Policy Research Working Paper 9821 Intimate Partner Violence and Household Decision Making Autonomy Effects of the Malian Conflict on Women Uche Eseosa Ekhator-Mobayode Jocelyn Kelly Amalia Rubin Diana J. Arango Gender Global Theme October 2021 Policy Research Working Paper 9821 Abstract Rates of intimate partner violence vary widely across but decreases women’s ability to decide how their earnings regions. Evidence suggests that some of this variation can are deployed. The results imply that to be successful, pro- be attributed to exposure to armed conflict. This study grams to mitigate these adverse effects of conflict on women exploits variation in the timing and location of conflict need to be context specific and rely on data-driven evidence events related to the war in Mali to examine the effect of from situations of conflict whenever possible. Policy makers conflict on intimate partner violence and some women’s are called to design programs that address harmful gender empowerment outcomes. The study used data from the norms and intimate partner violence at the individual/ Demographic and Health Survey spatially linked to con- household and community levels, especially for women flict data from the Armed Conflict Location and Events residing in areas with high-intensity conflict. Measurement Database. Wartime conflict increases the prevalence of of women’s empowerment should consistently include sev- women’s experiences of intimate partner violence. It also eral domains of women’s lives to gauge progress in voice increases women’s household decision making autonomy and agency, financial autonomy, and violence reduction. This paper is a product of the Gender Global Theme. It is part of a larger effort by the World Bank to provide open access to its research and make a contribution to development policy discussions around the world. Policy Research Working Papers are also posted on the Web at http://www.worldbank.org/prwp. The authors may be contacted at uekhator@worldbank.org, jkelly@hsph.harvard.edu, Amalia.h.rubin@gmail.com, and darango@worldbank.org. The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about development issues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished. The papers carry the names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the views of the International Bank for Reconstruction and Development/World Bank and its affiliated organizations, or those of the Executive Directors of the World Bank or the governments they represent. Produced by the Research Support Team Intimate Partner Violence and Household Decision Making Autonomy: Effects of the Malian Conflict on Women Uche Eseosa Ekhator-Mobayode[†], Jocelyn Kelly[‡], Amalia Rubin,[†] and Diana J. Arango[†] JEL classification: B54; J12; J16; O12 Keywords: Conflict, Forced Displacement, Violence against women and girls, Gender Based Violence, Intimate Partner Violence, Mali The authors of this paper conducted their research under Gender Dimensions of Forced Displacement project. The project is co-led by Lucia Hanmer and Diana Arango under the guidance of Hana Brixi, Global Director, Gender Unit, The World Bank Group. Funding: Building the Evidence on Protracted Forced Displacement: A Multi-Stakeholder Partnership''. The program is funded by UK aid from the United Kingdom's Foreign, Commonwealth and Development Office (FCDO), it is managed by the World Bank Group (WBG) and was established in partnership with the United Nations High Commissioner for Refugees (UNHCR). The scope of the program is to expand the global knowledge on forced displacement by funding quality research and disseminating results for the use of practitioners and policy makers. This work does not necessarily reflect the views of FCDO, the WBG or UNHCR. 1. Introduction 1 While IPV may be more common than sexual violence perpetrated by armed actors during war (Stark & Ager, 2011; Global Women’s Institute, 2019), it is often far less visible than conflict- related sexual violence (Stark & Ager, 2011; Hynes et al., 2004). For example, a survey of displaced women in northern Uganda found rates of past-year rape by partners was 41 percent compared to 5 percent by non-partners (Stark et al., 2010). Yet, violence faced by women and girls is a continuum, and there is often a link between experiencing multiple forms of violence in conflict settings. For example, in Uganda, Thailand, Côte d’Ivoire and Liberia, women who experience higher levels of conflict-related abuses from non-partners also report higher levels of IPV victimization during and after conflict (Saile et al., 2013; Falb et al., 2013; Gupta et al., 2012; Kelly 2018), and refugee women affected by the Burmese conflict who experienced conflict victimization were 5.9 times more likely to report past-year IPV than women who did not report conflict victimization (Falb et al., 2013). Exposure to conflict related violence in itself has also been linked to increased IPV risk as evidenced by recent analysis of the Liberia Civil War which found that living in a district that experienced fatalities during conflict increases the risk of experiencing multiple types of interpersonal violence in the post-conflict period. After adjusting for known individual-level correlates of IPV, residence in a fatality-affected district was significantly associated with a 50 percent increase in risk of abuse, even 5 years after conflict (Kelly, 2018). In Nigeria, the Boko Haram insurgency increased women’s risk of both experiencing of IPV and experiencing controlling behavior from their husbands and partners, after controlling for individual, partner, household and country specific factors, as well as comparing women with similar characteristics using a quasi-experimental analysis design (Ekhator- Mobayode et al. 2020). These findings suggest that the dynamics within the home that are risk factors for IPV may be exacerbated by armed conflict outside the home. To better understand the factors that underlie IPV in conflict, scholars have employed the ecological framework adapted by Heise (1998) to understand the interaction between conflict and IPV (Swaine et al. 2019). The adapted ecological framework for conflict and violence against women and girls (Swaine et al. 2019) suggests that exposure to conflict can impact risk factors for experiencing IPV at the individual, interpersonal, institutional, community and societal levels. At the interpersonal level, conflict can impact household stressors, controlling behavior, unequal decision making, division of labor, and men’s perception of their lack of ability to fulfill traditional masculine roles, among others (Swaine et al. 2019). These shifts may trigger changes in gendered roles within the household and increase risk factors also associated with increased perpetration of IPV (Ekhator-Mobayode et al. 2020; Falb et al, 2018). Studies that look at variation in the levels of conflict have indeed found that political conflict can lead to higher levels of IPV (Otsby, 2016; Kelly, 2018). Evidence from the Democratic Republic of Congo (DRC) suggests that work and income are central to masculine identities - the loss of either can pose significant risks with regards to men’s mental health and their desire to exert power within the household (Slegh et al. 2014). In DRC both men and women reported lack of work and income as a trigger for men’s use of violence. [†] World Bank, Washington, D.C, USA. [‡] Harvard Humanitarian Initiative. 