Domestic Violence and Humanitarian Crises: Evidence from the 2014 Israeli Military Operation in Gaza

Abstract

Using qualitative data from 21 group discussions and unique survey data from a representative cross-section of 439 women in the Gaza Strip, we investigate how the Israeli military operation “Protective Edge” in 2014 influenced domestic violence (DV), accounting for risk factors at different levels of the ecological model. We combine our survey data with secondary data on infrastructure destruction across Gaza’s neighborhoods, and use propensity score matching techniques to address endogeneity concerns. Our results show that the military operation increased DV, and that this effect manifests itself at relatively low-levels of destruction. Our analysis suggests that the mechanisms are displacement, a lowered ability of married women to contribute to household decision-making, and reduced social support networks.

Keywords

domestic violence conflict complex emergencies

Introduction

There is a growing recognition that prolonged exposure to violent conflict adds to the daily stressors faced by individuals and risks triggering or aggravating violence within the family structure (Annan & Brier, 2010; Horn, 2010; Human Security Research Group, 2012; Llosa et al., 2012; Wirtz et al., 2014). Family relations can become tense following conflict-related changes in gender roles, physical insecurity, and economic uncertainty (Al-Krenawi, Graham, & Sehwail, 2007), and violence may be seen as a “normal” tool when facing disputes and conflict (Jewkes, 2002). Furthermore, violent conflicts tend to generate forced displacement of populations, separation of families, and disruption of community and institutional protection structures, which limit access to justice for survivors and thus render refugees and internally displaced persons particularly vulnerable to violence (United Nations High Commissioner for Refugees [UNHCR], 2003).

Yet, there are at least three gaps in our knowledge of the links between violent conflict and domestic violence (DV). First, only a very small body of literature examines the causal relationship between different forms of violent conflict and DV. Existing studies, such as La Mattina (2017) on Rwanda, Gallegos and Gutierrez (2016) on Peru, and Noe and Rieckmann (2013) on Colombia, find a significant relationship between civil war and genocide on one hand, and DV—particularly intimate partner violence (IPV)—on the other hand. However, these studies face a number of challenges: (a) assumptions have to be made about individual or household experiences of conflict as conflict data are only available at province or district levels, (b) information on DV and IPV pre-conflict is missing, and (c) information to investigate potential mechanisms through which conflict affects the likelihood of DV within the household is missing. Second, due to insecurity and safety issues for respondents and researchers, data used to inform programs and interventions in contexts of complex emergencies¹ resulting from conflicts are largely drawn from non-probability samples—which often include service-based data that inflate perpetration rates by non-partners (Global Women’s Institute and International Rescue Committee, 2016) or focus only on specific sections of the population, for example, refugees and displaced people in camps or towns (Stark & Ager, 2011), who are often more easily accessible. Third, where studies document prevalence rates in complex emergencies, they mostly focus on physical or sexual intimate partner violence, and sexual non-partner violence—as do most of the studies on DV. Yet, though IPV is the most prevalent type of DV, non-IPV in domestic settings is not negligible. For example, a recent survey of DV in Ghana showed that parents and siblings play a significant part in perpetrating all forms of psychological, social, economic, physical, and sexual violence, both when women are married and particularly when they are unmarried and living in their parents’ homes (IDS, GSS, & Associates, 2016). Other research in India, Jordan, Palestine and other countries in the Middle East and North Africa also show that a limited focus on IPV neglects the experiences of younger, never-married women who might be subject to physical or nonphysical violence by fathers, brothers, mothers, sisters, or other family members (Assaf & Chaban, 2013; Baxter, 2007; Kulczycki & Windle, 2011), or married women who suffer at the hands of both intimate partners and non-intimate family members (Fernandez, 1997; Clark, Silverman, Shahrouri, Everson-Rose, & Groce, 2010).

This article aims to address these gaps and contribute to our knowledge of the links between violent conflict and complex emergencies, on one hand, and DV, on the other hand, by studying the Israeli military operation “Protective Edge” in the Gaza Strip (hereafter referred to as Gaza) in July and August 2014. We exploit a unique dataset collected by the authors on a representative sample of 439 married and non-married women, which documents changes in livelihoods and places of residence; experiences of DV before, during, and after the Israeli military operation “Protective Edge”; and information on risk factors associated with DV. The quantitative data are complemented by secondary data on neighborhood destruction due to the military operation, and primary qualitative data, which allow us to investigate the perceptions about causes and drivers of DV, and channels through which DV could be affected by the ongoing conflict.

Using these data allows us to analyze the impact of violent conflict on DV types, prevalence rates, and drivers for a representative sample of women in the context of a complex emergency. It is important to stress that due to the long-standing nature of the political violence in Gaza, the quantitative data analysis measures the impact of the intensification of the conflict and not the impact of the ongoing conflict.

The article is structured as follows: The section “Context” gives a brief introduction to the context of our study and the challenges of existing studies. The section “Data” describes the data, and the section “DV and Its Drivers in Gaza—Qualitative and Descriptive Results” uses qualitative insights to describe women’s experiences of DV in Gaza. The empirical strategy is explained in the section “Empirical Methods,” and the section “Empirical Results” describes the results. A final discussion with implications for research and policy is offered in the section “Summary and Discussion.”

Context

Gaza experiences a prolonged humanitarian crisis due to the ongoing land, air, and sea blockade, which severely restricts movement of goods and people, and repeated outbreaks of acute violence by Israeli military operations. Overall, 70% of Gaza’s population are refugees (and descendants of refugees) from the Arab–Israeli conflict of 1948. After years of Egyptian military rule over the Gaza Strip during which the movement of most of its residents was already restricted, Israel re-occupied the Gaza Strip in 1967 during the war among Israel, Syria, Jordan, and Egypt. After 20 years of occupation and growing numbers of Israeli settlements in Gaza (and the West Bank), the First Intifada (1988-2000) broke out with devastating consequences for Gaza’s population, infrastructure, and agriculture. Between 1990 and 1991 alone, the number of families receiving food assistance increased from 9,383 to 120,000.² During the Second Intifada (2000-2005), around 15% of the remaining agricultural land was destroyed, and two military operations in 2006 cost the agricultural sector around US$35 million (Alexander, 2007). Although Israel disengaged from Gaza in 2005, it continued to severely restrict the movement of people and goods. Gaza has seen three large military operations since 2008: Operation “Cast Lead” (the “Gaza War”) in 2008-2009, Operation “Pillar of Defense” in November 2012, and Operation “Protective Edge” in summer 2014. Both violent outbreaks and the continuing socio-economic pressures on families and individuals exert profound negative impacts on people’s well-being and social relations. The World Bank (2015) estimates that the real per capita income in Gaza in 2010 was 31% lower than it was 20 years ago, and that losses due to the blockade imposed in 2007 represent more than 50% of the gross domestic product (GDP). The military operation in summer 2014 alone is estimated to have led to a 15% contraction of Gaza’s GDP. Gaza’s unemployment rate reached 43% toward the end of 2014—“probably the highest rate in the world” (World Bank, 2015, p. 5)—with a youth unemployment rate of about 60%. In total, 1.4 million people in Gaza were targeted with humanitarian assistance in 2018, out of an estimated 1.9 million inhabitants in 2017 (Palestinian Central Bureau of Statistics [PCBS], 2018; United Nations Office for the Coordination of Humanitarian Affairs [UNOCHA], 2017). The ongoing political and economic tensions exert enormous pressure on individual, family, community and gender relations, resulting in a “crisis of masculinity and femininity,” where “a man is valueless and unable to secure his family, and a woman is humiliated by being forced to leave her children and her respected position as domesticated wife and mother, to act as a beggar for humanitarian aid in the public sphere” (Muhanna, 2010, p. 41). It has also resulted in decreased levels of freedom for women due to the militarization of the Second Intifada and growing Islamization of the occupied Palestinian territories (Assaf & Chaban, 2013; Shalhoub-Kevorkian, 2009).

