host communities themselves [28, 40].

The literature on linkages between P-NSC and IPV in humanitarian contexts is also limited. A qualitative study conducted among three refugee camps in South Sudan, Kenya, and Iraq, noted social cohesion was weaker in camps than pre-displacement and consequently that low social cohesion was a potential driver of IPV [6]. Moreover, in testing a social norms approach to improve IPV-related outcomes based on community-generated solutions, social cohesion was found to increase in Malaysia but not Lebanon, highlighting the influence of contextual factors, such as the local policy environment, camp spatial living configurations, and other cultural and displacement related characteristics on social cohesion [29, 35]. Additional research is needed to expand understanding of the influence of neighborhood processes on IPV risk in humanitarian contexts.

This study sought to build evidence on social cohesion and IPV risk within a refugee community. This study examines the association between P-NSC and women’s experience and men’s perpetration of IPV and controlling behaviors against women among Somali refugees living in Bokolmayo camp in Dollo Ado, Ethiopia. We hypothesized that low P-NSC is associated with both women’s experience and men’s perpetration of IPV and controlling behaviors.

Methodology

Quantitative data were analyzed from a social network study conducted to inform a podcast-based adaptation of Unite for a Better Life (UBL), a gender-transformative program designed to prevent IPV and HIV transmission in rural Ethiopia [41, 42]. The data were collected in 2019 at Bokolmayo camp, one of five refugee camps that opened in 2010 near Dollo Ado, Ethiopia – a small-town bordering Somalia [42,43,44]. At the time of the study, over 200,000 Somali refugees were registered in the five camps, with over 40,000 refugees residing in Bokolmayo [43,44,45]. The study was implemented by researchers at Women and Health Alliance (WAHA) International in Ethiopia, Addis Ababa University School of Public Health, Beth Israel Deaconess Medical Center (BIDMC) and the Harvard T.H. Chan School of Public Health (HSPH). The study was reviewed and approved by institutional review boards at HSPH in Boston, Massachusetts (IRB17-0867, June 2017), as well as the Addis Ababa University in Ethiopia (044/16/SPH, September 2017). Permissions to conduct the research were also obtained from UNHCR and the Administration for Refugee and Returnee Affairs (ARRA). Verbal informed consent was obtained from all study participants.

The study was primarily designed to assess social networks in the camp, and due to resource constraints focused on assessing networks of a subset of the population by randomly selecting a fixed number of households (or seeds) and identifying their connections using snowball sampling [46]. The overall target sample size was set at 300 surveys and was determined by resource availability. An initial sample of 16 households (or seeds) (8 women, 8 men) was randomly selected in two purposively sampled zones out of the 20 zones of the camp. Assuming an average of 4 close connections per respondent, the initial sample of 16 seeds would allow inclusion of all first- and second-degree connections of the initial sample. In each of the sampled seed households, one eligible man or woman was invited to participate in the study, with half of the households randomly sampled for a male interview and the other half sampled for a female interview. As part of the survey, each seed respondent was asked to identify their close relationships of any sex and provide contact information for these individuals. These close contacts (or first-degree connections) were subsequently invited to participate in the survey. Similarly, those participants’ close contacts were then invited to participate in the survey (second-degree connections). This process continued until a sample size of at least 300 residents was met. Inclusion criteria included: men or women who were at least 18 years of age and that identified as a Somali refugee residing in Bokolmayo refugee camp, and who were able to provide informed consent.

