Gendered childcare norms — evidence from rural Swaziland to inform innovative structural HIV prevention approaches for young women

Addressing discriminatory gender norms is a prerequisite for preventing HIV in women, including young women. However, the gendered expectation that women will perform unpaid childcare-related labour is rarely conceptualised as influencing their HIV risk. Our aim was to learn from members of a rural Swazi community about how gendered childcare norms. We performed sequential, interpretive analysis of focus group discussion and demographic survey data, generated through participatory action research. The results showed that gendered childcare norms were firmly entrenched and intertwined with discriminatory norms regarding sexual behaviour. Participants perceived that caring for children constrained young women’s educational opportunities and providing for children’s material needs increased their economic requirements. Some young women were perceived to engage in “transactional sex” and depend financially on men, including “sugar daddies”, to provide basic necessities like food for the children they cared for. Our results suggested that men were no longer fulfilling their traditional role of caring for children’s material needs, despite women’s traditional role of caring for their physical and emotional needs remaining firmly entrenched. The results indicate that innovative approaches to prevent HIV in young women should incorporate structural approaches that aim to transform gendered norms, economically empower women and implement policies guaranteeing women equal rights.


Introduction
Gender-based inequality and discrimination are implicated in women's disproportionate risk of HIV in sub-Saharan Africa, and are considered an important area for structural HIV prevention approaches. In particular, inequalities in sexual relationships, which constrain women's power to negotiate, have been widely acknowledged as contributing directly to women's HIV risk (Jewkes & Morrell, 2010). However, HIV prevention interventions have largely focused on individual behaviour change, for example, encouraging individual men and women to use condoms or abstain from sex. Continuing high rates of incident HIV in young women in Southern Africa indicate these interventions have been largely unsuccessful (Harrison, Newell, Imrie, & Hoddinott, 2010). There is increasing recognition that structural factors, which shape and constrain individuals' behaviour, must be addressed to create environments in which interventions promoting HIV protective behaviours (e.g., condom use) can be successful (Keleher & Franklin, 2008). Gender discrimination is an important structural factor.
One important and pervasive dimension of gender discrimination that has received limited attention in relation to HIV risk, is the expectation that women, including young women, will perform unpaid, childcare-related labour (Harrison, Short, & Tuoane-Nkhasi, 2014). Childcare is highly gendered labour and widely considered "women's work" (Sepulveda-Carmona, 2013). The significant burden of care, both for children orphaned and parents sick as a result of AIDS that has fallen to older women (grandmothers), has received attention, mainly from the non-governmental sector. Campaigns have advocated the need to: recognise and support female caregivers (emotionally and financially through social support and job creation); involve men in caregiving; and promote gender equality (UNAIDS, 2008). That young women, defined in this article as those aged 13-25 years, might share in the burden of caring for children affected by AIDS has received limited attention.
However, many plausible mechanisms exist through which this gendered division of labour might indirectly increase young women's HIV risk via economic marginalisation. For example, devoting time to unpaid work reduces the time women have to engage in employment or income generating activities (Antonopoulos, 2010). Providing for children's material needs may exacerbate women's poverty and increase their financial dependence on male sexual partners (Sherr et al., 2014), in ways that increase the likelihood of "transactional sex" (Fielding-Miller, Mnisi, Adams, Baral, & Kennedy, 2014). For young women who experience pregnancy before completing their education, caring for their own child may contribute to their risk of being infected with HIV, because it is a barrier to them returning to school following pregnancy (Grant & Hallman, 2008). There is also evidence that young women may drop out of school if they are needed to care for the children of adults who become ill, or die, of AIDS-related causes (Sherr et al., 2014). However, staying in school is a well-established protective factor for HIV and has been the focus of numerous HIV prevention interventions for young women (Keleher & Franklin, 2008). This evidence suggests that research to better understand how gender inequality in the division of unpaid childcare related labour influences young women's HIV risk is warranted. Better understanding of this pervasive dimension of gender-based discrimination has potential to provide evidence to inform innovative, structural interventions to prevent HIV in young women and girls.
"Scaling up" innovative HIV prevention approaches for girls and young women is a key component of Swaziland's Umgubudla fast track programme towards an AIDS free generation (GoKS, 2016). Structural approaches, which address discriminatory gendered norms that contribute to young women's HIV risk, have great potential to contribute to this strategic goal (Keleher & Franklin, 2008). Young Swazi women's social and economic roles, in community and domestic settings, vary greatly to those of men. Women have less power to negotiate in sexual relationships, and may risk violence from their intimate partners if they assert agency (Brear & Bessarab, 2012). This gendered imbalance in power is implicated in and demonstrated by women's, and particularly young women's, disproportionate risk of HIV infection (Shannon et al., 2012) (e.g., in females aged 18-24 years it is 26%, compared to 5% for males in the same age group (Bicego et al., 2013)). Women, including young women, perform most of the unpaid care-related labour and may be expected to care for their own or other people's children (Nhleko, 2009). The need to support child carers in Swaziland has been acknowledged with the establishment of "neighbourhood care points" (facilities that provide children meals and other services) in over 1 500 Swazi communities (UNICEF Swaziland, 2012). However, the neighbourhood care points model itself relies heavily on women providing their labour voluntarily (Brear, Shabangu, Fisher, & Keleher, 2016). The pervasive expectation that women will prioritise care-related labour (Nhleko, 2009) suggests that like other young women in sub-Saharan Africa (Sherr et al., 2014), young Swazi women's human development potential, including their educational attainment, may be constrained by the performance of care-related labour.
Although ubiquitous and internationally recognised as a barrier to women's human rights (Sepulveda-Carmona, 2013), the gendered division of care-related labour has received limited attention vis-à-vis HIV prevention. However, there is an increasing reliance on women to perform unpaid childcarerelated labour in settings like Swaziland where HIV is highly prevalent (Akintola, 2010;Harrison et al., 2014). In some instances this reliance is perpetuated through international policies, for example, those advocating community-based care of children in family settings rely heavily on women's unpaid care-related labour (UNICEF, 2007). Given this increasing reliance, understanding how performing unpaid childcare-related labour might influence women's risk of HIV is important. Our aim was to learn from members of a rural Swazi community caring for children affected by AIDS, about how gendered childcare norms and expectations influence young women's HIV risk and might be addressed by innovative, structural approaches to HIV prevention.