A small number of studies in Palestine have investigated the effect of violence related to the Israeli occupation on violence against women and associated risk factors. For example, UN Women and the Institute of Women’s Studies (2014) found a link between the Second Intifada and increased emotional, physical, and sexual violence against women. Clark, Everson-Rose, et al. (2010) found a strong correlation between economic consequences of the ongoing conflict and IPV in Gaza: Women whose households were affected by political violence were 139% more likely to report psychological IPV. Müller and Barhoum (2015) suggest that the military operation of 2014 restricted the ability of men to provide for their family, resulting in stress and frustration that were eventually vented in the form of violence against their wives. This is a similar finding as in Muhanna’s (2010) qualitative study in Gaza, where most women linked men’s violence toward women (and their children) with males’ feelings of helplessness and “loss of their sense of their manhood” (p. 49). The Institute of Women Studies (IWS, 2008) has found that the loss of jobs by husbands and the deterioration of households’ economic situations are related to higher levels of psychological, physical, and sexual IPV. However, all these studies are essentially descriptive and do not establish a causal relationship between conflict-related violence and DV. Indeed, it may be that women most exposed to DV belong to households that are also most exposed to the conflict. Without an exogenous source of variation in political violence, it is difficult to ascertain whether the association between conflict and IPV is causal or not (e.g., if both types of violence are driven by similar processes). Another limitation of existing studies is the lack of information on intra-household dynamics and other violence risk factors.

We overcome the first challenge—the lack of evidence to establish a causal relationship between conflict-related violence and DV—by combining our original data with secondary data on the extent of destruction at the neighborhood level to estimate the effect of conflict violence on DV through propensity score matching (PSM). Indeed, conditional on household characteristics, we show that the extent of damage at the neighborhood level is not correlated with previous levels of DV. In relation to the second challenge—the lack of information on intra-household dynamics and other violence risk factors—we are able to include information on women’s involvement in decision-making, their social support networks, and their perceptions of gender roles and norms. We further supplement our quantitative analysis with a rich set of qualitative insights from 229 women and men in 21 focus group discussions (FGDs). The following section describes these data in more detail.

Data

In 2015, one of the authors led a study aimed at measuring the prevalence of different types of violence against women in Gaza and understanding the underlying drivers of DV, support mechanisms, and the relationship between DV and the military operation. A mixed-methods approach, where findings from qualitative and quantitative data could complement each other to form a complete picture (Bryman, 2012), was employed. The sequencing was designed in a way that also allowed a small number of FGDs to inform the instrument development of the quantitative component. Overall, qualitative FGDs and interviews with more than 300 women and 130 men, 28 members of local organizations, and seven key informants were carried out to better understand channels and dynamics related to violence against women in Gaza. The quantitative survey was designed as a two-stage cluster sample. In the first stage, neighborhoods were selected with probability proportional to size, and households were randomly chosen in the second stage. Within each household, one woman aged 17 years or older was randomly chosen using the Kish Grid method³ (see Müller & Barhoum, 2015, for more information on the survey design and sampling). The national survey was completed by 439 women and is representative of women across Gaza aged 17 years and older.

For this article, we use information from the DV module of the questionnaire. This module captures IPV committed by husbands of married women and violence by parents, siblings, or other people living in the same household as unmarried women. Respondents were asked multiple behaviorally based questions about violence, that is, questions about concrete acts such as “has your husband twisted your arm” or “has someone in your household called you ugly, stupid, worthless or fat?” Such questions are less prone to interpretation bias than single, more general questions as to whether a woman had “been attacked” or “experienced violence” (see United Nations [UN], 2014, for a discussion of guidelines on producing statistics on violence against women). Both married and unmarried women answered the same questionnaire; thus, the perpetrators of the acts could differ, but the acts and types of DV are comparable. The choice of acts included in our survey was informed largely by the Violence Survey implemented by the PCBS in 2011 (see Appendix D in Müller & Barhoum, 2015, for the questions in the module). In addition, we used findings from six qualitative pilot FGDs preceding the survey to inform the violence module. This pilot also served as a training exercise and was used to test and improve the qualitative instruments.

The qualitative data for the analysis in this article were collected through 21 qualitative FGDs, with 138 women and 91 men across Gaza during the main phase of data collection in May and June 2015. Both the quantitative and qualitative data were collected in Arabic by 14 social workers with extensive experience in conducting data collection for social science research and who were specifically trained on the project’s survey tools in May 2015. Local organizations helped arrange the FGDs, which usually lasted about 1.5-2 hr and were carried out with 8-14 participants each. One fieldworker would facilitate the FGD and another fieldworker would take notes. The transcripts were translated by one of the authors of Müller and Barhoum (2015) between May and July 2015. The quantitative responses from the survey questionnaire were entered into a specifically designed database in early June, immediately after the data collection at the end of May (11 months after the end of the military operation). The quantitative data collection too was preceded by a small number of pilot interviews (30). Survey interviews generally took about 1 hr. The survey data were analyzed using the statistical analysis software Stata 14, and the qualitative transcripts were coded and analyzed with the qualitative data analysis software NVivo 10.

For the quantitative analysis, we matched information on the residence of the survey respondents before the military operation in 2014 with the “Gaza Crisis Atlas” (UNOCHA, 2014). This document gives a detailed damage assessment from Israeli bombardments at the neighborhood level across Gaza. “The Gaza Platform,” an online mapping tool developed in a collaboration between Amnesty International and Forensic Architecture, shows that more than 1,200 attacks on residential homes alone were made during summer 2014 (Amnesty International, 2015). For each neighborhood, it records the number of moderately damaged, severely damaged, and destroyed structures. Using this information allows us to estimate the impact of the intensification of the conflict on DV through PSM, described in the section “Empirical Methods.”

DV and Its Drivers in Gaza—Qualitative and Descriptive Results

According to our survey, 39.6% of respondents reported having experienced at least one act of DV since the summer of 2014. Psychological violence, in the form of cursing, insults, yelling, and screaming was the most commonly reported form of DV (34.1%). The second most reported category was economic abuse (18.9%): Women reported being refused sufficient funds for daily expenses even if funds were available, and being threatened with withdrawal of financial support and/or the control of their expenses. These were followed in importance by physical abuse (14.3%), controlling behavior (13.9%), and sexual abuse (7.5%).