2 This research contributes to the emerging literature examining the relationship between armed conflict and IPV, analyzing the effect of conflict in Mali on women’s experience of various forms of IPV and their agency within the household. We employ a quasi-experimental method as it is possible that unobserved characteristics affect both conflict and socioeconomic factors, making it difficult to isolate the effect of armed conflict using other approaches. The effect of armed conflict on IPV may be context specific. Therefore, evidence using quasi experimental methods where the data makes it possible to isolate the effect of armed conflict on IPV outcomes is important to strengthen the emerging empirical literature, and can inform evidence-based long and short-term policies to prioritize assistance in conflict affected settings. Since 2012, Mali has been faced with a complex political, security, and humanitarian crisis which began when groups of armed Tuaregs—a semi-nomadic ethnic minority in northern Mali—joined forces with extremist groups to declare the establishment of the independent state of “Azawad.” As the conflict progressed, jihadist groups took control of the rebellion and, after a successful coup d’état by national soldiers, international armed forces from France, neighboring countries, and the United Nations intervened. Despite an initial peace deal in June 2013 and a ceasefire in 2015, the conflict has continued (Lamarche, 2019). Before this unrest, Mali was recognized as a country with a peaceful democracy. Data gathered for conflict data sets support this reality. The Armed Conflict Location and Events Database (ACLED) shows an average of 16 conflict events reported between 2006 and 2010 compared to 278 events reported in 2011 alone (figure 1). The conflict also led to increased displacement, with over 227,000 newly forcibly displaced people in 2012 (Internal Displacement Monitoring Center, n.d). High levels of gender inequality pre-date the conflict in Mali. Mali ranked 158 of 162 countries in the most recent Gender Inequality Index (GII), little different from its rank of 143 of 146 countries in 2012 when the conflict began. According to the World Bank Gender Data portal, women’s economic participation has declined since 2000 in all industries except services, and women’s unemployment has increased. In contrast, women’s involvement in politics and decision making has increased since the war began. During the war (2012-2015), the number of seats held by women in parliament decreased. In 2020, political participation of women surpassed pre-war levels (World Bank, n.d). We use available data from the Demographic and Health Survey (DHS) in a period before and during Mali’s conflict and spatially link the DHS data to conflict data from ACLED. Trends in the ACLED data allow for the clear distinction between periods and locations of high conflict and relatively little or no conflict. This provides an opportunity to exploit the variation in the timing and location of wartime conflict in Mali and isolate the effect of wartime conflict on the study outcomes. This study adds to existing quantitative studies using quasi experimental methods to examine the effect of armed conflict on various human development outcomes (Ekhator- Mobayode and Asfaw 2019; Chukwuma and Ekhator-Mobayode 2019). It builds on Ekhator- Mobayode et al. (2020), which uses a similar method to isolate the effect of the Boko Haram (BH) insurgency in Nigeria on women’s experiences of IPV, which finds increased experiences of IPV in BH affected areas after controlling for individual, partner, household, and country specific characteristics. 3 Figure 1: ACLED Conflict Events in Mali, 2007 – 2018 Source: Authors own calculation using data from ACLED and Mali shape files from the Database of Global Administrative Areas – GADM. 2. Methodology 2.1 Data and sample construction The study sample is drawn from the 2006 and 2018 Mali Demographic and Health Surveys (DHS)- both of which are nationally representative household surveys that provides data on population, health, and nutrition for women aged 15-49 in Mali. Both surveys include information on the location of the interview and its GPS coordinates as well as a Domestic Violence (DV) module asking women about their experiences of IPV. The 2006 Mali DHS provides data for the period before protracted conflict in Mali (referred to as “peacetime” hereafter). The 2018 Mali DHS provides data for the period during protracted conflict in Mali (referred to as “wartime” hereafter). Exposure to wartime conflict is measured using conflict events reported by ACLED. ACLED collects real-time data on the locations, dates, actors, fatalities, and types of all reported political violence and protest events across various countries. These events are recorded by date and type whether they generate fatalities or otherwise. We spatially link observations from the 2006 and 2018 Mali DHS to events recorded in the ACLED for Mali using the GPS coordinates provided in both the DHS and ACLED. Although anecdotal reports suggest that the current crisis in Mali began in 2012, we consider all events after the period of peace in our study, i.e., we consider all events between 2007 to 2018. We identify 1,926 events during this period with over 95 percent of these occurring between 2012 and 2018. 2.2 Measuring Exposure to Conflict Studies on micro level outcomes exploiting the heterogeneity within country have measured conflict exposure as residence in an administrative area with high intensity of conflict or residence within a target distance of conflict events, i.e., a conflict buffer zone. Ekhator-Mobayode et al. (2020) measure exposure to conflict as residence within 10km of any conflict event in the period of interest. We refine this definition and posit that events within the conflict radius do not have the 4 same weights –closer events are likely to have greater impacts than events further away and so should have greater weights. Hence, we geographically discount the events using an exponential decay function as in Bozolli (2011). This strategy applies a lesser weight the further away a conflict event is from an individual/ household. The sum of events within the 10km buffer is thus given as follows 2: ∑ −( ( ,ℎ)) ℎ = =1 (1) Where ( , ℎ) is defined as the square of the distance (d) in kilometers between the household (h) and each of the ACLED events ( ) and is the distance-discount factor. We estimate the exponential function in equation (1) using nonlinear least squares and calculate the predicted values for events within 10km for each woman in the DHS. These predicted values are then summed to determine the sum of geographically discounted events. Exposure to wartime conflict in Mali is then defined as residence within at least 1 geographically discounted ACLED event within 10km during wartime. Applying this definition of exposure, we find that about 36 percent of the women in our study sample were exposed to wartime conflict. These are the women living in the DHS clusters affected by wartime conflict in Mali and make up the treatment group in our study (see figure 2). Figure 2: Wartime conflict occurring within 10km of communities surveyed by the Mali Demographic and Health Survey 2 We check that the results are robustness to changes in the definition of the distance away from conflict events by exploiting other distance definitions- 20km to 50km. This is especially important since the GPS coordinates for randomly selected respondents in the DHS are displaced to protect the confidentiality of respondents. 