Methods
This article reports sequential analysis of data related to gendered childcare norms and young women's HIV risk. The data were generated through a qualitativelydriven mixed-method design (Morse & Cheek, 2014), participatory action research study about health in a rural Swazi community caring for children affected by HIV. The study was approved by the Swaziland Scientific and Ethics Committee (MH599C) and the Monash University Human Research Ethics Committee (CF13/994 -2013000486). Community-level consent was obtained from the local governance authority (umphakatsi) and the community preschool which was the partner organisation in the participatory action research study.

Methodology
Participatory action research is an approach that engages members of the study community in all aspects of the research process (Israel, Eng, Schulz, & Parker, 2013). In this participatory action research the community was purposively selected by the second author MB because she had been participating in community development interventions for approximately five years. A history of participation in the study community is considered an essential foundation for participatory action research, and the key criteria for selecting a study community (Israel et al., 2013). The study engaged a group of 10 community researchers (including the first author PS) in all aspects of the research process (as defined by Brear, Hammarberg, & Fisher, 2017). The group comprised 4 males and 6 females aged 18-40 years at the beginning of the 14-month long study. The co-researchers were selected purposively for age and sex diversity, from amongst a group who submitted an expression of interest to participate in the study. These co-researchers participated in learning about research, co-designing (with MB) the study (including instruments, sampling techniques and ethical procedures) and collecting and analysing data. They also participated in implementing actions in their community, based on the results of the research.

Research design -methods and research instruments
The mixed-method participatory action research involved focus group discussions (the core method) and a demographic and health census survey (the supplementary method). The focus group discussion schedule included three core questions: about the benefits, problems and solutions to problems in the community. Focus group facilitators pre-developed a series of probing questions to stimulate further discussion about each of the core questions (e.g., what are the benefits/problems compared to urban areas and what are the solutions that could be implemented with/without help from outside the community). Facilitators were also given latitude to follow up with spontaneous probes they developed in response to the information participants provided (e.g., to ask for solutions to problems identified by participants, or to ask other participants to share their opinions on a benefit or problem identified by another participant).
The survey included two levels of data, individuallevel demographic and household-level health-related information. It primarily included standard, fixed choice demographic and health indicators (e.g., from the Swaziland Demographic and Health Survey (CSO & MII, 2007)). It also included context-specific fixed choice questions (e.g., demographic variables to collect data about members of the family who lived away from the household most of the time), open-ended questions about problems in the community and a comments section in which the survey enumerator recorded additional relevant information. The survey was developed in English, translated to siSwati by community researchers (including PS) and back-translated to English by a research assistant from outside the community who had not seen the original version of the survey. Following pilot testing and modification, inter-rater reliability testing showed 98% overall agreement between three enumerators.