Existing studies have uncovered a range of risk factors for DV. The ecological model of violence against women (Heise, 1998, 2011; Jewkes, 2002) posits that at the individual and household levels, characteristics such as age of respondent and partner, education levels, socio-economic status of the household, and household size can play a significant role in determining risks (Aizer, 2011; Bobonis, González-Brenes, & Castro, 2013; Clark, Everson-Rose, et al., 2010; García-Moreno, Jansen, Ellsberg, Heise, & Watts, 2005; Jewkes, 2002; Jewkes, Levin, & Penn-Kekana, 2002). At the level of family and relationship, cross-cultural studies have cited male control of wealth and decision-making within the family and marital conflict as determinants of DV. Furthermore, women’s access to (social) support, gender roles, tolerance of physical punishment of women and children, and the wider acceptance of violence as a conflict-solving mechanism have been shown to play a significant role (Heise, 2011; Heise, Ellsberg, & Gottmoeller, 2002).

Analyzing the themes that emerged from qualitative FGDs with women and men shows that the perceived causes of violence discussed by participants correspond well to each of the levels of the ecological model mentioned above. Purported causes were broadly related to personal characteristics, interpersonal relationship dynamics, and structural factors. With respect to personal characteristics, the mentioned “causes” of violence against women were mainly as follows: youth and lack of maturity, the influence of friends, psychological pressure, lack of responsibility, bad upbringing, psychological “troubles” of the perpetrator, and women’s behavior. A lack of mutual understanding or trust, controlling behavior, intergenerational transmission of violence, bad communication, jealousy, and rushed marriages were mentioned on the interpersonal level. Structural factors included scarce job opportunities and unemployment, “bad” traditions (including social norms), lack of respect toward women, stigma, the Israeli siege, the government, illiteracy and a lack of education, and gender roles and norms. It is worth noting that although most of these issues were identified as important contributors to violence in several discussions, the issues of “bad tradition” and social norms, and poverty and unemployment came up in every FGD.

In relation to the above-mentioned factors, the quantitative survey uncovered pervasive patriarchal gender norms and relations between the sexes. For example, only 42% of women disagreed with the notion that “a woman’s purpose in life is only fulfilled if she gives birth to a son”; only 33.9% disagreed with the statement that “the man should have the final say in the home”; and only 32% disagreed that a “woman’s most important role is to take care of her home and cook for the family.” However, there were strong feelings about the unacceptability of violence: 77.8% of respondents disagreed that it would be “okay for a man to beat his wife if she refuses to serve him” and 79.8% disagreed that “an occasional slap does not amount to DV.” Yet, 16% of women were indecisive, and 26% agreed (strongly) with the notion that it is okay if “a man beats his wife because he is angry of the political situation”; and respondents were almost completely divided over the issue of accepting violence to keep the family together (45.5% disagreed, 36.8% agreed, and the remaining 17.7% were undecided).

Some interesting themes emerged during group discussions that revealed several layers and interconnections between the different levels of the ecological model such as: discussions about age, needs, and vulnerabilities; solutions to prevent or reduce violence; and the use of services. For example, there was no general consensus on whether girls and young women are exposed to more or less violence at certain ages; however, study participants agreed that the types of violence and the “reasons” of the perpetrators could change, as reasons were often related to life experience and needs. One critical juncture in the life cycle of individuals is marriage, when—as framed by Assaf and Chaban (2013, p. 427)—the authority of women’s families is “transferred to the authority of their husbands’ home.” Indeed, violence against younger girls was often portrayed as corrective or protective, whereas violence against older, married women—who supposedly “know better”—was framed as a punishment. Furthermore, most participants explained that at younger ages, women and girls lack awareness and understanding of the needs and behavior of men, which could result in violence if they “make mistakes” or because they “are immature and can be taken advantage of” (FGD with men, aged 18-30 years, on May 23, 2015). However, from the age of 18 years onwards—and particularly in marriage—women have “enough experience to bear the responsibilities” (Assaf & Chaban, 2013). Some men also suggested that women should “obey,” “be patient toward her husband’s behaviors and never make mistakes,” and “avoid making the man angry” to reduce the risk of violence (FGDs with men, aged 18-33 years, on June 21, 2015; men, aged 18-30 years, on June 1, 2015; and men, aged 20-32 years, on May 24, 2015). Some women themselves advised to “try to avoid fights with husbands when they’re angry,” to show “patience so that conflicts don’t get bigger,” or to “act like she didn’t hear” when being verbally abused (FGDs with women, aged 32-60 years, on May 23, 2015; women, aged 23-52 years, on May 23, 2015; and women, aged 17-24 years, on May 21, 2015).

The fact that the perpetrators of DV change with age could point toward the role of intra-household bargaining or power dynamics, as this typically changes with age and a woman’s status or role in the household. The quantitative survey provides information on decision-making power within the household in five domains that serve as a proxy for these kinds of dynamics. Examining the average number of decisions women are involved in across different age groups, it seems that women do “learn to negotiate” better with age as women below 29 years are least likely to be involved in decisions and women between 50 and 59 years are most likely to be involved in all decisions.

Discussions on the availability and use of outside support again reflected gender norms and expectations toward women. In several FGDs, participants agreed that it was okay for women to seek outside help, but only if their family members did not help. Mukhtars—responsible for restoring justice and solving disputes between members of families and communities—were mentioned as acceptable mediators, but most mukhtars are males with the same traditional values as their “constituents.” Furthermore, they work within the boundaries dictated by the values of their “clientele” and “always favor the man in conflicts” (FGD with men, aged 24-27 years, on May 22, 2015). Many women explained that “solutions” offered by different actors did not differ much and that in most cases women go back to the home of their husbands (Müller & Barhoum, 2015). Usually, getting the police involved would be entirely unacceptable, and if women seek help from legal associations “mostly they will get divorced” (FGD with men, aged 19-51 years, on May 21, 2015). Our survey data show 78.9% of women agreed with the statement that “family problems should not be discussed outside the home”—48.4% of them strongly.

Under these circumstances, many women chose not to say anything or to keep the matter as closely contained within the family as possible. Overall, 28.1% of survivors of DV in our survey did not tell anyone about it, and the women who did talk chose to speak with parents or relatives (38.5%), left to stay with their family (34.8%), or tried to solve the issue themselves by talking to the perpetrator (38.5%) or refusing to talk to the perpetrator for several days (50.3%). However, some women do seek outside help from friends, colleagues, or neighbors (26.7%). In the quantitative analysis, we will indicate the degree of potential support by including information on the number of good friends the respondent has in the neighborhood (about 2.5 on average) and the number of social events attended in the week preceding the survey (about 2 on average).

Finally, we discussed the relationship between the ongoing conflict with Israel and violence against women with all study participants. With respect to the 2014 military operation, all focus group participants spoke about the stress of the destruction, overcrowding, lack of services, loss of livelihood and income, and constant fear. According to the participants, many witnessed fights over resources, verbal abuse, and divorces, particularly among displaced couples; and most focus group participants felt that violence against women increased during the war. Women discussed at length the anger of brothers and husbands during and after the military operation, even though some also spoke about how families stuck together. Our survey sample indicates that the prevalence of DV was lower during the conflict (32.5% of respondents reported DV during this period) compared with the 11 months that followed the end of the military operation (39.5%).