5 2.3 Outcome Variables We examine the impact of political conflict on women’s experience of IPV, experience of controlling behavior from their husbands or partners, earnings decision making autonomy and household decision making autonomy. Both the 2006 and 2018 Mali DHS ask women questions about specific behaviors from the husbands/partners to determine their experience of physical, sexual and emotional IPV. To measure experience of physical IPV, women are asked if they experienced pushing, slapping, arm twisting, punching, kicking, choking and attacking with weapons. For sexual IPV, women are asked whether they experienced forceful sexual intercourse or performance of other sexual acts. For emotional IPV, women are asked whether their husbands or partners humiliated them, threatened them with harm or insulted them to make them feel bad. Because the questions are based on recall and our methodology depends on knowing the timing of the experience of IPV, we focus on women’s experiences of IPV in the 12 months before the survey as a proxy for current experience of IPV. 3 We create variables measuring past year experience of physical IPV, past year experience of sexual IPV, past year experience of emotional IPV, past year experience of physical or sexual IPV and past year experience of physical, sexual or emotional IPV. Each of these are binary variables, equal to 1 if a woman answers yes to experiencing any behaviors about the various forms of IPV from her husband/partner in the past 12 months and 0 otherwise. To determine whether a woman experiences controlling behavior from her husband/partner, the DHS asks if a husband/partner gets jealous when she talks to other men, accuses her of unfaithfulness, does not permit her to meet female friends, limits her contact with family or insists on knowing where she is at all times. We use this to create the variable for controlling behavior, which equals 1 if a woman answers yes to any of these questions and 0 otherwise. To determine earnings decision-making autonomy, women who earned cash for work in the 12 months preceding the survey are asked who usually decides how their earnings are spent. The variable capturing women’s earnings decision-making autonomy is equal to 1 if a woman makes decisions alone or jointly with her husband/partner and 0 otherwise. Finally, women’s household decision-making autonomy is measured by answers given to questions about three types of household decisions: health care, large household purchases and visit to family. A woman has household decision- making autonomy if she makes decisions alone or jointly with her husband/partner. We create a variable for women’s household decision-making autonomy by summing the binary variables indicating autonomy in the three types of decisions. The index ranges from 0 to 3, corresponding to the number of household decisions in which a woman has joint or sole autonomy. 2.4 Control Variables We account for factors other than conflict that may affect the outcome variables discussed in section 2.3. The covariates are classified as follows: individual characteristics, partner characteristics, household characteristics as well as attitudes and experiences of IPV during childhood (see table 1). The empirical model is described in the next section. 3 Previous work has also used experience of IPV in the past year as proxy for current experience of IPV (see Abramsky et al. 2011 and Ekhator-Mobayode 2020). 6 Table 1: Variables used in the analysis A. IPV 1. Past year physical IPV Woman experienced physical IPV in the 12 months preceding the NDHS. 2. Past year sexual IPV Woman experienced sexual IPV in the 12 months preceding the NDHS. 3. Past year emotional IPV Woman experienced emotional IPV in the 12 months preceding the NDHS. 4. Past year physical or sexual IPV Woman experienced physical or sexual IPV in the 12 months preceding the NDHS. 5. Past year physical, sexual or Woman experienced physical, sexual or emotional IPV in the 12 months emotional IPV preceding the NDHS. B. Controlling behavior 6. Controlling behavior by husband/ Woman’s husband/partner exhibits any control issue. ** partner C. Decision making autonomy 7. Women’s household decision- Number of household decisions made alone or jointly with making autonomy index husband/partner 8. Women’s earnings decision- Woman decides how her earnings are spent alone or jointly with making autonomy husband/partner. D. Covariates Individual characteristics Age in years*; Education*; Employment status*; Type of earnings****; Total number of children born; Married as a child Partner characteristics Age in years*; Education*; Employment status*; Sometimes or often Drunk Household characteristics Polygamous household; Household headship status; Household size*; Location (rural/urban*); Wealth* Attitudes and experiences of IPV Father beat mother; Condones wife beating. during childhood Other variables Religion and regional dummies capturing differences in socio-economic conditions between regions Note: * denotes variables used in the matching stage. ** defined as answering yes to whether a husband/partner gets jealous when she talks to other men, accuses her of unfaithfulness, does not permit her to meet female friends, limits her contact with family or insists on knowing where she is at all times. **** employment type i.e., sector of employment is not available for both years of DHS survey hence type of earnings is used as proxy. 3. Model Specification In the absence of war, changes in the rates of IPV and the other outcome variables should be similar for both the population exposed to war and those not exposed. However, if exposure to wartime conflict events increases IPV, this will not be true. To test this hypothesis, we estimate the following difference-in-difference (DD) model: = 0 + 1 + 2 + 3 * + 4 + (2) The DD model exploits the variation in timing and location of the wartime conflict events discussed in section 2.2. The unit of observation for equation (2) is a woman. is any of the outcome variables discussed in section 3.3 for a woman i, interviewed in an area l, and DHS year y. is a binary variable indicating the DHS year the woman was interviewed. It is 1 for observations from the 2018 Mali DHS during the war and 0 otherwise. indicates 7 whether the woman resides in a wartime conflict affected area discussed in section 2.2. The covariates listed in table 1 are represented by . Finally, is the stochastic error term. The parameter of interest is 3, the difference-in-difference estimate. It measures the average treatment effect of residing in the affected area during wartime in Mali on the outcome variables. 