Population, sampling and data collection
The population for the census was all households within homesteads in the geographic community. In rural Swaziland homesteads are units of land allocated to a married male (the homestead head) by community headmen, and extended family social units (Russell, 1993a). As a social unit a homestead may include more than one household (a social unit including people who eat from the same kitchen or "pot"). Each household is typically headed by a relative of the homestead head. The population was identified by the community researchers through participatory mapping. The accuracy of the maps was checked by tindvuna (community headmen who allocate land to homestead heads). The homesteads on the final map formed the population for the census. To collect the census survey data a pair of community researchers visited each homestead and invited the head to participate. Those that agreed were asked to identify all households in their homestead, and for permission to interview the head cooker (defined as the person in charge of domestic work, including childcare) of each household. If permission was granted, the head cooker was invited to participate in the household survey, which was administered in siSwati by face-to-face interview, if the head cooker consented. A bar of soap was provided to all survey respondents as a token of appreciation for the time and knowledge they contributed to the study.
The head cooker was asked to indicate if they consented for household members to be invited to focus group discussions. Members of households where the head cooker consented, formed the sampling frame for the focus group discussions. To include perspectives of male and female community members of different ages, and construct groups in which members had similar age and sex characteristics, the list of household members was stratified into groups by age (13-17, 18-25 and over 25 years) and sex for sampling. Ten potential participants from each strata were selected from a hat. In addition, potential key informant participants (community health workers, community headmen/governors and teachers) were identified and invited to participate. All focus group discussions were facilitated by the community researcher who most closely matched the participants' age and sex characteristics, in siSwati, in the classroom of the community preschool that was the partner organisation for the study. The location was selected because it was located approximately at the centre of the community. All participants provided written informed consent and were provided SZL10 (~US$1) as a token of appreciation.

Data preparation and analysis
Focus group discussion audio files were translated twice, once in handwriting by a community researcher (including PS) and once, typed, by a research assistant from outside the community. Discrepancies were highlighted and reconciled by the community researchers (including PS). Identifying information was removed and PS allocated each participant a pseudonym, which was used for referencing quotations (in the format "Pseudonym-sex-age" or "Pseudonym-key informant group-sex"). The qualitative data (including survey comments) were imported into NVivo version 10 (QSR International, Victoria, Australia) to facilitate data analysis.
Primary content analysis of the entire qualitative data set highlighted that gendered discrimination, including in the performance of care-related labour, was a salient problem in the community (Brear et al., 2016). We therefore performed qualitatively-driven, sequential analysis, that is, a focused further analysis of the same data, using a different analytical technique (Simons, Lathlean, & Squire, 2008). The sequential analysis we present here focused on care-related labour and young women's HIV risk. It used an interpretive analytical approach, which situated ordinary people as possessing expert knowledge about the meanings of their experiences (Denzin, 2001). Narratives were not considered absolute truths, but unique perspectives which through theoretically-informed interpretation, could provide novel insights into complex social arrangements which may go unnoticed by outside professionals (Stringer, 2003). Using this framework we conducted interpretive analysis of data coded to gender discrimination, focusing on key quotations (exemplars or epiphanies) and analysing them in relation to the broader data to reveal implicit meanings for young women's HIV risk (Denzin, 2001).
The analysis was informed by the capabilities approach to women's human development (Nussbaum, 2000). This theoretical framework outlines 10 capabilities considered essential for healthy human development, and fundamental entitlements for every human being. These capabilities are: full length life; bodily health; bodily integrity; senses imagination and thought; positive emotions; practical reason; respectful affiliation; coexistence with other species; time to play; and control over one's environment. It highlights the ways in which gendered norms, including the expectation that women will provide childcare without pay, constrain women from attaining these entitlements. We considered the capabilities approach particularly suitable for this analysis because of its explicit acknowledgment of the ways in which gendered norms and expectations within the family (i.e., the setting in which most children are cared for) prevent women achieving capabilities and their human development potential.
Quantitative survey data were entered to a database with equivalent English and siSwati screens, and numerically coded. We conducted focused analysis of relevant quantitative data using univariate statistics and cross-over analysis (i.e. the use of qualitative techniques like textual description to analyse quantitative data) (Onwuegbuzie & Combs, 2010). The point of interface (Morse & Niehaus, 2009) for the quantitative and qualitative findings (i.e. the point at which the results were brought together) is the ensuing results section.