Many focus group participants felt that the underlying causes of DV were chronically present so that DV would occur even in the absence of the military operation. However, the ongoing Israeli–Palestinian conflict and its economic implications were consistently perceived across all focus groups as contributing and aggravating factors that lead to DV. Our data are consistent with both narratives: Whereas 7.6% of women in our survey only experienced DV since the outbreak of the 2014 military operation and 2.7% experienced DV exclusively during the military operation, most women who responded to the survey and who identified as survivors of DV since the end of the military operation had previously experienced DV.

Empirical Methods

Following the description of the qualitative findings and descriptive statistics, we now turn toward the empirical analysis of the impact of the military operation of 2014 and DV in Gaza in 2015. We conduct two sets of estimations:

We look at drivers of DV since summer 2014, taking into account different factors from the ecological model discussed at the beginning of section “DV and Its Drivers in Gaza—Qualitative and Descriptive Results.” Formally, this estimation is written as

D_{i} = β_{0} + β_{1} + β_{2} P_{i} + β_{3} H H_{i} + β_{4} S o c_{i} + ε_{i}

(1)

where D_i denotes respondent i’s exposure to DV since the end of the military operation in August 2014. D_i takes the form either of (a) a binary variable equaling 1 if she has experienced any type of DV and 0 if she has not, or (b) a discrete variable that equals the number of different types or acts of physical, sexual, psychological, or economic violence or controlling behavior the respondent has been exposed to (referred to later as “multiplicity”). In further specifications, D_i will also correspond to specific types of psychological, physical, sexual, or economic violence, as well as domestic control. P_i denotes personal and household characteristics, such as age, education, marital status, whether the head of household is working, and so on. HH_i is an indicator of the respondent’s decision-making power within the household. It ranges from 0-5, counting the number of decisions a woman has a say in, that is, the higher the index, the more decisions she is involved in. Soc_i indicates the degree of potential social support proxied by the number of good friends in the neighborhood and the number of social events attended in the last week. Finally, Att_i is an index of the respondent’s acceptance of violence against women and of her perceptions of the patriarchal values discussed in the section “DV and Its Drivers in Gaza—Qualitative and Descriptive Results.” The respondent was asked 10 questions that could be answered on a 5-point scale ranging from 0-4. Higher scores indicate higher levels of violence acceptance and agreement with patriarchal values.

We investigate the impact of the military operation in summer 2014 on subsequent experiences of DV. For that purpose, we use information on whether the respondent and her family members experienced high levels of destruction in their neighborhood and their displacement status during and after the military operation, and we relate this information to their experiences of DV. However, we recognize that the scale of damage in the neighborhood and displacement status are likely correlated with drivers of DV, such as the household’s location and socio-economic characteristics. Similarly, the extent of support available from neighbors, friends, and family could explain both the likelihood of displacement and DV. Thus, any simple statistical association between exposure to conflict and DV would not equate the true estimate of the impact of the war on the experience of DV.

We address this potential endogeneity bias using PSM, a popular quasi-experimental technique to estimate causal effects in nonexperimental settings. The idea is still to compare rates of DV between areas with different levels of destruction, but the matching method ensures that we only compare women with otherwise similar characteristics. Based on a set of pre-2014 military operation variables, the PSM estimator computes for all women the probability of living in areas affected by high levels of destruction (in technical terms this would be the “treatment group”). The estimator then finds women who did actually live in highly affected areas and matches them with women who did not live in highly affected areas but have a similar predicted probability of living in highly affected areas (in other words, who otherwise share similar individual and household characteristics with women in the treatment group).

We examined different cutoff points to classify areas as “highly affected by the 2014 military operation.” Specifically, we considered an area to be highly affected if it belongs to the upper 30, 40, 50, 60, and 70% of distribution of the extent of destruction, as measured by the “Gaza Crisis Atlas.” We expected that the impact of the 2014 military operations would be higher when considering stricter thresholds for classifying an area as “highly affected” (i.e., the upper 30 or 40% of the distribution) than when considering looser thresholds (i.e., the upper 60 or 70% of the distribution). The rationale was that, in the former case, we would use a more intense definition of the “treatment” (only the most affected areas would be part of the treatment group), leading to stronger impact of the treatment.

In fact, we found the reverse to be true. Our empirical analyses reveal that the 2014 military operation exerts a statistically significant impact on women’s risk of experiencing DV when we consider both the upper 60 and 70% of the distribution to define treatment status but not when we consider the upper 30, 40, or 50% of the distribution. This indicates that the effects of the war on DV “kick-in” at relatively low levels of destruction, and that any increased extent of destruction beyond the 60th percentile of the distribution does not generate additional effect on DV.

In what follows, we report in detail the results when we use the threshold of the upper 60% of the distribution to define areas as highly affected (a minimum of 32 destroyed structures per neighborhood). The results are qualitatively unchanged when we use the alternative upper 70% (a minimum of 21), and the estimated impact of the 2014 military operations is not different from zero when we use the 50th, 60th, and 70th percentile of the distribution as alternative thresholds. All these results are available upon request.

The quantity of interest we attempt to estimate with the PSM estimator is the average treatment effect on the treated (ATT), which is defined as the effect of the 2014 military operation on DV experiences for those women who have been living in highly affected areas. Formally, it is defined as the mean difference in prevalence of DV between “treated women” (those living in highly affected areas) and “untreated” women (those living in other areas) with the same predicted probability of being treated.

It is important that the level of destruction (the “treatment variable”) is not correlated with the likelihood of DV in the household before the 2014 military operation for the PSM estimate to be credibly causal. Indeed, our argument throughout this analysis is that the extent of destruction through the bombardment is unrelated to the women’s likelihood of suffering DV before the military operation. If there was a systematic association between the likelihood of living in a highly affected area and prevalence rates of DV before the military operation, then our argument would not be tenable. Figure 1 shows the results of a multivariate estimation of the likelihood of living in neighborhoods with high levels of destruction. It shows that the risk of experiencing DV before the military operation is not correlated with the level of destruction in the neighborhood (the confidence interval of the DV variable crosses the vertical line indicating that the estimated coefficient is not statistically different from zero). The figure also shows that based on household characteristics before the military operation, the likelihood of living in a neighborhood with high levels of destruction is rather “random” with respect to individual characteristics, such as age, marital status, or education, and incidence only significantly increases with respect to household size and agricultural livelihoods (their confidence intervals do not cross the vertical line). The former—household size—is a particularly good example of why we use PSM. Because research shows that household size is positively related to levels of DV, simply comparing means between displaced and non-displaced women, or women in neighborhoods with high levels of destruction to those in neighborhoods with less destruction, would result in an overestimation of the effect of the military operation. The PSM strategy ensures that we control for these and other confounding variables before the military operation.

Panel A in Table 1 shows some household and respondent characteristics that may have been affected by the 2014 military operation, and which could also have an impact on the likelihood of experiencing DV. These are part of the estimation in Equation 1. None of these variables are significantly different between neighborhoods with lower or higher levels of destruction.

Table 1.

Comparison of Pre and Post Characteristics of Respondents Living in Neighborhoods With Higher Levels of Destruction and Those in Areas Less Affected.