3.1 Estimation Bias Selection bias across time and groups is common with standard DD models and may bias the estimates from equation (2) (Stuart et al. 2014). It is expected that the difference between women in affected and non-affected areas is constant before and during wartime. However, population movements to and from areas affected by war may impact the composition of women in the affected area. We address this potential source of bias by estimating a kernel-based propensity score matching method to ensure that the differences in composition of women in the affected and unaffected areas before and during wartime are constant based on observables. Specifically, the kernel-based propensity score matching (PSM) method matches each of the individual observations in the group of women who reside in the affected area in peacetime, women who reside in the unaffected area in peacetime and women who reside in the unaffected areas in wartime to the observations of women in the affected areas in wartime. The weights from the balanced sample are then used to estimate the effect of war on the outcomes using the common support of the matched sample in the DD model. Caliendo et al. (2005) suggests that previous research should be relied on in building the matching model. Hence, we rely on evidence from Ceriani and Verme (2018) who find that age, education, employment status, and ownership of household assets are associated with the decision to stay or flee from a conflict affected area. 4 Thus, the variables used for matching are individual characteristics of the women and their partners using the following variables: age, education, household size, type of settlement (urban versus rural), and household wealth status. The results from estimating the model in equation 2 on the matched sample are presented in the next section. 4. Results 4.1 Difference-in-difference (DD) estimates 4.1.1 Women’s experience of IPV In table 2, we compare IPV trends between the affected and unaffected areas in peacetime and wartime in the study sample. The matched sample includes data on 2,059 women living in war affected areas and 3,710 women living in non-affected areas. When looking at the combined measure of IPV, prevalence rates increased significantly in both affected and non-affected areas. However, the increase from peacetime to conflict in conflict- affected areas was notably higher than non-conflict affected areas (19 versus 7 percent). All forms of IPV separately also increased significantly in the conflict-affected areas from peacetime to wartime (physical, sexual and emotional IPV). The same is not true in non-affected areas, where physical IPV significantly decreased and the combined measure of physical or sexual IPV stayed the same. Only past-year sexual and emotional IPV increased significantly over the time period in non-affected areas. 4 Larger household size, being male, being older and being unemployed were associated with not migrating during conflict while more years of education was associated with migrating during conflict. 8 Table 2: Rates of Intimate Partner Violence (IPV) in Mali in peace and wartime, percent Affected Area Non-Affected Area Variable Peacetime Wartime Difference Peacetime Wartime Difference N (A) (B) (C=B-A) (D) (E) (F=E-D) (Total) Past year physical IPV 13 17 4** 23 18 -5*** 9 N 1503 556 2835 875 5769 Past year sexual IPV 3 8 5*** 5 10 5*** 6 N 1502 556 2835 875 5768 Past year emotional IPV 9 32 23*** 13 29 16*** 16 N 1503 556 2835 875 5769 Past year physical or sexual IPV 15 21 6*** 24 22 -2 21 N 1503 556 2835 875 5769 Past year physical, sexual or 18 37 19*** 28 35 7*** 28 emotional IPV N 1503 556 2835 875 5769 Source: Authors own calculation using data from the 2006 and 2018 Mali Demographic and Health Survey Note: Affected Area defined as DHS clusters exposed to any geographically discounted event within 10km between 2007 and 2018. *** Difference in means is significant at the 1% level. Using the DD model described in equation 2, we examine whether the differences in IPV trends between the affected and unaffected areas from peacetime to wartime can indeed be explained by exposure to wartime conflict. The results are presented in table 3. 9 Table 3: Difference-in-differences estimate of the effect of war on women’s experience of Intimate Partner Violence (IPV) in Mali Variables Past year Past year Past year Past year physical Past year physical IPV sexual emotional or sexual IPV physical, sexual IPV IPV or emotional IPV * 0.138*** -0.007 0.115*** 0.119*** 0.178*** (0.031) (0.026) (0.038) (0.038) (0.052) Individual characteristics Type of earnings (reference category: no earnings) Cash only 0.005 -0.031 0.017 0.005 0.007 (0.024) (0.017) (0.018) (0.026) (0.027) Cash and kind 0.052** -0.018 0.054** 0.048* 0.059* (0.026) (0.016) (0.023) (0.028) (0.029) Kind only 0.032 -0.015 0.039* 0.032 0.040 (0.039) (0.016) (0.022) (0.033) (0.035) Total number of children born <-0.0001 -0.004** -0.004 -0.001 -0.003 (0.003) (0.002) (0.002) (0.003) (0.004) Married as a child (before 18 years 0.008 -0.001 0.001 0.001 -0.006 of age) (0.020) (0.015) (0.019) (0.022) (0.026) Partner characteristics Sometimes or often drunk 0.234*** 0.114** 0.189*** 0.241*** 0.178** (0.081) (0.056) (0.064) (0.081) (0.077) Household characteristics Polygamous household 0.035** 0.010 0.023 0.035** 0.039* (0.015) (0.010) (0.014) (0.016) (0.020) Household head -0.028 -0.018 -0.039* -0.038 -0.044 (0.035) (0.011) (0.022) (0.026) (0.029) Attitudes and experiences of IPV during childhood Father beat mother 0.141*** 0.092*** 0.092*** 0.188*** 0.183*** (0.031) (0.034) (0.025) (0.041) (0.041) Condones wife beating 0.030 0.031*** 0.048*** 0.046*** 0.062*** (0.019) (0.010) (0.013) (0.021) (0.022) Others Religion (reference category: No religion) Christianity -0.007 0.033 -0.101** -0.022 <0.001 (0.054) (0.027) (0.043) (0.056) (<0.001) Islam 0.028 -0.029 -0.043 0.013 0.032 (0.035) (0.020) (0.028) (0.036) (0.049) Other religion <0.001 <0.001 <0.001 <0.001 0.034 (<0.001) (<0.001) (<0.001) (<0.001) (0.059) Regional Fixed effects Yes Yes Yes Yes Yes N Before War (Non-Affected Area) 2835 2835 2835 2835 2835 Before War (Affected Area) 1503 1502 1503 1503 1503 Wartime (Non-Affected Area) 875 875 875 875 875 Wartime (Affected Area) 556 556 556 556 556 Total 5769 5768 5769 5769 5769 R-squared 0.06 0.05 0.09 0.06 0.07 Note: Variables used to estimate the kernel propensity scores in the matching stage are dropped in the DD estimation. Estimations are done using weights provided in the DHS data. The kernel density function is Epanechnikov with a bandwidth of 0.06. Probit estimation is used for the propensity score in the first stage. ***Significant at the 1% level, **Significant at the 5% level, *Significant at the 10% level. Standard errors clustered at the DHS cluster level in parentheses. 10 The results of the DD model show that living in a conflict-affected area increases the probability of a woman’s experience of physical IPV alone by 13.8 percent, emotional IPV alone by 11.5 percent, physical or sexual IPV by 11.9 percent and physical, sexual, or emotional IPV by 17.8 percent after controlling for a woman’s partner and household characteristics, her attitude towards the acceptability of wife beating, her witnessing IPV in childhood within the household, religion and regional fixed effects. We find no effect of the war on sexual IPV alone. Two covariates are consistently associated with increased risk of women’s experience of IPV. The first is having a sometimes or often drunk partner (23.4, 11.4, and 18.9 percent increase for physical IPV alone, sexual IPV alone and emotional IPV alone respectively and 24.1 and 17.8 percent increase for physical or sexual IPV and physical, sexual or emotional IPV, respectively). The second variable is witnessing a father who physically abused his wife/partner (14.1 percent increase for physical IPV alone, 9.2 percent increase for sexual IPV alone and emotional IPV alone respectively and 18.8 and 18.3 percent increase for physical or sexual IPV and physical, sexual, or emotional IPV respectively). 