WHO recommendations for researching violence against women and girls were adhered to at all stages of the project [47]. A total of four male and four female trained enumerators collected the data. All data collectors completed a six-day training focused on protection of human subjects, quantitative and qualitative methods, interviewing techniques, and risk mitigation. All data collectors were fluent in Somali and local dialects, had experience working for an NGO or conducting research, and were recruited from within Bokolmayo camp. Data collectors were supervised by a field manager hired from within the same community. Engaging team members from the same community facilitated trust-building and sharing of views. The study team did not interview any participants they knew personally, followed a code of conduct, and followed the WHO ethical recommendations for researching VAWG guidelines to safeguard data collectors [47]. A research manager based in Addis Ababa directly managed all data collection procedures and training. Interviewers of the same sex as the participants administered the questionnaires in Somali language. Interviews were conducted in participants’ homes in a private location, with no one else present. A list of local medical, legal, and other relevant support services was given to participants upon completion; referrals for psychological support were also provided. Surveys were administered via electronic questionnaires programmed using SurveyCTO on mobile devices and registered on password-protected digital devices. Electronic files were uploaded to a password-protected, encrypted laptop by the study team and stored on a secure, password protected server.

Questionnaire and measures

The social network questionnaire included sections on demographics (age, ethnicity, literacy and education level, relationship [marital status, partner characteristics]), displacement, alcohol and khat use, social networks (including information about individuals that the respondent discusses important matters with, and socializes with, neighborhood cohesion, social norms), women’s experience of physical, sexual and emotional IPV and men’s perpetration of physical IPV. The questions on neighborhood social cohesion were adapted from the Conjoint Community Resilience Assessment Measurement (CCRAM) tool for assessing community resilience in emergency settings [48]. This tool was selected because it was designed and used for assessing community cohesion, trust and other dimensions in settings similar to the study location. Because of length considerations and because of the study’s focus on cohesion, only the questions that capture community connectedness, trust, safety and cohesion were included (a total of 9 questions). These were adapted and contextualized to the population based on stakeholder feedback and piloting to generate a measure of P-NSC and are shown in Appendix 1. The women’s IPV module was adapted from the WHO multi-country study on domestic violence and women’s health questionnaire [49]. In addition, measures for experience of past-month physical IPV among women and perpetration of past-month physical IPV by men were included to capture more recent IPV. The main measures analyzed for this paper are described in detail in Table 1. In addition, demographic and social status variables included: age in years, ethnicity (majority vs minority clan), whether the respondent was in a polygamous relationship, the type of area of residence prior to displacement (suburban, urban or nomadic), literacy level, average monthly income (including cash transfers), any formal education, current employment (unemployed, working for trade or working for pay), reason for displacement (natural disaster, safety, other), ever use of alcohol or khat, partner’s ever use of alcohol or khat, and the number of years living in the camp. Age was converted to a categorical variable with three categories (< 30 years, 30–39 years, > 39 years). The questionnaire was translated to Somali language and back translated by a native Somali speaker who was also fluent in English and had expertise in gender. The instrument was piloted in the camp.

Table 1 Key outcome measures

Analysis

The final data analysis was conducted in Stata SE version 16.0. For the present analysis, men and women whose reported partnership status was not married or partnered at the time of enrollment (n = 1 woman, n = 16 men) were excluded due to the study’s focus on IPV. The final study sample included 155 women and 147 men. Separate models were constructed for women’s experience and men’s perpetration of IPV. Bivariate analyses, with chi square tests for categorical variables and analysis of variance tests for continuous variables, were completed to examine associations between demographic characteristics with P-NSC and IPV separately. Multiple logistic regression models were used to calculate adjusted odds ratios and 95% confidence intervals for the association between P-NSC and IPV. Following this, a change-in-estimate approach was used to create a parsimonious model. Variables that were not associated with the outcomes (p > 0.10) and whose removal did not change the main effect by more than 10% were removed. The final adjusted models for women’s experiences of IPV controlled for years displaced, ethnicity, polygamy, neighborhood type before displacement, literacy, years in camp, monthly income and cash transfer amount, and reason for displacement. The final adjusted model for men’s perpetration of IPV controlled for employment, education, polygamy, ethnicity, and personal khat use.

Role of the funding source

The funders of the study had no role in study design, data collection, data analysis, data interpretation, or writing of the report. The corresponding author had full access to all the data in the study and had final responsibility for the decision to submit for publication.