	High levels of destruction	Lower levels of destruction	Difference
Panel A: Characteristics that could have been affected by the military operation
Head of household working	0.57	0.65	0.08
Role in decision-making	2.89	3.21	0.32
Number of friends in the neighborhood	2.52	2.39	−0.12
Number of social events in last week	1.99	2.08	0.09
Gender roles	17.22	16.99	−0.23
Panel B: Characteristics of respondents and households before the military operation
Age of respondent	35.02	35.45	0.43
Highest level of education	4.03	4.07	0.05
Respondent is married	0.73	0.73	0.01
Number of household members	6.24	5.36	−0.88***
UNRWA food basket recipient household	0.31	0.24	0.07
Agriculture	0.11	0.04	−0.07**
Family business	0.09	0.08	0.01
Government employee	0.38	0.36	−0.02
Private sector	0.21	0.21	−0.00
UNRWA employee	0.09	0.04	−0.05*
Cash for work	0.07	0.03	−0.04
Refugee status	0.66	0.58	−0.08
Observations	289	150	439

Source. Gaza VAW survey 2015; own computation based on 439 observations.

Note. Cutoff point for “high” levels of destruction is the 40th percentile (32 or more buildings were completely destroyed). UNRWA = United Nations Relief and Works Agency.

p < .1. **p < .05. ***p < .01.

Panel B in the table shows summary statistics for a number of key respondent and household variables, which could potentially influence the likelihood of living in more affected neighborhoods and of being displaced (as used in the multivariate regression underlying Figure 1). As can be seen, comparing respondents from neighborhoods with higher and lower levels of destruction—which is highly correlated with the level of displacement—does not show many significant differences between groups. Similar to what we observed in Figure 1, women from households who lived in highly affected neighborhoods are significantly more likely to have depended on agriculture as their main source of income before the military operation (11 vs. 4%) and are more likely to live in larger households (6.2 vs. 5.4 household members on average).

Empirical Results

Drivers of DV After the Military Operation in 2014 (Equation 1)

Table 2 shows six different estimations. Columns 1 and 2 show the relationship between DV and a set of explanatory variables reflecting risk factors at the individual, household, and relationship levels of the ecological model, as well as information on attitudes and beliefs toward gender roles and norms (see Equation 1) for all women in the sample. The dependent variables in columns 1 and 2 are the experience of any type of DV during the 11 months following the military operation in summer 2014 (column 1), and the multiplicity of DV, that is, the number of different types of domestic physical, sexual, psychological, economic, or social violence experienced during the same period (column 2), respectively.

Table 2.

Determinants of Domestic Violence After the 2014 Military Operation.

	All women		Married women		Non-married women
	(1)Any type of DV	(2)Multiplicity of DV	(3)Any type of DV	(4)Multiplicity of DV	(5)Any type of DV	(6)Multiplicity of DV
Age	0.99	−0.002	1.00	0.0007	0.97	−0.009
Age	(0.0079)	(0.0053)	(0.012)	(0.0077)	(0.018)	(0.0084)
Education	0.95	−0.047	0.95	−0.082	0.84	−0.006
Education	(0.093)	(0.066)	(0.11)	(0.076)	(0.15)	(0.11)
Head working	1.32	0.14	1.32	0.18	0.93	−0.031
Head working	(0.37)	(0.17)	(0.48)	(0.22)	(0.56)	(0.32)
Married	0.99	−0.15
Married	(0.26)	(0.17)
Household size	1.16***	0.083***	1.11	0.06**	1.37***	0.15***
Household size	(0.065)	(0.029)	(0.07)	(0.029)	(0.14)	(0.053)
Decision-making	0.93	−0.089*	0.92	−0.11*	1.04	0.016
Decision-making	(0.068)	(0.049)	(0.079)	(0.060)	(0.18)	(0.10)
Number of friends	0.85***	−0.034**	0.85***	−0.040	0.80*	−0.020
Number of friends	(0.054)	(0.016)	(0.051)	(0.028)	(0.10)	(0.018)
Social events	1.00	−0.024	1.02	0.010	0.88	−0.14
Social events	(0.055)	(0.028)	(0.03)	(0.031)	(0.12)	(0.083)
Gender roles	1.02	0.007	1.03	0.008	0.97	−0.011
Gender roles	(0.019)	(0.012)	(0.025)	(0.013)	(0.034)	(0.025)
Cash-for-work	0.85	0.36	0.37	−0.16	8.80**	1.74***
Cash-for-work	(0.59)	(0.41)	(0.26)	(0.28)	(8.09)	(0.62)
UNRWA food basket	0.60*	−0.37**	0.67	−0.42**	o.28**	−0.44
UNRWA food basket	(0.18)	(0.17)	(0.24)	(0.20)	(0.14)	(0.29)
N	439	439	320	320	119	119
R ²	.075	.091	.065	.091	.218	.242

Source. Gaza VAW survey 2015; own computation based on 439 observations.

Note. Exponentiated coefficients in columns 1, 3, and 5; standard errors in parentheses; all other sources of household main income as listed in Panel B of Table 1 were also controlled for, but results not shown here for sake of space (and lack of significance)—detailed results can be obtained upon request. UNRWA = United Nations Relief and Works Agency.

p < .1. **p < .05. ***p < .01.

Many of the individual and household characteristics, such as age, education level, marital status, or employment status of the head of household, do not play a significant role in determining the risk of DV. This indicates that DV is not restricted to any particular group in terms of socio-economic status, and also confirms FG discussants’ observations that the risk of DV does not change across ages, and that only the nature or the perpetrator of DV may change. Nevertheless, women in larger households are more likely to report having experienced DV: an increase in household size by one member is associated with an increased risk of DV by 16%. This relationship could be explained by increased numbers of household members “simply” creating more stress and potential interpersonal conflicts, which could increase the risk of violence (Flake, 2005; Straus, 2010); for example, higher dependency ratios increase economic stress when there are insufficient resources (Carlson, 1984).

Women’s decision-making power within the household and the extent of a social support network have protective effects. Being able to rely on an additional good friend in the neighborhood reduces the risk of experiencing any type of DV by 15%, and a one unit increase in women’s decision-making power index is associated with a decrease in the risk of DV, although the effect is not statistically significant. We find similar results for the multiplicity of DV, although the magnitude of all the effects tends to be small (i.e., less than 10% of the standard deviation).

Columns 3-6 show the results of a disaggregated analysis between married and unmarried women. They show that the first two columns in Table 2 were masking important differences between married and non-married women. For example, the relationship between household size and risk of DV completely disappears when restricting the sample to married women and is much larger and highly significant when turning to non-married women. One more household member increases the risk of experiencing DV by 37%. Given that this result is restricted to unmarried women, it is comparable to findings in the literature on violence against children, where increasing numbers of children (which are highly correlated with the number of household members in our study) are associated with higher levels of violent discipline (see, for example, Dietz, 2000). This interpretation would also be consistent with Muhanna’s (2010) findings that much of the violence by men in Gaza is directed against their children.

Furthermore, the relationship between the decision-making power of women and the risk of violence completely disappears in the case of non-married women. This is most likely due to the fact that most non-married women (74.8%) have never been married and live with their parents. Also divorced, separated, and widowed women (25.2% of non-married women) often live with family members who are in charge of the household as it is very uncommon for women in Gaza to live alone. Most of these unmarried women will have very little influence on decision-making; and with very little variation in levels of involvement within this group and its small sample size, it is not surprising to be unable to detect statistically significant differences.