4.1.2 Women’s experience of controlling behavior and decision-making autonomy In table 4, we compare trends in experience of controlling behavior as well as trends in decision making autonomy for women in the study sample living in areas affected by the Mali civil war to their counterparts in areas not affected. Table 4: Women’s experience of controlling behavior by husband/partner and women’s decision- making autonomy in Mali before and during Wartime, percent Variables Affected Area Non-Affected Area Peaceti Warti Difference Peacetime Wartime Difference N me me (C=B-A) (D) (E) (F=E-D) (Total) (A) (B) Controlling behavior by husband/partner 76 68 -8*** 74 63 -11*** 72 N 1503 556 2835 875 5769 Women’s earnings decision-making autonomy 95 82 -13*** 92 86 -6*** 91 N 1225 461 1914 546 4146 Women’s household decision- making autonomy 0.79 1 0.21*** 0.75 0.60 -0.15*** 0.76 Autonomy over healthcare 22 29 7*** 17 19 2 20 Autonomy over major household purchases 22 31 9*** 21 19 2 22 Autonomy over visits to friends and family 35 40 5** 37 22 15*** 34 N 1502 556 2832 875 5765 Source: Authors own calculation using data from the 2006 and 2018 Mali Demographic and Health Survey Note: Affected Area defined as DHS clusters exposed to any geographically discounted event within 10km between 2007 and 2018. *** Difference in means is significant at the 1% level. ** Difference in means is significant at the 5% level. * Difference in means is significant at the 10% level. In both the affected area and non-affected area, the proportion of women reporting experiencing controlling behavior from their husbands/ partners reduced from peacetime to wartime. However, in the non-affected area the reduction was 3 percent more. Also, in both the affected area and non- affected area, the proportion of women reporting having some autonomy over their own earnings reduced from peacetime to wartime. However, the reduction was 7 percent more in the affected area. Finally, while the degree of women’s autonomy over household decisions reduced from 11 peacetime to wartime by 0.15 units in the non-affected area, it increased by 0.21 units in the affected area. We estimate the DD model in equation 2 to examine if these trends can be explained by the war. The results from this estimation are presented in table 5. The results of the DD model show that the presence of the war decreases the probability that a woman has some autonomy over her own earnings by 6.9 percent after controlling for a woman’s, partner and household characteristics, attitudes towards the acceptability of using violence towards a partner and experiences of witnessing IPV in childhood, religion and regional fixed effects. We also find that the presence of the war increases the degree of women’s autonomy over household decisions by 0.79 units. The component of the household decision making index driving this result is autonomy on decisions about large household purchases and autonomy on decisions about visits to family. We find no effect of the war of women’s autonomy on decisions about their own health care. We also find no effect of the war of women’s experience of controlling behavior from their husbands/partners. 12 Table 5: Difference-in-differences estimates of the effect of War on women’s experience of controlling behavior and decision-making autonomy in Mali Variables Controlling Earnings Women’s household decision- making autonomy behavior decision-making autonomy Household Autonomy on Autonomy on Autonomy on decision decisions about decisions about decisions making healthcare large household about visit to index purchases family * -0.019 -0.069* 0.790*** 0.064 0.125*** 0.231*** (0.058) (0.038) (0.176) (0.047) (0.045) (0.058) Individual characteristics Type of earnings (reference category: no earnings) Cash only 0.013 - 0.161* 0.071*** 0.039* 0.035 (0.027) (0.087) (0.022) (0.020) (0.031) Cash and kind -0.064* - 0.384*** 0.114*** 0.133*** 0.092** (0.035) (0.121) (0.032) (0.033) (0.040) Kind only -0.005 - 0.214* 0.003 -0.031 0.095** (0.035) (0.113) (0.029) (0.031) (0.048) Total number of children born -0.009*** 0.003 0.043*** 0.010*** 0.011*** 0.011*** (0.003) (0.002) (0.009) (0.003) (0.003) (0.003) Married as a child (before 18 0.002 0.006 0.036 0.014 0.054*** -0.022 years of age) (0.021) (0.018) (0.071) (0.020) (0.017) (0.023) Partner characteristics Sometimes or often drunk 0.067 0.028 0.197 0.063 0.052 0.032 (0.052) (0.044) (0.230) (0.062) (0.059) (0.079) Household characteristics Polygamous household -0.007 0.005 0.038 0.016 0.003 0.002 (0.024) (0.013) (0.063) (0.019) (0.017) (0.021) Household head 0.002 0.054*** 0.451*** 0.144*** 0.108*** 0.131*** (0.035) (0.018) (0.106) (0.036) (0.032) (0.036) Attitudes and experiences of IPV during childhood Father beat mother 0.071*** -0.074*** 0.089 -0.041** -0.033* 0.033 (0.027) (0.035) (0.084) (0.020) (0.019) (0.030) Condones wife beating 0.139*** -0.023* -0.174** 0.006 -0.009 -0.099*** (0.030) (0.014) (0.085) (0.023) (0.022) (0.028) Others Religion (reference category: No religion) Christianity <0.001 <0.001 <0.0001 <0.001 <0.001 <0.001 (<0.001) (<0.001) (<0.001) (<0.001) (<0.001) (<0.001) Islam 0.001 -0.007 0.025 0.040 0.004 -0.014 (0.035) (0.015) (0.143) (0.044) (0.041) (0.050) Other religion 0.019 -0.116** -0.237 0.012 -0.025 -0.096 (0.053) (0.060) (0.179) (0.054) (0.049) (0.060) Regional Fixed effects Yes Yes Yes Yes Yes Yes N Before War (Non-Affected Area) 2835 1914 2835 2834 2833 2832 Before War (Affected Area) 1503 1225 1503 1503 1503 1502 Wartime (Non-Affected Area) 875 546 875 875 875 875 Wartime (Affected Area) 556 461 556 556 556 556 Total 5769 4146 5769 5768 5767 5768 R-squared 0.05 0.06 0.05 0.05 0.06 0.05 Note: Variables used to estimate the kernel propensity scores in the matching stage are dropped in the DD estimation. Estimations are done using weights provided in the DHS data. We log-transform the index of women’s household decision-making autonomy before estimating the model to allow for the reporting of the estimates in terms of percentages. To take care of zero values, we add 0.1 to the index before taking the log. Thus, the estimates reported calculated as ( − 1). The kernel density function is Epanechnikov with a bandwidth of 0.06. Probit estimation is used for the propensity score in the first stage. ***Significant at the 1% level, **Significant at the 5% level, *Significant at the 10% level. Standard errors clustered at the DHS cluster level in parentheses. 13 4.2 How does migration affect the results? It is not possible to know the proportion of women who moved out of the conflict-affected area during wartime as the DHS does not ask women about their previous place of residence, However, the DHS asks about the number of years women have lived in their current place of residence. To check that our results are robust to movements into the affected area during wartime, we restrict our sample to women who have lived in their place of residence for the entire duration of the war, i.e., only women who lived in the same location for at least seven years. For women in the 2018 survey, this translates to residing in the same residence for the entire duration of the period of war, i.e., between 2012 and 2018. We estimate equation 1 on the matched restricted sample. The effect of wartime on women’s experience of IPV using the restricted sample is consistent with the analysis of the whole sample. This is not surprising as about 84 percent of the sample have lived in their current place of residence for 7 or more years, over 14 percent have remained in the same place for between 1 and 6 years and less than 2 percent with have lived for 1 year or less in their current place of residence. The results for this estimation for IPV outcomes are presented in Table 6 while the results for controlling behavior and decision-making autonomy are presented in Table 7. As before, the war increased women’s experience of physical IPV alone, emotional IPV alone, physical, or sexual IPV, and physical, sexual, or emotional IPV. It is also consistent with increased household decision making autonomy in the main analysis. As before, we find no effect on controlling behavior. While the direction of the effect of the war on earnings decision-making autonomy is the same as in the main analysis, it loses its statistical significance. 