**1. Introduction**

Violence against women and girls remains a major public health threat the world over. It is estimated that 35% of women worldwide experience either physical and/or sexual violence during their lifetime (WHO 2017). A significant amount of violence experienced by women and girls is that perpetrated by their intimate partners. According to the WHO (2017), almost one third of women who have been in a relationship in their lifetime have been subjected to physical and/or sexual violence at the hands of their intimate partner. Global and country-specific evidence shows that males are the main perpetrators of the intimate partner violence (IPV) meted out to women. For instance, a systematic review of global and regional evidence found that close to 40% of all murdered women were killed by a male intimate partner (WHO 2013). In the USA, 99% of the IPV against women reported in 2008 was perpetrated by men (National Data on Intimate Partner Violence 2012). In Uganda, demographic and health survey data of 2011 cite men as the main perpetrators of IPV (UBOS and ICF 2012).

The risk of experiencing IPV is amplified for women and girls who ge<sup>t</sup> married or into union before turning 18. Global statistics show that girls who marry before the age of 15 are 50% more likely to face physical or sexual violence from a partner (Girls Not Brides 2020). An analysis of data from a population-based survey in Ethiopia found that girls who married before the age of 15 were almost four times more likely to have experienced forced first marital sex compared to those who

married at 18 and 19 years (Erulkar 2013). In Bangladesh, a longitudinal multilevel analysis of how the community prevalence of very early child marriage influenced a woman's risk of IPV found that almost 70% of the women who reported experiencing physical abuse had married before the age of 18 (Yount et al. 2016). Demographic and health survey data of 34 low and middle income countries outside South Asia show that incidences of physical and sexual IPV were higher among women who married as children (29%) compared with those who married as adults (20%). Similarly, a randomised controlled trial of an IPV prevention programme in rural Cote d'Ivoire found that all forms of IPV were higher among women who married as child brides compared to those who go<sup>t</sup> into union as adults (Falb et al. 2015). In Uganda, the demographic and health survey of 2016 shows that slightly higher proportions of women who married before their 18th birthday experienced physical violence compared to their counterparts who married at the age of 18 and above (UBOS and ICF 2018).

Previous studies show that leaving a violent intimate relationship is often a di fficult and complex decision. Abused women usually balance personal, familial and even communal interests in decisions to stay or leave (Barnett 2000; Khoury and Wehbi 2016). Several barriers are associated with the entrapment of women in violent intimate relationships. These include economic limitations such as poverty and financial dependence and socio-cultural considerations such as social proscriptions that discourage divorce/separation and openness about experiences of IPV, and social and gender norms that value family continuity at all costs, normalise IPV against women, put the responsibility of family preservation on women and present marriage as a source of status and respect for women (Barnett 2000, 2001; Lacey 2010; Khoury and Wehbi 2016; Willan et al. 2019). Other inhibiting factors include the investment of significant amounts of time in the relationship, fears over the safety of the survivor and their family, fear of reprisal and lack of strong formal and informal support systems, among others (Barnett 2000, 2001; Lacey 2010). This is not to sugges<sup>t</sup> that women are passive victims of IPV when they do not report or leave violent relationships. Studies show that abused women often take steps to minimise or address the violence during the relationships. These include seeking the intervention of, and emotional support from, family, friends and community organisations, adjusting behaviours to meet the violent partner's expectations, threatening to report the abusive partner, limiting contact with the abuser in the home and seeking professional help (Ruiz-Pérez et al. 2006; Zink et al. 2006; Khoury and Wehbi 2016; Willan et al. 2019).

On the other hand, women's decisions to leave abusive intimate partner relationships are associated with several social, economic and psychological factors. An analysis of data from the domestic violence experiment in Omaha, Nebraska, USA, found that women who were financially independent and had a high self-esteem, an internal locus of control and less fear for their safety or reprisals were more likely to leave violent intimate partner relationships (Kim and Gray 2008). A qualitative study of how women in Lebanon made the decision to leave violent domestic relationships underlined the significance of family support in their decisions to leave (Khoury and Wehbi 2016). In South Africa, Willan et al. (2019) show that young women typically left violent love-relationships when they had strong emotional and economic support from family, the violence became public and their partners openly displayed their infidelity and no longer fulfilled their expectations of providing financial and material support to them and/or their children.