The protective effect of social network is similar among married and non-married women and, although the size of the “effect” is slightly larger for non-married women, the statistical significance of the effect is much stronger for married women (1% level significance). Having more friends could be an indicator of having a more “liberal” partner who might also be less inclined to use violence. It could also be related to an observation made by Muhanna (2010) in Gaza: In the absence of men’s ability to provide income, women find themselves searching for livelihood outside of their household, which reduces their dependence on their husbands, which in turn protects them from their marital violence. It is possible that women with more friends are more successful in their search. Yount (2011) describes how women in Egypt often advise each other on how best to avoid spousal violence, resulting in “strategic conformity.” In the section “DV and Its Drivers in Gaza—Qualitative and Descriptive Results,” we discussed the advice often given to women of being patient and tolerant and not “provoking” men, indicating that strategic conformity might also be at play here.

Finally, Table 2 also shows the coefficients of two household variables related to the main source of income after the military operation, an indicator whether a household receives its main income through cash-for-work, and an indicator of whether the main source of income comes through the United Nations Relief and Works Agency (UNRWA) food basket.⁴ A household’s dependence on the food basket has a very large protective effect on the likelihood of women experiencing any type of DV, as well as on the likelihood of experiencing multiple types. This could be related to the way the food basket is “delivered” as part of the larger UNRWA Relief and Social Services program (RSSP), which also involves regular work with social workers, skills training, and capacity building, mainly for women, children, and youth.

The Israeli Military Operation and DV

Figure 2 shows the impact of the military operation on experiencing DV for all women in highly affected neighborhoods since the end of the military operation in August 2014 as measured through our PSM estimation. We show the overall probability of experiencing DV (“Any DV”), as well as the risks of experiencing at least one act of each single measured type of DV (physical, sexual, psychological, and economic violence, and control) separately. The graph on the left shows the estimated impact when women are matched based on their personal and household socio-economic characteristics. The graph on the right shows the estimated impact of the 2014 military operation obtained while also controlling for prior experiences of DV, that is, taking into account that experiences of DV might have already begun before the military operation. Indeed, DV before and after the military operation is highly correlated and therefore should be controlled for to not over- or under-estimate the impact of the bombardment. However, matching on this additional dimension generates technical challenges, particularly when smaller numbers of respondents are available, as we will see below when we discuss the results for unmarried women.

The impact of the military operation decreases when we take former DV experiences into account; nevertheless, the risks of psychological violence and controlling behavior are still statistically significantly elevated because of the military operation, both by 10%. Comparing results between married women and non-married women (Figure 3) shows that these averages are driven by the experiences of married women, who face increased psychological violence and marital control after the military operation (even when taking their prior experience of DV into account).

The risk of experiencing psychological violence after the military operation is statistically different from 0 for all women in Figure 2, but it only “shows” for married women when controlling for prior violence (see column 2 in Panels A and B in Table 3). This indicates that it is predominantly married women in highly affected neighborhoods who had not experienced DV before the bombardments, who experienced this type of violence afterwards. In addition to increased risks of experiencing psychological violence (by 13%), married women in highly affected neighborhoods also face a higher likelihood of experiencing marital control (by 12%).

Table 3.

Effect of High Levels of Neighborhood Destruction on DV for Married and Non-Married Women.

	(1)Any DV	(2)Psycho-logical DV	(3)Physical DV	(4)Sexual DV	(5)Economic DV	(6)Control^a
Panel A: Married women, no control for prior DV
ATT	1.12*	1.11	1.04	0.98	1.09*	1.14***
	(0.072)	(0.073)	(0.045)	(0.046)	(0.053)	(0.033)
Observations	320	320	320	320	320	320
Panel B: Married women, control for prior DV
ATT	1.09**	1.13***	1.05	1.03	1.01	1.12***
	(0.039)	(0.046)	(0.039)	(0.031)	(0.043)	(0.036)
Observations	320	320	320	320	320	320
Panel C: Unmarried women, no control for prior DV
ATT	1.18**	1.16**	1.05	1.04	1.12*	1.10
	(0.091)	(0.078)	(0.070)	(0.045)	(0.068)	(0.078)
Observations	119	119	119	119	119	119
Panel D: Unmarried women, control for prior DV
ATT	1.04	1.04	0.93	1.03	1.06	0.95
	(0.076)	(0.089)	(0.094)	(0.053)	(0.10)	(0.070)
Observations	119	119	119	119	119	119

Source. Gaza VAW survey 2015; own computation based on 439 observations.

Note. Exponentiated coefficients; standard errors in parentheses. ATT = average treatment effect on the treated.

For married women, “control” equals spousal control; for unmarried women, it equates to controlling behavior by any other household member.

p < .1. **p < .05. ***p < .01.

For non-married women, we note an increase of overall DV after the 2014 military operation (18% “any”), as well as an increase in psychological and economic violence in particular (see Panel C Table 3). However, when we control for prior experiences (Panel D Table 3), the effect sizes decrease so that the effects are no longer statistically significant. We do not know with certainty whether the effect sizes are indeed zero or whether we do not have enough statistical power to detect differences for the small subgroup of unmarried women. Indeed, as can be seen in Figure 3 and Table 3, the confidence intervals (in the graph) and standard errors (in the table) are much larger compared with those for the married women sample. It is also the case when matching unmarried women not only on socio-economic background (Panel C Table 3) but also on whether they experienced DV before the military operation (Panel D Table 3).

Similar to our strategy in Equation 1, we also looked at the impact of the military operation on the multiplicity of DV, that is, the likelihood of experiencing multiple types of DV (Figure 4). Taking into account the experiences of DV before the military operation, married women in highly affected neighborhoods face 0.3 more types of DV on average (corresponding to 25% of the standard deviation), 0.3 more acts of psychological violence (29% of the standard deviation), and 0.2 more acts of marital control (29% of the standard deviation). We do not observe statistically significant effects on the multiplicity of violence for unmarried women, which is in line with our previous discussion on the overall risk of experiencing DV.

Channels

To understand the mechanisms through which the 2014 military operation impacted on DV, we implemented the PSM estimation on a number of different potential drivers of DV as discussed in the ecological model, that is, investigated whether high levels of destruction had any direct impact on the involvement in decision-making of the respondent, on the number of friends she has in the neighborhood, her attitude toward violence and patriarchal norms, the perceived stability of the husband’s job (or that of other male household members if the respondent is unmarried), and the head of household’s level of stress or depression. We found no such direct effect to be statistically significant. However, we did find a very large—and expected—significant positive relationship between the level of neighborhood destruction and the likelihood of household displacement. Married women who lived in highly affected neighborhoods were 42% more likely to be displaced than married women in neighborhoods that were less affected by destruction, and the effect on unmarried women was a 52% increase in the incidence of displacement.