Results
Ninety-nine per cent of the 126 invited homestead heads agreed for the head cooker/s (person in charge of domestic work, including childcare) of each household in their homestead, to be invited to participate in the household survey. All the 151 invited head cookers provided consent, completed the survey, and gave permission for other household members to be invited to participate in focus group discussions. Collectively they provided demographic information for 1 031 household members, including 331 who lived away from the household most of the time (e.g., for work or school). Of the 120 potential participants invited to attend 1 of 12 focus group discussions, overall three-quarters actually participated (53% of males, 83% of females and 73% of key informants). Each focus group included a mean of 7 (range 4-11) participants and lasted for a mean duration of 61 (range 48-71) minutes.

Gendered expectations regarding childcare
The results demonstrated that women in the community were expected to care for children. For example, a 59-yearold father, who was also the head cooker of a household with neither adult women nor children living in it, commented that he decided to move his children to live with their grandmother after his wife died because children need a mother's care (comments-household survey-026). Another man reported he was dissatisfied with women taking positions in committees because, "Tomorrow this person will have to look after a sick child instead of working… [when you ask her to work for the committee], she will, say, 'my child is sick'" (Themba, male, 53 years).
Although it was not their traditional role, providing financially for children was, at least some times, conceptualised as women's (not their male partners') responsibility. For example, one participant said, "very young children that you don't even imagine they can take care of another child, you hear that she is pregnant. It's a big problem because… they expect help from us… we have to give her clothes. You can't ignore her" (Nombuso, female, 18 years) and another, "I wish I could beat the mother [for not providing clothing for her child]" (Lindiwe, female, 22 years). This tendency to blame the mother was paradoxical, given that the young female participants also recognised the ways in which gendered norms and expectations, and poverty influenced young women's sexual behaviours (discussed further below).

Gendered norms in the performance of childcarerelated labour
The expectation that women would take responsibility for the care of children was reflected in the actual division of care-related labour in the community. Unpaid childcarerelated labour at schools and community facilities was highly gendered. For example, being a volunteer community health worker at a neighbourhood care point (a communitybased facility that provides meals to young children) was reported to be "women's work" (Zanele-Female-36). Women, including young women, were more likely to be nominated as the head cooker than men were (Table 1). The households with female head cookers included more children, and more children who were not the biological children of the head cooker (Table 1). For young women head cookers the non-biological children were typically their siblings, however, there were exceptions. For example, one young woman was looking after two children that her husband had fathered to two other women, as well as two children they had conceived together. All the 4 children were under 5 years of age.

Caring for biological children
Many young women and girls in the community had given birth and were caring for their biological children. At least some had conceived pregnancies in age-disparate relationships (i.e., to men who were much older than them when they were teenagers). For example, one woman had conceived her first child when she was 17 years of age and her male partner, a polygamist, was 43 years old. There were three 3 and no male parents who were under 18 years of age. None of the mothers under 18 years of age had finished or were attending high school. This was perceived to be at least partly due to childcare responsibilities. For example, one participant said, "Most of them drop out from school because maybe they are pregnant, so they will leave school and stay home to raise the child" (Mayibongwe, primary teacher).