14 Table 6: Difference-in-differences estimates of the effect of war on women’s experience of intimate partner violence – restricted to sample of women living in place of residence for at least 7 years. (1) (2) (3) (4) (5) Variables Past year Past year Past year Past year Past year physical IPV sexual IPV emotional physical or physical, sexual IPV sexual IPV or emotional IPV * 0.130*** -0.014 0.129*** 0.097** 0.167*** (0.032) (0.035) (0.039) (0.046) (0.057) Covariates Individual characteristics Yes Yes Yes Yes Yes Partner characteristics Yes Yes Yes Yes Yes Household characteristics Yes Yes Yes Yes Yes Regional Fixed effects Yes Yes Yes Yes Yes N Before Wartime (Non-Affected 2365 2365 2365 2365 2365 Area) Before Wartime (Affected Area) 1264 1264 1264 1264 1264 During Wartime (Non-Affected 750 750 750 750 750 Area) During Wartime (Affected Area) 458 458 458 458 458 Total 4837 4837 4837 4837 4837 R-squared 0.05 0.06 0.09 0.06 0.07 Note: Variables used to estimate the kernel propensity scores in the matching stage are dropped in the DD estimation. Estimations are done using weights provided in the DHS data. The kernel density function is Epanechnikov with a bandwidth of 0.06. Probit estimation is used for the propensity score in the first stage. ***Significant at the 1% level, **Significant at the 5% level, *Significant at the 10% level. Standard errors clustered at the DHS cluster level in parentheses. 15 Table 7: Difference-in-differences estimates of the effect of War on Women’s experience of controlling behavior by husband/partner and women’s decision-making autonomy in Mali- restricted to sample of women living in place of residence for at least 7 years. (1) (2) (3) (4) (5) (6) Variables Controlling Earnings Household decision making autonomy behavior decision- making autonomy Household Autonomy Autonomy on Autonomy decision on decisions on making decisions about large decisions index about household about visit healthcare purchases to family * -0.016 -0.049 0.840*** 0.061 0.127*** 0.239*** (0.065) (0.045) (0.174) (0.047) (0.046) (0.058) Covariates Individual characteristics Yes Yes Yes Yes Yes Yes Partner characteristics Yes Yes Yes Yes Yes Yes Household characteristics Yes Yes Yes Yes Yes Yes Regional Fixed effects Yes Yes Yes Yes Yes Yes N Before Wartime- placebo (Non-Affected 2365 1575 2365 2364 2363 2366 Area) Before Wartime- placebo (Affected Area) 1264 1009 1264 1264 1264 1263 During Wartime placebo (Non-Affected 750 472 750 750 750 750 Area) During Wartime placebo (Affected Area) 458 404 458 458 458 458 Total 4837 3460 4837 4836 4835 4837 R-squared 0.06 0.08 0.07 0.04 0.05 0.05 Note: Variables used to estimate the kernel propensity scores in the matching stage are dropped in the DD estimation. Estimations are done using weights provided in the DHS data. The kernel density function is Epanechnikov with a bandwidth of 0.06. Probit estimation is used for the propensity score in the first stage. ***Significant at the 1% level, **Significant at the 5% level, *Significant at the 10% level. Standard errors clustered at the DHS cluster level in parentheses. 4.3. Evidence in support of the DD estimation method One key identifying assumption of the DD model specified in equation (1) is that in the absence of war, the differences in the trends of the outcome variables between the affected and unaffected areas would be similar. Although it is not possible to empirically test this assumption, we present evidence in support of this assumption. First, in figures 3 and 4, we present graphs of the mean levels of the outcome variables in peacetime and wartime. The graphs suggest that in peacetime, the trends in outcomes are parallel between the affected area and unaffected area, suggesting that the identifying assumption of parallel trends holds. 16 Figure 3: Peace vs. wartime trends - IPV Figure 4: Peace vs. wartime trends – controlling behavior, and decision making Second, we estimate equation (1) for IPV, decision making and controlling behavior using the sample of women in peacetime only, i.e., using only the 2006 DHS for which eligible women were interviewed between April and December 2006. The results are presented in tables 8 and 9. We assume that women who were interviewed in the first half of the survey period (between April and July) make up the placebo sample in peacetime, while those interviewed in the second half of the survey period (between August and December) make up the placebo sample in wartime. We find no spurious significant coefficient. Although some of coefficients of women’s experience of 17 particular types of IPV are significant, this seems unrelated to the war since the signs are different from the main analysis, again suggesting that the assumption of parallel trends holds. Table 8: Difference-in-differences estimates of the effect of War on Women’s experience of intimate partner violence – Placebo analysis. (1) (2) (3) (4) (5) Variables Past year Past year Past year Past year Past year physical IPV sexual IPV emotional physical or physical, sexual IPV sexual IPV or emotional IPV * -0.035 -0.036** -0.080*** -0.050 -0.075* (0.035) (0.018) (0.027) (0.037) (0.040) Covariates Individual characteristics Yes Yes Yes Yes Yes Partner characteristics Yes Yes Yes Yes Yes Household characteristics Yes Yes Yes Yes Yes Regional Fixed effects Yes Yes Yes Yes Yes N Before Wartime- placebo (Non- 1295 1295 1295 1295 1295 Affected Area) Before Wartime- placebo 692 692 692 692 692 (Affected Area) During Wartime placebo (Non- 1538 1538 1538 1538 1538 Affected Area) During Wartime placebo (Affected 802 802 802 802 802 Area) Total 4327 4327 4327 4327 4327 R-squared 0.06 0.04 0.04 0.07 0.07 Note: Variables used to estimate the kernel propensity scores in the matching stage are dropped in the DD estimation. Estimations are done using weights provided in the DHS data. The kernel density function is Epanechnikov with a bandwidth of 0.06. Probit estimation is used for the propensity score in the first stage. ***Significant at the 1% level, **Significant at the 5% level, *Significant at the 10% level. Standard errors clustered at the DHS cluster level in parentheses. 18 Table 9: Difference-in-differences estimates of the effect of War on Women’s experience of controlling behavior by husband/partner and women’s decision-making autonomy in Mali- Placebo Analysis. (1) (2) (3) (4) (5) (6) Variables Controlling Earnings Household decision making autonomy behavior decision- making autonomy Household Autonomy Autonomy Autonomy decision on on decisions on making decisions about large decisions index about household about visit healthcare purchases to family * 0.074 0.005 -0.200 -0.034 -0.033 -0.063 (0.047) (0.027) (0.158) (0.040) (0.080) (0.250) Covariates Individual characteristics Yes Yes Yes Yes Yes Yes Partner characteristics Yes Yes Yes Yes Yes Yes Household characteristics Yes Yes Yes Yes Yes Yes Regional Fixed effects Yes Yes Yes Yes Yes Yes N Before Wartime- placebo (Non-Affected 1295 850 1295 1295 1295 1295 Area) Before Wartime- placebo (Affected 692 580 692 692 692 691 Area) During Wartime placebo (Non-Affected 1538 1034 1538 1537 1536 1538 Area) During Wartime placebo (Affected Area) 802 647 802 802 802 802 Total 4327 3111 4327 4326 4325 4326 R-squared 0.09 0.03 0.05 0.06 0.05 0.05 Note: Variables used to estimate the kernel propensity scores in the matching stage are dropped in the DD estimation. Estimations are done using weights provided in the DHS data. The kernel density function is Epanechnikov with a bandwidth of 0.06. Probit estimation is used for the propensity score in the first stage. ***Significant at the 1% level, **Significant at the 5% level, *Significant at the 10% level. Standard errors clustered at the DHS cluster level in parentheses. 