We then investigated whether there were any indirect effects of the military operation on drivers of DV, through the channel of displacement. Again, we used the PSM estimator; however, we matched women based on their probability of being displaced and not on their probability of living in a neighborhood with high levels of destruction. As can be seen in Table 4, displacement had a very large negative impact on married women’s involvement in household decision-making, and it significantly reduced their number of friends in the neighborhood. Both women’s control and social capital are known violence-reducing factors (and were found to protect women from DV in our first set of estimations [see Table 2]).

Table 4.

Effect of Displacement on DV Factors for Married and Non-Married Women (Indirect Effect of Destruction).

	(1)Stress husband	(2)Decision-making	(3)Number of friends	(4)Social events	(5)Gender roles	(6)Unstable work
Panel A: Married women
ATT	0.077	−0.77***	−0.76**	0.065	−0.66	−0.006
	(0.20)	(0.25)	(0.34)	(0.37)	(0.96)	(0.081)
Observations	320	320	318	320	320	277
Panel B: Unmarried women
ATT	0.11	−0.15	0.23	−0.19	1.50	0.022
	(0.31)	(0.43)	(0.57)	(0.62)	(1.78)	(0.13)
Observations	119	119	115	119	119	88

Source. Gaza VAW survey 2015; own computation based on 439 observations.

Note. Standard errors in parentheses. ATT = average treatment effect on the treated.

p < .1. **p < .05. ***p < .01.

Summary and Discussion

Using unique qualitative and quantitative data on DV in Gaza, this article adds to our understanding of DV against women in violent conflict and complex emergencies in a number of ways. We provided a causal estimate of the impact of the intensification of the conflict in the form of the military operation in 2014 on DV experiences using representative data from women aged 17 years and older across the whole Gaza Strip. We empirically investigated mechanisms through which this impact is made and complemented these findings with qualitative data from 21 FGDs with 229 women and men.

Three key observations stand out. First, DV against women and girls in the context of Gaza’s complex and ongoing emergency is widespread. Overall, 37.5% of respondents have experienced DV in the 12 months before, and 39.6% have experienced DV in the 11 months after the military operation. This has immediate implications for service providers, practitioners, and funders, who often focus their advocacy and funding against gender-based violence in humanitarian emergencies on violence perpetrated by strangers or non-domestic relatives (Stark & Ager, 2011). Screening for risk factors of domestic as well as non-domestic violence could help protect many women and girls in such settings.

Second, some of the risk factors of DV differ across the life cycle; for example, household size as a risk factor only has a significant impact on DV for unmarried women. Also, the protecting effect of having friends is more pronounced for married than for unmarried women. Most studies to date focus on IPV as the most frequent type of DV and rarely disaggregate their findings by age or life cycle stage. Understanding how risk factors differ across the life cycle could greatly improve how vulnerable populations are identified and targeted.

More research is also needed to understand the mechanisms through which these factors increase or decrease the risk of DV across different contexts. For example, our results suggest that some livelihoods are related to higher, and some others to lower, risks of DV. This could be because some livelihoods are more insecure and thus stressful than others, or because certain ways of combining aid delivery with social work such as the United Nations Population Fund (UNFPA) food basket helps dealing with stress or feelings of hopelessness, or because the role of women in gaining certain types of aid increases husbands’ “cooperation” (Muhanna, 2010). This relates to other research that has shown a link between economic hardship and DV (see, for example, Renzetti, 2009), which was also made in the qualitative part of this study. Findings from such research would be very important for the design of aid and development programs in humanitarian settings.

Third, we find that an intensification of the Israeli–Palestinian conflict—here, the military operation of summer 2014—had a significant negative effect on married women’s ability to participate in household decision-making and the size of their social network, mediated through household displacement. This in turn led to a significant increase in overall prevalence rates of DV, mostly on psychological violence and marital control. What is important to recognize is that only 10 out of 176 respondents were still personally displaced at the time of the survey. Yet, the impact of even temporary displacement on DV proved strong and lingering. This shows the importance of trying not to restrict data collection efforts to subsets of the population, for example, in camps, only. Furthermore, future studies in the field of gender and displacement should look at the effect of temporary, short-lived, displacement experiences.

Although the study recognizes that DV is not solely explained by the military operation, it does play a significant aggravating role. As such, renewed and continued efforts to resolve this ongoing conflict are necessary not only to reduce DV, but also for the overall improvement of the lives of the people of Gaza.

Footnotes

Acknowledgements

The authors would like to thank participants at the Annual International Conference of the Research Group on Development Economics 2017 in Göttingen, Germany, and the Jan Tinbergen European Peace Science Conference 2017 in Antwerp, Belgium, for their thoughtful comments and feedback. We would also like to thank seminar participants, colleagues, and peer reviewers at the Institute of Development Studies and two anonymous reviewers for their comments and suggestions on earlier drafts of this article.

Authors’ Note

The views expressed in this article, as well as all errors and omissions, are the responsibility of the authors alone.

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

Notes

Author Biographies

Catherine Müller is a trained economist with extensive skills in the econometric analysis of (large) datasets using quasi-experimental methods (e.g., instrumental variables, difference-in-difference, matching). As an applied researcher, she has gained an in-depth knowledge of both quantitative and qualitative research design and data analysis in different country contexts across Europe, Africa, the MENA region, South Asia, and Latin America. She is particularly interested in women’s economic empowerment and gender-based violence related issues, including in conflict and humanitarian settings. She has recently joined Social Development Direct as a senior technical specialist.

Jean-Pierre Tranchant is a research fellow at the Institute of Development Studies. He is an applied economist, specializing in household surveys and in the use of quasi-experimental and experimental methods, including as part of impact evaluations. His main research area is the relationships between human development outcomes and conflict at the micro- and meso-levels. This includes research on conflict and child development outcomes, on the impact of humanitarian aid, and on the vulnerability to modern slavery in contexts of displacement. He also has an interest in decentralized development and the role decentralization plays in shaping ethnic conflict.

References

Aizer

(2011). Poverty, violence, and health. The impact of domestic violence during pregnancy on newborn health. Journal of Human Resources, 46, 518-538.

Alexander

(2007, October). Conflict, economic closure and human security in Gaza (Briefing paper). London, UK: Oxford Research Group.

Al-Krenawi

Graham

J. R.

Sehwail

M. A.

(2007). Tomorrow’s players under occupation: An analysis of the association of political violence with psychological functioning and domestic violence, among Palestinian youth. American Journal of Orthopsychiatry, 77, 427-433.

Amnesty International. (2015, July 8). Launch of innovative digital tool to help expose patterns of Israeli violations in Gaza (News release). Retrieved from https://www.amnesty.org/en/latest/news/2015/07/launch-of-innovative-digital-tool-gaza/

Annan

Brier

(2010). The risk of return: Intimate partner violence in northern Uganda’s armed conflict. Social Science & Medicine, 70, 152-159.

Assaf

Chaban

(2013). Domestic violence against single, never-married women in the occupied Palestinian territory. Violence Against Women, 19, 422-441.

Baxter

(2007). Honor thy sister: Selfhood, gender, and agency in Palestinian culture. Anthropological Quarterly, 80, 737-775.

Bobonis

G. J.

González-Brenes

Castro

(2013). Public transfers and domestic violence: The roles of private information and spousal control. American Economic Journal: Economic Policy, 5, 179-205.

Bocco

(2009). UNRWA and the Palestinian refugees: A history within history. Refugee Survey Quarterly, 28, 229-252.

