Endemic poverty, gender-based inequalities including intimate partner violence (IPV) and economic migration constrain the effectiveness of HIV prevention efforts in much of southern Africa. The link between these structural factors and the behavioural and biological processes that fuel HIV transmission is mediated by complex systems of social norms, networks and relationships. Social capital has been put forth as a theoretical construct to guide assessment of such dynamics. Between 2001 and 2004, we implemented a prospective cluster randomized trial of a structural intervention for the prevention of HIV and IPV – the Intervention with Microfinance for AIDS & Gender Equity (IMAGE Study). The intervention combined a microfinance-based poverty alleviation programme with a gender and HIV training curriculum. Through working with groups of the poorest women in target communities, the intervention was an attempt to engage structural dimensions of the epidemic, in an attempt to stimulate individual and collective responses to priority concerns. The training component of the intervention emphasizes leadership, communication, and community mobilization around priority concerns. The delivery of the intervention through group-based microfinance, with its emphasis on trust and solidarity, has the potential to further stimulate social capital – “the system of norms, networks and trust relationships that can improve the efficiency of society by facilitating coordinated action.” This dissertation examines changes in social capital in the context of IMAGE, and the associations between social capital and health outcomes, in this case IPV and HIV infection. A multidimensional assessment of social capital based on the World Bank’s Social Capital Assessment Tool was conducted at baseline and after two years among households participating in the intervention and matched comparison households from control villages. Quantitative data on both structural and cognitive social capital examined changes in social capital. A portfolio of qualitative research described the nature and extent of these changes. Associations between household social capital and health outcomes were assessed among two cohorts: a group of older women (IMAGE participants) and matched controls where relationships to empowerment and IPV were examined (cohort 1), and; a group of 14-35 year old household members (cohort 2) in whom HIV-related psychosocial attributes, risk behaviour, HIV prevalence and incidence data were collected. 843 cohort 1 women were interviewed at baseline with an average age of 42 years. Two year follow-up rates were 87%.1455 14-35 year old cohort 2 males and females were interviewed at baseline with an average age of 22 years. Two year follow-up rates were 73% in this group. Among direct programme recipients, effect estimates suggested positive changes in pre-defined indicators of structural and cognitive social capital relative to matched controls, including social networks (aOR 1.85 95% CI 0.95-3.61), perception of community solidarity (aOR 1.65 95% CI 0.81- 3.37), collective action (aOR 2.06 95% CI 0.92-4.49), and reciprocity/community support (aOR 1.11 95% CI 0.38-3.24). Qualitative data suggest economic and social gains from IMAGE enhance participation in both financial and non-financial social networks in the community. Microfinance loan groups, when functioning well, were an important source of bonding social capital, providing emotional and material support, collective identity and normative guidance. Bridging social capital was evidenced by numerous examples of individual and collective mobilization – linking loan centres to wider community structures. However, repayment problems, leadership struggles, and gossip were obstacles to these efforts in some instances. Among cohort 1 women, there were few associations between social capital and IPV at baseline. However, at follow-up, after adjusting for socio-economic status (SES) and potential confounders, there were significant positive associations between greater cognitive social capital and 5/6 empowerment indicators, and borderline associations with lower levels of IPV. In 4/6 cases, these
relationships held after adjusting for the confounding exposure to the IMAGE intervention. Higher levels of structural social capital were significantly associated with lower levels of IPV (aOR 0.51 95% CI 0.25-0.99), 3/6 indicators of empowerment, and borderline associations with 2/6 additional empowerment indicators. However, interaction tests suggest relationships between structural social capital and risk of IPV were strongly influenced by the presence of the intervention. Among 572 males from cohort 2, there were few associations between social capital and vulnerability to HIV at baseline. However, at follow-up, after adjusting for SES, potential confounders and residing in an IMAGE household, there was substantial evidence that males from household with higher levels of cognitive social capital were less vulnerable to HIV infection, including lower HIV prevalence (aOR 0.39 95% CI 0.15-0.99), higher levels of condom use consistently (aOR 2.4 95% CI 1.13-5.06) and at last sex (aOR 2.15 95% CI 1.11-4.14), greater levels openness to discuss sex in the home (aOR 2.18 95% CI 1.24-3.82) and lower self-perceived HIV risk (aOR 0.44 95% CI 0.26-0.74). There were no associations between household levels of structural social capital and HIV risk in this group. Among 883 females from cohort 2, higher levels of household cognitive social capital were associated with protective HIV-related psychosocial mediators at both baseline and follow-up. At follow-up there were also associations with consistent condom use in adjusted models (aOR 1.95 95% CI 1.03-3.69). Relationships with structural social capital were more complex. No baseline associations were evident. Follow-up data suggested strong associations between greater structural social capital and protective risk behaviour including condom use consistently (aOR 2.78
95% CI 1.52-5.09) and at last sex (aOR 1.75 95% CI 1.1-2.77), alongside generally protective HIVrelated psychosocial attributes. However, after adjusting for SES, confounders and exposure to IMAGE, HIV prevalence (aOR 1.83 95%1.04-3.2) and incidence (aOR 2.9 95% CI 1.2-7.1) were both significantly increased in the group of young women from households with wider social networks. There was strong evidence to suggest that the effect of the IMAGE intervention was to enhance the strength of associations between social capital and reduced vulnerability to IPV and HIV. : This research suggests that even in the short term, microfinance-based poverty eductions programmes can indeed generate social capital. Higher levels of social capital at the household level were linked to reductions in vulnerability to IPV in older women and to HIV in younger male co-residents. However, among young women while higher levels of social capital may positively influence protective attitudes and risk–behaviour, exposure to wider social networks may also serve to increase HIV transmission.