Results

Participant characteristics are shown in Table 2. The women were predominantly between 18–39 years old, largely unemployed, from nomadic and urban areas, and displaced due to natural disaster or threat to safety. The men were mostly over 39 years old, currently employed, from suburban and urban areas only, and displaced due to threat to personal or family safety in Somalia. Both women and men were mainly from minority ethnic clans. Furthermore, reported lifetime personal and partner use of alcohol (personal – 3%, partner – 1%) and khat (personal – 2%, partner – 8%) was low among women participants. Among male participants, reported lifetime personal and partner use of alcohol (personal – 5%, partner – 2%) and khat (personal – 11%, partner – 1%) was slightly higher. Women and men reported residing in the camp an average of 8 and 10 years, respectively. A substantial proportion of women (24%) and a smaller proportion of men (6%) reported being in polygamous unions.

Table 2 Select demographic characteristics of women and men in the study, experience and perpetration of IPV and P-NSC

P-NSC differed considerably between men and women and is shown in Fig. 1. The women in the study reported a range of responses for P-NSC, with most women (51%) reporting moderate P-NSC. In contrast, 72% of men reported the highest possible score of 9 on the P-NSC suggesting that most men perceived their neighborhood as having a very high social cohesion. Accordingly, we categorised P-NSC based on the distribution of responses for women and men separately. For women, P-NSC was categorised as high (score 8–9), medium (6–7), and low (0–5). For men, P-NSC was binary with a score of 9 categorised as high and 0–8 as low.

Fig. 1

Distribution of P-NSC Scores among Women and Men

Approximately 37% of women reported experiencing physical IPV in the past month. In the past year, 19% of women reported experiencing physical IPV; 10% reported sexual IPV; 16% reported emotional IPV; and 43% reported controlling behaviors by a partner. Among the men, 44% reported perpetration of physical IPV in the past month.

Bivariate analysis

The association between P-NSC and selected covariates for women and men are shown in Table 3. Women’s demographic characteristics that were associated with P-NSC included: neighborhood type before displacement, reason for displacement, monthly income and/or cash transfer amount, and partner khat use. Literacy, polygamy, and years in camp also were marginally associated with women’s P-NSC. Men’s demographic characteristics that were associated with P-NSC included: education, employment, personal khat use, age, ethnicity, and neighborhood type before displacement.

Table 3 Association between selected covariates and P-NSC

The association between women’s experience and men’s perpetration of IPV and selected covariates are shown in Tables 4 and 5, respectively. Among women, living in an urban neighborhood before displacement, being displaced for a reason other than natural disaster or safety, and partner khat use were all associated with experience of emotional, sexual, and physical IPV in the past year as well as physical IPV in the past month. Higher monthly income and/or cash transfer amount was associated with experience of physical IPV in the past month. Polygamy was associated with physical IPV. Controlling behaviors were also associated with more years in the camp and lower income level. Among men, ethnicity, age, neighborhood type before displacement, education, reason for displacement, and personal khat use were all associated with perpetration of physical IPV in the past month.

Table 4 Association between selected covariates with women’s experiences of physical and sexual IPV

Table 5 Association of selected covariates with men’s perpetration of physical IPV

Multivariable analysis

The logistic regression models for P-NSC and women’s experience and men’s perpetration of IPV are shown in Tables 6 and 7, respectively. Women who reported medium and low P-NSC had an increased odds of experiencing physical IPV in the past month, and physical, sexual, and emotional IPV in the past year in crude analyses, but none of these associations were statistically significant in adjusted models. However, women who reported low P-NSC had 90% lower odds (OR = 0.1; 95% CI: 0.0–0.5) of experiencing controlling behaviors by a partner compared to those reporting high P-NSC which persisted (OR = 0.1; 95% CI: 0.0–0.6) after adjustment for covariates. Men who reported low P-NSC had 20.1 times higher odds (95% CI: 7.2–56.2) of perpetrating physical IPV in the past month compared to men who had high P-NSC in crude analyses. These remained statistically significant when adjusting for covariates (OR = 22.0, 95% CI: 6.5–74.2).