10.

Bryman

(2012). Social research methods (4th ed.). Oxford, UK: Oxford University Press.

11.

Carlson

B. E.

(1984). Children’s observations of interparental violence. In Roberts

A. R.

(Ed.), Battered women and their families (pp. 147-167). New York: Springer.

12.

Clark

C. J.

Everson-Rose

S. A.

Suglia

S. F.

Btoush

Alonso

Haj-Yahia

M. M.

(2010). Association between exposure to political violence and intimate partner violence in the occupied Palestinian territory: A cross-sectional study. Lancet, 375, 310-316.

13.

Clark

C. J.

Silverman

J. G.

Shahrouri

Everson-Rose

Groce

(2010). The role of the extended family in women’s risk of intimate partner violence in Jordan. Social Science & Medicine, 70, 144-151.

14.

Dietz

T. L.

(2000). Disciplining children: Characteristics associated with the use of corporal punishment. Child Abuse & Neglect, 24, 1529-1542.

15.

Fernandez

(1997). Domestic violence by extended family members in India: Interplay of gender and generation. Journal of Interpersonal Violence, 12, 433-455.

16.

Flake

D. F.

(2005). Individual, family, and community risk markers for domestic violence in Peru. Violence Against Women, 11, 353-373.

17.

Gallegos

J. V.

Gutierrez

(2016). The effect of civil conflict on domestic violence: The case of Peru (RAND Working Paper WR-1168). Santa Monica, CA: RAND Corporation.

18.

García-Moreno

Jansen

H. A. F. M.

Ellsberg

Heise

Watts

(2005). WHO multi-country study on women’s health and domestic violence against women: Initial results on prevalence, health outcomes and women’s responses. Geneva, Switzerland: WHO.

19.

Global Women’s Institute and International Rescue Committee. (2016). Evidence brief: What works to prevent and respond to violence against women and girls in conflict and humanitarian settings? Washington, DC: George Washington University, and London: IRC.

20.

Heise

Ellsberg

Gottmoeller

(2002). A global overview of gender-based violence. International Journal of Gynecology & Obstetrics, 78, S5-S14.

21.

Heise

L. L.

(1998). Violence against women: An integrated, ecological framework. Violence Against Women, 4, 262-290.

22.

Heise

L. L.

(2011). What works to prevent partner violence? An evidence overview (Working Paper). STRIVE Research Consortium, London School of Hygiene and Tropical Medicine. Retrieved from https://www.oecd.org/derec/49872444.pdf

23.

Horn

(2010). Responses to intimate partner violence in Kakuma refugee camp: Refugee interactions with agency systems. Social Science & Medicine, 70, 160-168.

24.

Human Security Research Group. (2012). Human security report 2012: Sexual violence, education, and war: Beyond the mainstream narrative. Vancouver, British Columbia, Canada: Simon Fraser University.

25.

IDS, GSS, & Associates. (2016). Domestic violence in Ghana: Incidence, attitudes, determinants and consequences (Project Report). Brighton: IDS.

26.

Institute of Women Studies. (2008). The impacts of Israeli mobility restrictions and violence on gender relations in Palestinian society: 2000-2007 (Background report for the World Bank). IWS, Birzeit University, Birzeit, West Bank, Palestine.

27.

Jewkes

(2002). Intimate partner violence: Causes and prevention. Lancet, 359, 1423-1429.

28.

Jewkes

Levin

Penn-Kekana

(2002). Risk factors for domestic violence: Findings from a South African cross-sectional study. Social Science & Medicine, 55, 1603-1617.

29.

Kulczycki

Windle

(2011). Honor killings in the Middle East and North Africa: A systematic review of the literature. Violence Against Women, 17, 1442-1464.

30.

La Mattina

. (2017). Civil conflict, domestic violence and intra-household bargaining in post-genocide Rwanda. Journal of Development Economics, 124, 168-198.

31.

Lewis-Beck

M. S.

Bryman

Futing Liao

(Eds.) (2004). The SAGE encyclopedia of social science research methods. Thousand Oaks, CA: Sage. doi:10.4135/9781412950589

32.

Llosa

A. E.

Casas

Thomas

Mairal

Grais

R. F.

Moro

M.-R.

(2012). Short and longer-term psychological consequences of operation cast lead: Documentation from a mental health program in the Gaza strip. Conflict and Health, 6, 8.

33.

Muhanna

(2010). Changing family and gender dynamics during the siege against Gaza: Spousal relations and domestic violence. Review of Women’s Studies, 6, 40-52.

34.

Müller

Barhoum

(2015). Violence against women in the Gaza Strip: After the Israeli military operation Protective Edge 2014 (Project report). Johannesburg: ActionAid and Madrid: Alianza por la Solidaridad.

35.

Noe

Rieckmann

(2013). Violent behaviour: The effect of civil conflict on domestic violence in Colombia (Poverty, equity and growth discussion paper). Göttingen, Germany: Courant Research Centre.

36.

Palestinian Central Bureau of Statistics. (2011). The main findings of violence survey in the Palestinian society (Technical report). Ramallah, Palestine: Author.

37.

Palestinian Central Bureau of Statistics. (2018). Population, housing and establishment census 2017: Census final results summary. Ramallah, Palestine: Author.

38.

Renzetti

C. M.

(2009). Economic stress and domestic violence. CRVAW Faculty Research Reports and Papers Paper, 1. Retrieved from https://uknowledge.uky.edu/cgi/viewcontent.cgi?article=1000&context=crvaw_reports

39.

Shalhoub-Kevorkian

(2009). Militarization and violence against women in conflict zones in the Middle East: A Palestinian case study. Cambridge, UK: Cambridge University Press.

40.

Stark

Ager

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

41.

United Nations. (2014). Guidelines for producing statistics on violence against women. New York, NY: Department of Economic and Social Affairs, Statistics Division, UN.

42.

United Nations High Commissioner for Refugees. (2003). Sexual and gender-based violence against refugees, returnees and internally displaced persons: Guidelines for prevention and response. Geneva, Switzerland.

43.

United Nations Office for the Coordination of Humanitarian Affairs. (2017, December). January-December 2018 Humanitarian Response Plan. Humanitarian Response Strategy.

44.

United Nations Office for the Coordination of Humanitarian Affairs (UNOCHA, August, 2014). Gaza crisis atlas. United Nations Office for the Coordination of Humanitarian Affairs occupied Palestinian territory. East Jerusalem.

45.

UN Women and Institute of Women’s Studies. (2014). Building ties. Towards integrated strategies policies for empowering Palestinian women. UN Women: Palestine country office.

46.

Wirtz

A. L.

Pham

Glass

Loochkartt

Kidane

Cuspoca

. . . Vu

(2014). Gender-based violence in conflict and displacement: Qualitative findings from displaced women in Colombia. Conflict and Health, 8, 10.

47.

World Bank. (2015). Economic monitoring report to the ad hoc liaison committee (vol. 2): Main report (English) (Working Paper 99646). Washington DC: World Bank Group.

48.

Yount

K. M.

(2011). Women’s conformity as resistance to intimate partner violence in Assiut, Egypt. Sex Roles, 64(1-2), 43-58.