Caring for non-biological children
One interesting form of non-biological childcare that the results showed affected young women was the care of children whose mothers had married men who were not their children's fathers. The head cooker in 7 of the 151 households in the community spontaneously reported that they were looking after a child born to a female relative out of wedlock, because their mother had left them after getting married to and moving to live with a man who was not the father of her child/ren. Furthermore, one young woman reported she was caring for her siblings because her mother remarried following her father's death, and moved to her marital homestead to which she could not take her children (comments-household survey-126).
The community researchers who participated in the study (including PS) explained this was because, unlike men, women are not allowed to bring children they conceived with other partners into their marital household. The fact that no participants bothered to explain this norm, indicates how pervasive it is. However, there was also evidence that this discriminatory tradition was changing. The community had three households in which a male head's step-children (or other children who were biologically related to the male head's wife or daughters-in-law, but not the head themselves) were living in and considered members of the household. For example, one respondent reported that she came to her husband's marital homestead with a child she had conceived to another man, before having three children with her husband. She noted explicitly (suggesting it is not a norm) that her husband had no problem with this (comments-household survey-121). The norm of women caring for non-biological children was perceived to have negative and positive effects on young women's lives, which might plausibly influence their HIV risk. One positive effect was that young women and girls whose parents had died might be able to access support and protection through their extended families. For example, one said, "let me say both my parents pass away… my mother's relatives, I can ask to stay with them" (Nokwanda, female, 17 years). Women's willingness to care for non-biological children also enabled young women to leave their children with female relatives (e.g., mothers or sisters) so that they could attend school or move to town and work. Nine households included young women members who lived away for work, whose child/ren lived in the household and were being cared for by another woman. Some participants perceived that working increased women's financial dependence on men and their HIV risk, . The practice of women caring for children they had not given birth to also had negative effects that might increase young women's HIV risk. In particular young women living in the rural community were expected to perform unpaid care-related labour for their siblings' and other relatives' children. Caring for children created a financial burden for caregivers, whose ability to generate income in the community was minimal. For example, one young woman said, Our brothers go to town… and find girlfriends [who] bear children and bring the children here at home and they leave us with their children. Then the one who stays are home will have to find… food for the children (Lindiwe, female, 22 years). Young women were also reported to take responsibility for caring and providing materially for younger siblings if their parents died. This practice was perceived to increase the young woman's HIV risk because it left young women vulnerable and financially dependent on men who expected sex in return for financial support.

Discussion
The results of our study provide evidence of the ways in which childcare-related norms might exacerbate young women's HIV risk by shaping their sexual behaviours, in societies with conservative and discriminatory gendered norms, widespread poverty and high HIV prevalence. The results indicate that in this context, caring for children may indirectly increase HIV risk, by preventing young women and girls from participating in education and work. This increases their financial needs and leaves them financially dependent on men in ways that make transactional sex more likely, and reduce young women's power to negotiate in sexual relationships. The results also suggest that living in a society where gender discrimination, including in childcare norms, is highly prevalent, socialises young women to accept gender inequality. Given the limited attention performing childcarerelated labour has received in HIV prevention approaches for young women (despite recognition that eradicating gender discrimination is essential for ending HIV), these findings have great potential to inform the development innovative HIV prevention interventions.

Strengths and limitations
However, our findings were generated through a case study, conducted in a single community with a specific socio-cultural and economic context. Our mixed-method data set, which was generated using instruments designed in partnership with members of the community, and analysed drawing on their insights, strengthens the findings. Drawing on qualitative and quantitative data enabled us to provide numbers to support the narratives that captured the participants' voices and perspectives. Capturing the voices of diverse community members enabled us to triangulate the analysis by considering the perspectives of men and women of various ages. The results reflect the perceptions of community members directly affected by the problem, about the complex and indirect ways in which caring for children might influence young women's HIV risk. However, the results must be interpreted, and their implications for other settings drawn out, with caution. They do not establish causality, nor is that the aim in this qualitatively-driven, mixed-method study. Rather our aim, within a context where gendered discrimination is recognised as contributing to women's disproportionate risk of HIV infection, was to highlight the ways in which: (1) gendered discrimination in the form of childcare norms, affect young women; and (2) how this type of gendered discrimination might be addressed through innovative HIV prevention approaches. We now turn to discussing the types of interventions that our results indicate might be effective in reducing young women's HIV risk, in Swaziland and other similar settings.