Table 6 Logistic Regression Models of P-NSC and Women’s Experience of IPV

Table 7 Logistic Regression Models of P-NSC and Men’s Perpetration of IPV

Discussion

This study finds that in a refugee context in Dollo Ado, Ethiopia, low P-NSC among men was associated with significantly increased odds of male perpetration of physical IPV in the past month; whereas, low P-NSC among women was associated with a significantly lower odds of experiencing of controlling behaviors by an intimate partner. In contrast, we did not find significant associations between women’s P-NSC and reported experience of past-month physical IPV or past-year physical, sexual, or emotional IPV in adjusted models.

The finding of increased odds of male perpetration of IPV among those with low P-NSC is consistent with other research in refugee and other populations [10, 12, 19, 35, 50, 51]. Coupled with the null findings regarding women’s P-NSC and experiences of IPV, our findings suggest social cohesion may be more directly associated with IPV perpetration, an action or behavior, rather than the experience of IPV. A similar finding was reported in a study conducted among Colombian immigrants in Spain, where variation in dating violence perpetration by young men was mainly explained by neighborhood factors, while variation in women’s experiences of IPV was primarily explained by individual factors [52]. More diverse research is needed to better understand the relationship between social cohesion and IPV in a variety of settings.

Our study found a strong and significant link between women’s P-NSC and controlling behaviors by an intimate partner but not with other forms of IPV. To our knowledge, no previous study examines controlling behaviors and few compare specific forms of IPV (i.e. physical vs sexual vs psychological) in relation to social cohesion. Thus, our research provides new insights and suggests that various forms of IPV experienced by women may be differentially related to social cohesion. In particular, controlling behaviors by an intimate partner, which measures a partner’s influence on a resident’s interactions with their neighborhood (e.g. keeping wife/partner away from family, expecting wife/partner to ask for permission to leave home or access healthcare) may be more directly related to social cohesion than other forms of IPV, at least in this context. The direction of the association between P-NSC and controlling behaviors, is interesting. Previous literature suggests that displacement reduces social cohesion, placing women at increased risk of IPV. For example, a qualitative study in three refugee camps in South Sudan, Kenya, and Iraq found lower social cohesion post-displacement, leaving women unsupported, isolated and less likely to seek help or access resources and, as a result, potentially at higher risk of IPV [6]. However, that study did not assess an association between social cohesion and IPV or controlling behaviors.

Several possible explanations may account for our findings. In contexts where IPV is normative, high social cohesion may not be protective against violence/controlling behaviors as neighbors may not feel an obligation to intervene or provide support. Rather, they may collectively support or reinforce such behaviors. For example, the same study in three refugee camps, reported how friends, family members and neighbors helped fuel violence or allowed it to occur [6]. Alternatively, controlling behaviors can be associated with more severe forms of IPV [53, 54]; women experiencing controlling behaviors may have been more likely to seek informal or formal social support and thus report higher levels of trust towards their community. Finally, the P-NSC questions used have not been validated in Somali language in the refugee camp context, potentially limiting their ability to fully capture social cohesion in this context. Other constructs which could influence the association between P-NSC and IPV, such as social disorganization and community norms supportive of violence, should be further studied.