While there is considerable literature on the factors that influence leave or stay decisions among women experiencing IPV (see Barnett 2000, 2001; Anderson and Saunders 2003; Burman and Chantler 2005; Kim and Gray 2008; Lindgren and Renck 2008; Lacey 2010), it is not clear how such decisions play out in the context of child marriage, particularly within low resource settings. Girls who ge<sup>t</sup> into union before 18 years tend to be portrayed as passive victims of IPV due to their young age, limited education, economic dependence and broader structural issues that may curtail their agency, such as patriarchal norms promoting male domination and female subordination (Mathur et al. 2003; Yount et al. 2016; Kidman 2017). This article, however, shows that girls can exercise agency by taking proactive steps to escape IPV inflicted within child marriages. We examine the factors that can help child brides to leave violent marriages in a low resource context. These data can provide insight into protective factors and supports that can be built or reinforced to empower girls in Uganda and similar contexts to recognise and swiftly escape violent intimate partner relationships during adolescence and later in life.

In the Ugandan local context, the term marriage is used fluidly to refer to both formal and informal unions. Formal unions include those that are customarily/traditionally recognised or conducted by religious leaders and governmen<sup>t</sup> o fficials. In informal unions the couple live together as husband and wife, without legalising the relationship through any of the formal mechanisms (Uganda Ministry of Gender, Labour and Social Development (MGLSD 2015a)). Borrowing from the local conception of marriage in Uganda, we define child marriage as the formal or informal union of girls below the age of 18 years.

### **2. Analytic Framework**

We have utilised structuration theory and the agency-structure debate to contextualise the experiences of the young women in their marital relationships and the attendant decisions that they made during the relationships (Giddens 1984; Sewell 1992; Wendt 1987). Giddens recognises the primacy of human agency in influencing the structures in place, conceived metaphysically (Giddens 1984). Structures are defined as entailing rules and resources (Giddens 1984), with the latter further expounded by Sewell (1992) as constituting inanimate and animate objects existing naturally or manufactured to control and lay claims on power. Human resources, on the other hand, are conceived of as dexterity, physical strength, knowledge and emotional intelligence (commitment), which perpetuates the accessibility and control of power. Giddens (1984, p. 9) asserts that "agency refers not to the intention people have in doing those things but to their capability of doing those things in the first place". This implies that power influences the situation and also suggests that intention alone is not equal to agency, as agency implies action, and can also bring about unintended consequences. Although Giddens recognises human agency and its position in societal interactions (Smith 1998; Turner 1986) and the influence on social structure, other scholars subscribing mainly to realist social theory schools, such as Margaret Archer (2000, 2003), tend to critique this view and that of symbolic interactionism and emphasise instead the critical (and independent) role of structure in regulating human behaviour and its constraining influence on agency.

Sewell (1992) points out that there are variances in the stock of agency across societies and within any given society, indicating how occupations of di fferent positions in society determine di fferent accessibility to, and exercise of, agency. This analysis by Sewell enables the appreciation of the situation of young women in a violent relationship, where their accessibility to power and resources to negotiate freedom was not simple due to the lack of control over the rules and structures prevailing in their marital homes. The current study lends credence to the literature by recognising the *agency, strengths* and resilience of the young women, while recognising the limitations arising from structural constraints (see Archer 2003; Dessler 1989; Sewell 1992). Our study further suggests that the fact that the young women leave violent relationships is a demonstration of agency which shows their e fforts to positively control the direction of their lives. However, the failure of the relationships and limited options for escape may sugges<sup>t</sup> constraint on the part of the young women's agency (agential powers) to exert control over social structures.

### **3. Materials and Methods**

### *3.1. Study Design, Setting and Population*

The article draws on data from a qualitative study of issues, beliefs and experiences of child marriage and gender-based violence (GBV) in Uganda. The study adopted a cross-sectional qualitative design, where data were collected at one point in time. Qualitative methodology is sensitive to unique personal experiences, perceptions, beliefs and meanings of individuals and was therefore considered to be the most appropriate approach for exploring the lived experiences of child marriage survivors. The study was conducted in a total of six districts, four of which were located in central Uganda (Kampala, Wakiso, Masaka and Nakasongola) and two (Gulu and Amuru) were in northern Uganda. The districts were targeted for their experience of military conflict, location (rural and urban) and the perceived risk and prevalence of child marriage.