Innovative HIV prevention interventions Transforming gendered norms and expectations
The results of this study indicate that there is an urgent need for HIV prevention interventions for young women and girls to challenge the pervasive expectation that women will care for children without being paid. Challenging this pervasive form of gendered discrimination is not only important for relieving the current generation of young women of their childcare burden and economic marginalisation, it is also important for socialising the next generation of girls and young women to expect better for themselves. The gender discrimination girls and young women are currently exposed to socialises them to expect and accept this discrimination in their own lives (Nussbaum, 2000). They are simultaneously socialised to accept other gender-discriminatory norms which interact with and mutually reinforce the gendered division of childcare-related labour (e.g., acceptance of men, but not women, bearing children with multiple partners). These findings concur with those in the increasing body of literature (Jewkes & Morrell, 2010;Keleher & Franklin, 2008;Mabaso et al., 2017), indicating a need for HIV prevention interventions that challenge and attempt to change discriminatory gendered norms and expectations in the Southern African context. They further highlight the need for these interventions to be included in HIV prevention programmes for young women. Labelled "gender transformative", interventions implemented with the aim of transforming ideologies about the roles of men and women typically focus on providing opportunities for participating men and women to think critically about and actively construct new more gender equitable notions of masculinity and femininity. Gender transformative interventions have been effective in changing attitudes about male power and violence in sexual relationships (Jewkes, Flood, & Lang, 2015). Applied to transforming gendered norms and expectations relating to women's role as child carers, the results of this study indicate that such interventions might encourage critical discussions about: why women are expected to perform childcarerelated labour without being paid; what skills women need to provide children care; who benefits and who experiences detriments from women's unpaid childcare-related labour; what role governments and large organisations (including charities and the corporations that children's parents work for) should play in providing care for children; and what a gender equitable childcare arrangement might look like, in terms of who performs the labour and how much they get paid (by whom) for doing so.
This study suggests that another important related area that gender transformative interventions could address is acceptance of age-disparate and transactional sexual relationships, which directly increase HIV risk for the young women (Cluver et al., 2013). The results of this study indicate that acceptance of older men entering sexual relationships with and sometimes impregnating young women is part of a broader set of gender discriminatory beliefs in which men are entitled to have sex with multiple partners and not required to take responsibility for their offspring, should pregnancy occur. These norms are associated with increased HIV risk for teenage women, and potentially also their children. Teenagers in age-disparate relationships are more likely to fall pregnant, partly because they are unable to negotiate condom use in these relationships (Toska, Cluver, Boyes, Pantelic, & Kuo, 2015). Not using condoms is also an important HIV risk behaviour. Children born to teenage mothers also have increased health and social risks, including, for example, being less likely to complete school (Fall et al., 2015), which is also a risk factor for HIV infection.
Although women in this study were sceptical, evidence from the region suggests value in gender transformative interventions that work with men and boys (Jewkes et al., 2015). One important norm that this study highlights gender transformative interventions might address, is the responsibility and living arrangements for children born out of wedlock. The norm of women leaving their children with other caregivers if they subsequently married a man who is not the father of their children, extends a Swazi tradition which considers children the property of men. According to this Swazi custom, fathers have a right to claim ownership of their offspring, by payment of cattle to the mother's father (the maternal grandfather). This arrangement leaves women free to marry other men. However, there is no onus on fathers to claim responsibility for their children and unless they do children remain the property and responsibility of their maternal household. Although men still maintain these rights under Swazi customary law, it is increasingly uncommon for men to claim ownership of and responsibility for offspring born out of wedlock (Russell, 1993b). These traditional childcare arrangements are concerning in relation to the perceived instability of relationships in which young women become pregnant to "sugar daddies" (i.e., in relationships with older, often married men, involving transactional sex). If older men do not marry the young women they impregnate, these women are likely to eventually marry other men, later in life, and may be forced to enact the gendered childcare norm that separates them from their children. That several men in this community were challenging this tradition by living with step children is encouraging. It highlights the potential for changing gender discriminatory ideologies and practices in men and boys, which is an essential component of achieving gender equality (Jewkes et al., 2015).