Our findings may help illustrate how to apply or consider neighborhood theory in relation to a refugee camp context which differs from other residential neighborhoods. Some potentially unique characteristics of refugee settlements are residential self-selection, residential instability, shared collective trauma, social and cultural norms, mixing of different ethnicities, tribes and clans, and institutional control of the environment. Firstly, residential self-selection or the ability to select one’s neighborhood may be limited in this type of setting [16]. Secondly, research has shown a range of positive, negative, and null associations between IPV and residential instability. Residential instability has typically been measured as either the percentage of residents living in their current households for five years or less and the percentage or ratio of houses occupied by owners as opposed to renters [10, 16, 18]. Refugee camps are, by definition, temporary settlements, however, many displacement situations are becoming increasingly protracted. For example, women and men in our study reported residing in Bokolmayo for an average of 8 and 10 years. Given this, the five-year threshold for the instability measure may require adaptation for use in different types of humanitarian contexts and may need to be considered differently in refugee versus hosting communities. For example, one study examining a refugee context in Ecuador defined residential instability as the percent of households that moved to the neighborhood in the last 20 years [13]. Because residential instability has been shown to increase social cohesion independent of neighborhood disadvantage, it could be further defined and assessed as part of future research [10]. Thirdly, as part of a displaced community that has experienced collective trauma, residents’ potential willingness to improve social cohesion may be different than residents in other contexts. Previous research has shown that shared interpretation of collective experiences and creating new social networks is integral amongst this population [33]. Social cohesion in a refugee setting, may be influenced by the mixing of different ethnic groups, tribes, or cultures. This could foster mutual understanding and collaboration, but it could also exacerbate divisions if historical tensions or differences in language and norms persist. Lastly, although social processes do not exist in a vacuum, refugee camps may have a higher degree of institutional control than other types of neighborhoods. In refugee settlements, social processes, such as social cohesion, may be greatly dependent on organizational resources, capacity, and management and consequently driven by the nature of relationships with and between humanitarian and other agencies [10, 16, 55]. These institutions and others whose work largely informs the structural context, should consider integrating a neighborhood approach to reduce IPV through social processes in conjunction with other effective IPV prevention strategies.

Limitations

Our findings should be interpreted with the following limitations in mind. First, IPV against women was examined because women more frequently experience IPV from severe acts of physical violence and chronic patterns of control and abuse. Violence against men, and other forms of GBV experienced by women such as non-partner sexual violence were not examined in this study. Secondly, the study only captured data on male perpetration of physical IPV in the past month. Perpetration of other forms of violence including sexual, psychological IPV and controlling behaviors were not captured. Perpetration of past-year IPV was also not assessed. Thirdly, snowball sampling does not guarantee representative sampling, which limits the external validity of our results. The initial sample however was randomly selected to mitigate potential bias and ensure as much diversity in sampling as possible. Snowball sampling could also have resulted in some men and women in the same households interviewed about IPV, which is not recommended in ethical guidelines on IPV research. However, we worked with local stakeholders and the local IRB on prior couples’ IPV research in the same setting and safely collected IPV data from both partners. Furthermore, the cross-sectional design precludes inferences on the directionality of the associations reported; this is a shared limitation with most of the existing literature [10, 18]. As in all observational studies, residual or unmeasured confounding cannot be fully discounted. Finally, limited sample size and potential under-reporting of key measures including those related to IPV may have reduced the statistical power. For example, social desirability bias could have led to underreporting, but interviewers were recruited from the camp and were trained in strategies to build trust and rapport, interviews were conducted in private with no one else present and confidentiality was maintained. Recall bias could also have been an issue, though this study focused on violence in the past 12 months. Despite these limitations, our study is one of the few to examine P-NSC and its relation to IPV and one of the first, to our knowledge, to examine this association in relation to both men’s perpetration and women’s experiences of IPV in a refugee camp context.

Implications for research and programming

Our findings have implications for humanitarian research, evaluation and programming. In order to design more effective IPV prevention strategies, further research is needed to better understand community-level drivers of IPV (including P-NSC, community collective action, etc.) and how they interact with each other and with individual- and household-level drivers of IPV. This will also require improving quantitative and qualitative measurement of social cohesion in conflict-affected and displaced populations through adaptation and validation of existing scales and tools. Humanitarian contexts vary greatly and settlement patterns could range from closed camps with complete physical separation between refugee and host communities, to open camps where movement in and out is permitted, to urban or rural cohabitation [28]. Indicators appropriate for different displacement scenarios, and for both intra- and inter-community social cohesion and related factors, and that are able to identify changes over time are needed. Additional research, particularly qualitative approaches, would be well suited to understand local definitions of social cohesion and suitable measurement approaches, as well as the appropriate timing and strategies to foster social cohesion given the realities of long-term displacement [28]. Additional research could also be useful in further exploring the gendered differences in perceptions of neighborhood social cohesion that were identified in our study findings.