19 4.4 Robustness to varying the radius of the affected area In Figures 5 and 6, we provide evidence demonstrating the robustness of our significant results in in sections 4.1.1 and 4.1.2 to variation in the radius of the affected area. Increasing the radius of the affected area increasingly includes women residing further away from the location of conflict events in the affected area and reduces the difference between the number of women in the affected and non-affected areas. We expect a reduction or no change in the magnitude of the adverse effect of conflict events since more women who do not reside in the immediate environment of conflict events are added to the affected area. Figure 5 shows that IPV outcomes are consistent with this expectation – increasingly expanding the affected area to include women residing further away from conflict events reduced the probability that women in the affected area experience IPV. Similarly, figure 6 shows that women closest to the conflict events lose more of their household decision making autonomy compared to their counterparts who reside further away. Finally, the adverse effect of conflict on earnings decision making autonomy remains fairly constant as the radius of the affected area expands, suggesting that conflict has no effect on earnings decision making autonomy for women who do not live in the immediate proximity of conflict events. Figure 5: Kernel-based DD Estimates for various BH Area buffer zones – IPV outcomes 20 Figure 6: Kernel-based DD Estimates for various BH Area buffer zones – controlling behavior and decision making 21 5. Discussion and Conclusion This study exploits variation in timing and location of conflict events during the ongoing civil war in Mali to identify the impact of conflict on women’s experience of IPV and decision-making autonomy. Two DHS surveys - one conducted before the conflict and one conducted during the ongoing violence - were spatially linked to conflict data. The difference-in-difference approach allows for matching of women in the two areas to isolate conflict as a driver of increased violence using a rigorous quasi-experimental research design. This approach represents one of the first quantitative, population-based assessments of how the ongoing instability may affect Malian women in key domains of violence and personal decision-making autonomy. These findings highlight the fact that women living in conflict-affected areas experienced notable and significant increases in all forms of IPV, and combined measures of IPV. In the non-conflict- affected areas, women faced lower levels of physical IPV and no change in the combined measure of physical and sexual IPV. Women in non-conflict affected areas in Mali still experienced increased emotional and sexual IPV. However, when examining the change from peacetime to wartime in the two areas, we see that women living in conflict-affected areas were significantly more likely than their non-affected counterparts to experience multiple forms of IPV, including a combined measure of physical and sexual IPV, as well as physical, sexual and emotional IPV combined. This paper also found an increase in risk of experiencing IPV if a woman has witnessed a father physically abuse his wife/partner or have a partner who is sometimes or often drunk, both known risk factors for IPV. This study also finds that women in conflict affected areas report greater decision-making within the home in at least some domains – specifically, decision-making related to large household purchases and paying visits to family. This analysis does not permit understanding of the potential pathways for increased autonomy in these areas. However, other scholarship has traced how conflict may lead to higher numbers of women-headed households during instability. Husbands and other male relatives may be injured, missing or killed during conflict and forced displacement. This reality may result in women having greater joint or sole decision-making power in households that are experiencing shifts in traditional power dynamics. It is also possible that the pressing realities of navigating political instability mean that women take more decisions upon themselves or gain greater say in joint decision making. The fact that men may be less present or engaged within the home also highlights the stark findings related to IPV; even while some households may experience changes in composition and certain power dynamics, domestic violence has still increased in conflict-affected areas. Given this, it is important to consider that improvements in autonomy at the household level alone may not indicate an increase overall gender equality in conflict-affected areas – as starkly highlighted by the results related to IPV. This is particularly salient given the post-conflict context in Mali. The country continues to rank at the bottom of the Gender Equality Index, 158th of 162 countries. Malian women continue to lack legal protections around mobility, employment, pay, GBV and marriage (Trumbic et al, 2020). Mali also continues to show high prevalence rates of IPV and failure to pass a draft GBV law drafted sin 2017 that has received strong opposition from the High Islamic Council. Lack of equality in these other areas, coupled with the findings of this paper speak to the need to undertake research that examines women’s experiences holistically when examining their safety and empowerment. Further research could support the understanding 22 of the pathways and dynamics underlying changes in empowerment domains for women and girls and how they are interrelated. This paper highlights the need for continued analyses by investigating different measures of autonomy with IPV. These findings make a strong case for bringing outcomes together to view larger pictures of women’s experiences. The study points to several areas where policy and programming may have the opportunity to reduce violence and increase gender equality within the household. Multi-sectoral programs that address harmful gender norms at the individual, household and community levels are vital to transform gender norms and ensure that women can fully and safely participate in society. These findings also point to the need to ensure women’s substantive participation in peace negotiations and state-building. Preceding the conflict, the Government of Mali launched a National Gender Policy (2010) and National Action Plan for Women, Peace, and Security (2012) – processes that included multi-day consultations with women’s civil society organizations while writing the document (Shephard, 2019). However, in the 2015 Bamako Agreement, only 5 percent of peace talk negotiators and 15 percent of the signatories were women (Council on Foreign Relations, 2019). To shift norms and attitudes around gender roles, women’s organizations need a significant presence in peacebuilding and state-building efforts. Without this engagement, women will continue to be left behind in peace processes – bearing the brunt of violence that moves from the public to the private sphere. 