Reducing young women's economic marginalisation
Although they hold great promise, ideologically transformative interventions are unlikely to be effective in the short term, nor in the absence of practical interventions that provide women with actual assistance and experiences of more gender equitable childcare arrangements (Nussbaum, 2000). Disparities in paid employment are an important form of gender discrimination, and are exacerbated by women performing unpaid care-related labour (Antonopoulos, 2010). From a human development perspective, expecting women to perform childcare-related labour without pay, denies them the ability to control their environment and to affiliate in respectful ways, which includes non-discrimination and dignity at work (Nussbaum, 2000). Evidence indicates that paying women to perform care-related labour has both ideological and practical benefits. It situates childcare as a social, rather than an individual responsibility, and women's work as valuable (Brennan, 1998). There is also some evidence that creating jobs for women increases their self-efficacy and participation in community political life and contributes to more equitable gender relations (Breuer & Asiedu, 2017). However, there may also be detrimental HIV-related effects of employment. There is evidence that in some contexts employment increases women's risk of intimate partner violence (Devries et al., 2013), which is in turn associated with HIV risk. The participants in this study perceived that relocating to an urban centre for employment, and the expenses associated with that, might increase HIV risk via transactional sex, indicating that job creation targeted in rural communities might be most effective.
However, the results of this study concur with evidence showing Swazi women who cannot access basic necessities (particularly food) for their children may enter "transactional" sexual relationships (Fielding-Miller et al., 2014) where the direct risk of HIV infection occurs. Creating jobs for women has the immediate effect of reducing women's economic marginalisation (Budlender & Lund, 2011), which our results indicate may reduce HIV risk, by reducing the need for women to engage in transactional sex. A strong social welfare system could reduce young Swazi women's HIV risk by reducing their economic marginalisation. For example, in South Africa receiving a cash transfer was associated with reduced transactional sex and involvement in age disparate relationships for adolescent girls (Cluver et al., 2013). Interventions that encourage or enforce men to take responsibility for the material and financial needs of children they father could also address women's economic marginalisation.
Creating and acting on policies and legislation to reduce gender discrimination Legislation that requires non-custodial parents to pay maintenance to custodial parents is already in place in Swaziland, and the Constitution guarantees women equality.
From the human development perspective that has informed our analysis, these legal documents are of little value if they do not enhance what "people are actually able to do and be" (Nussbaum, 2000, p. 5), or their "substantial freedom to choose and act" (Nussbaum, 2011, p. 24). The results of this study suggest that women, including young women, have little freedom to choose when it comes to sharing the burden of childcare-related labour with men or receiving financial support from men who father their children.
Another area of gendered discrimination that must be addressed to achieve gender equality in practice is in access to education. This study demonstrated that young women's equal access to education is constrained by pregnancy and subsequent childcare, a phenomenon that only affects young women. Both pregnancy and the need to care for children born to women of school-going age are recognised more broadly to increase school drop-out (Grant & Hallman, 2008) which is associated with increased HIV risk (Keleher & Franklin, 2008). Interventions which enable and encourage young women to stay in school during pregnancy and return to school after child birth could reduce this gendered form of school dropout, and in turn young women's HIV risk.

Conclusion
The results of this study show that the division of child-care related labour in our study community was highly gendered. In this context of poverty and limited opportunity, caring for children contributed to young women's HIV risk by increasing their economic dependence on men and gendered power inequalities in sexual relationships, where the direct risk of HIV infection occurred. These included transactional sexual relationships with sugar daddies. Caring for children was also shown to reduce young women's opportunities to participate in activities that might develop their human potential, for example, education, and to socialise them to accept and expect gendered forms of discrimination. The results strongly indicate that interventions to address gender norms and expectations regarding who will perform childcare-related labour and how much they will get paid for it will be an important component of innovative structural HIV prevention strategies for girls and young women in Swaziland. Although specific to a small community, given the ubiquity of norms and expectations regarding women's roles as children's carers in Swaziland, sub-Saharan Africa and internationally, the results of this study provide theoretical insights for preventing HIV in many other settings.
Our results suggest that gender transformative interventions, which actively challenge the notion that women should care for children without being paid, are needed to prevent HIV. However, these ideological interventions will likely be most effective if they are implemented alongside interventions that: economically empower women by increasing their incomes and financial independence; create opportunities for women to participate in paid work, including by creating jobs; and creating environments that support the implementation of legislation and policy promoting gender equality and women's rights. The results we have presented suggest that the success of these types of structural interventions will be a prerequisite for enabling young women to make healthy choices about the sexual behaviours which directly influence their risk of HIV. They also have great potential to contribute to more gender-equitable social relations, an important goal in itself.