GBV risk mitigation interventions that reduce exposure to GBV and ensure that humanitarian response actions and services themselves do not cause harm or increase risk of violence are recommended across all humanitarian sectors [56, 57]. However, social cohesion and other community-level risk factors for violence may not be sufficiently understood or considered in relation to humanitarian aid or GBV risk mitigation. As an example, in “Good Shelter Programming, Tools to Reduce the Risk of GBV in Shelter Programming” recommended strategies to mitigate GBV risk in camp design and layout include among others, engaging women and other groups at high risk of GBV in the shelter design process, using shelter materials that maximize safety and privacy and considering distance between shelters and from services [57]. This guidance describes the potential for increased GBV risk when there is overcrowding or integration of groups that do not have a history of social cohesion but does not suggest approaches for strengthening social cohesion as a risk mitigation measure. While in some refugee contexts, interventions to enhance and promote social cohesion have been implemented, these may not take into consideration how social cohesion and GBV interact, or gendered dynamics of social cohesion – and in some cases could be exacerbating exclusion and insecurity for women and girls [58]. There is a need for innovative social cohesion interventions that take a gender transformative approach, considering the intersection of gender, GBV and social cohesion. Such interventions should be designed and evaluated in collaboration with local women’s organizations and women and girls. As there is limited evidence on whether such strategies could help to mitigate GBV risks in humanitarian contexts, research and evaluation to assess the impact of these interventions is needed.

In parallel, community-level drivers should be considered when developing GBV and IPV programming in these settings. As prevention programming often focuses on individual- and household-level factors, considering social cohesion and other community-level drivers such as community norms around violence could be an important avenue to reduce perpetration of IPV. For example, analysis of data from the SASA! trial found that community norms on the acceptability of violence and progressive gender relations were the major mediators in the intervention’s impact on reducing men’s perpetration and women’s experience of IPV [59]. The authors suggest that peer pressure and perceived threats of sanctions among men are potentially critical mechanisms for preventing IPV perpetration [60]. Thus, the same social and peer networks which can enable violence could potentially be strengthened to prevent or reduce violence perpetration. Since some research has shown that social support plays a protective role in reducing IPV risk among women [61], strengthening social support mechanisms in refugee camps where social cohesion is low may also be a critical IPV prevention strategy. For example, this could include strengthening women’s social support networks through safe spaces where women can rebuild their networks and access trusted information and services [6]. Approaching social cohesion as it relates to GBV via multiple pathways from basic services, financial independence, economic integration, language training, and the built environment may also improve effectiveness of GBV programming in refugee camps [62].

Conclusion

In summary, this study reports that high P-NSC among men is associated with reduced perpetration of physical IPV in Bokolmayo refugee camp in Ethiopia. The association between P-NSC and women’s experiences of IPV needs to be further theorized and researched. Further research is needed to improve measurement of social cohesion in humanitarian contexts and to understand community-level drivers of IPV (including P-NSC) and how they interact with other factors to inform IPV programming. Humanitarian and other organizations responsible for camp administration and programing should consider and address neighborhood factors in GBV risk mitigation efforts, and could consider programming focused on promoting social cohesion that appropriately considers gender and GBV [28, 56, 57]. Understanding how to better foster constructive neighborhoods in humanitarian contexts with displaced populations could reduce IPV structurally, promote protective social processes to create and maintain beneficial neighborhood environments and foster other positive health outcomes.