23 References Abramsky T, Watts CH, Garcia-Moreno C, Devries K, Kiss L, Ellsberg M, et al. (2011). “What factors are associated with recent intimate partner violence? Findings from the WHO multi-country study on women's health and domestic violence.” BMC Public Health 111:1. Buvinic, M., Das Gupta, M., Casabonne, U., & Verwimp, P. (2013). Violent Conflict and Gender Inequality: An Overview. The World Bank Research Observer, 28(1), 110–138. https://doi.org/10.1093/wbro/lks011 Bozolli C Bruck T. Muhumuza T. (2011). Does war influence individual expectations? Economic Letters 113 (2011) 288–291 Caliendo M., Kopeinig S., (2005). “Some practical guidance for the implementation of propensity score matching.” IZA Discussion Paper No. 1588 Ceriani L., Verme P., 2018 “Risk preferences and the decision to flee conflict.” The World Bank Group Policy Research Working Paper 8376 Chukwuma A., Ekhator-Mobayode U.E., (2019). Armed conflict and maternal health care utilization: evidence from the Boko Haram Insurgency in Nigeria. Social Science & Medicine 226, 104- 112 https://doi.org/10.1016/j.socscimed.2019.02.055. Council on Foreign Relations (2019). “Explore the Data”, CFR: Women’s Role in Peace Processes. Retrieved from: https://www.cfr.org/interactive/interactive/womens-participation-in- peaceprocesses/explore-the-data DHS STATcomplier, (2018). The DHS Program STATcompiler. Funded by USAID. http://www.statcompiler.com. [Accessed June 14, 2021]. Ekhator-Mobayode, U.E., Abebe Asfaw, A., (2019). The child health effects of terrorism: evidence from the Boko Haram Insurgency in Nigeria. Appl. Econ. 51 (6), 624–638. Ekhator-Mobayode, UE and Hanmer, L and Rubiano Matulevich, EC and Arango, DJ (2020). Effect of Armed Conflict on Intimate Partner Violence: Evidence from the Boko Haram Insurgency in Nigeria. World Bank Policy Research Working Paper No. 9168, Available at SSRN: https://ssrn.com/abstract=3547923 Ellsberg, M., Arango, D.J., Morton, M., Gennari, F., Kiplesund, S. et al. (2015). Prevention of violence against women and girls: what does the evidence say? Lancet, 385(9977), 1555-1566. Falb K.L., McCormick M.C., Hemenway D., Anfinson K., & Silverman J.G.(2013). Violence against refugee women along the Thai-Burma border. Int J Gynaecol Obstet 120:279–83. Gupta, J., Acevedo-Garcia, D., Hemenway, D., Decker, M. R., Raj, A., & Silverman, J. G. (2009). Premigration exposure to political violence and perpetration of intimate partner violence among immigrant men in Boston. American journal of public health, 99(3), 462-469. 24 Gupta J, Reed E, Kelly J, et al (2012). Men’s exposure to human rights violations and relations with perpetration of intimate partner violence in South Africa. J Epidemiol Community Health 66:e2 Hagberg S. Korling G. (2012) . Socio-political Turmoil in Mali: The Public Debate Following the Coup d’État on 22 March 2012 . Africa Spectrum IISN 1868-6869 https://journals.sagepub.com/doi/pdf/10.1177/000203971204702-306 Heise, L. L. (1998). Violence Against Women: An Integrated, Ecological Framework. Violence Against Women, 4(3), 262–290. https://doi.org/10.1177/1077801298004003002 Hynes M., Ward J., Robertson K. (2004). A determination of the prevalence of gender-based violence among conflict-affected populations in East Timor. Disasters. https://doi.org/10.1111/j.0361- 3666.2004.00260.x Internal Displacement Monitoring Center (n.d). Mali: Annual conflict and disaster displacement figures 2012-2020. Norwegian Refugee Council/Internal Displacement Monitoring Centre. https://www.internal- displacement.org/countries/mali Justino, P. (2017). Violent Conflict and Changes in Gender Economic Roles (F. Ní Aoláin, N. Cahn, D. F. Haynes, & N. Valji, Eds.; Vol. 1). Oxford University Press. https://doi.org/10.1093/oxfordhb/9780199300983.013.7 Kelly J.T.D., Colantuoni E., Robinson C., & Decker M.R. (2018) From the battlefield to the bedroom: a multilevel analysis of the links between political conflict and intimate partner violence in Liberia. BMJ Glob Health 3:e000668. doi:10.1136/bmjgh-2017-000668 Lamarche, A. (2019) Mali’s Human Rights Crisis: Overmilitarized and Overshadowed. Refugees International. https://www.refugeesinternational.org/reports/2019/12/10/malis-humanitarian-crisis- overmilitarized-and-overshadowed Lecocq B., Mann G., Whitehouse B., Badi D., Pelckmans L., Belalimat N., Hall B. & Lacher W. (2013) One hippopotamus and eight blind analysts: a multivocal analysis of the 2012 political crisis in the divided Republic of Mali, Review of African Political Economy, 40:137, 343- 357, DOI: 10.1080/03056244.2013.799063 Østby, G. (2016). Violence begets violence: Armed conflict and domestic sexual violence in sub-Saharan Africa. In SVAC (Sexual Violence and Armed Conflict) Workshop, Harvard University, Cambridge (pp. 2- 3). Peterman A., Palermo T., Bredenkamp C. (2011) Estimates and determinants of sexual violence against women in the Democratic Republic of Congo. Am J Public Health. 101(6):1060–1067. doi:10.2105/AJPH.2010.300070. Saile, R., E.V. Ertl, F. Neuner, and C. Catani. (2013). “Prevalence and Predictors of Partner Violence against Women in the Aftermath of War.” Social Science & Medicine 86: 17–25. Shephard, L. (2019). National Action Plan for the Implementation of Resolution 1325 of the Security Council of the United Nations on Women, Peace, and Security. Unofficial Translation. Funded by ARC DP160100212. Retrieved from https://www.wpsnaps.org/app/uploads/2019/09/Mali-NAP-2012-2014- English-translation-DP160100212.pdf 25 Slegh, H., Barker, G., & Levtov, R. (2014). Gender Relations, Sexual and Gender-based Violence and the Effects of Conflict on Women and Men in North Kivu, Eastern Democratic Republic of the Congo: Results from the International Men and Gender Equality Survey (IMAGES). Promundo US and Sonke Gender Justice. https://promundoglobal.org/wp-content/uploads/2015/01/Preliminary-Results-from-IMAGES- DRC.pdf Stark L. and Ager A. (2011). A Systematic Review of Prevalence Studies on Gender-Based Violence in Complex Emergencies. Trauma, Violence and Abuse, 12(3) 127–134. Stark, L., Roberts, L., Wheaton, W., Acham, A., Boothby, N., & Ager, A. (2010). Measuring violence against women amidst war and displacement in northern Uganda using the “neighborhood method”. Journal of Epidemiology & Community Health, 64(12), 1056-1061. Swaine A, Spearing M, Murphy M, Contreras-Urbina M. (2019). Exploring the Intersection of Violence Against Women and Girls With Post-Conflict Statebuilding and Peacebuilding Processes: A New Analytical Framework. Journal of Peacebuilding & Development. 14(1):3-21. doi:10.1177/1542316619833877 Trumbic,T.;Loayza,N.V.;Ramalho, R.;Kraay, A.C.;.2020. Women, Business, and the Law 2021. Washington, D.C. : World Bank Group. https://imagebank2.worldbank.org/search/32857340 United Nations Development Programme (2020). Human Development Report 2020: The next frontier. New York, NY. Retrieved from http://hdr.undp.org/en/content/gender-inequality-index-gii USAID (2012). Mali Gender Assessment. https://www.usaid.gov/sites/default/files/documents/1860/Mali-Gender-Assessment.pdf World Bank Group (n.d) Gender Data Portal: Mali. World Bank Group; Washington, DC. https://www.worldbank.org/en/data/datatopics/gender/country/Mali 26