Unequal power relations and partner violence against women in Tanzania: A cross-sectional analysis 16 Studies in Human Society 1608 Sociology 16 Studies in Human Society 1699 Other Studies in Human Society

Background: Research on factors associated with partner violence against women is often framed within the context of gender inequality and power imbalances between husbands and wives - inequalities that are considered products of broader structural systems. Tanzania, a patriarchal society where high levels of partner violence exists, has gone through rapid economic and social changes over the past two decades. Increasing numbers of women are seeking paid work, and men's ideals of manhood have reshaped with evidence of extra marital relations and alcohol use. Nationally representative population-based data documents 46.2% of ever-married women have experienced physical or sexual partner violence in their lifetime; 29.6% in the past year. In order to plan appropriate interventions to end violence against women, factors consistently associated with abuse need to be understood. Methods: This study uses "couples" data from the 2015 Tanzania Demographic and Health Survey to examine correlates of past year partner violence against women. Multivariate regression analysis was used to explore individual and relational-level variables - including socio-demographic characteristics and history of abuse among women, partner behavioural characteristics, and indicators of gender and economic inequality - among 1278 married and cohabiting couples. Results: At the individual level, women's experiences of non-partner violence (sexual abuse by a non-partner and witnessing violence in childhood) was strongly associated with risk and highlights that all forms of violence against women serve to keep them subordinated. Partner behavioural characteristics (polygamy and problematic alcohol use) were also associated with risk. Household socio-economic status, however, was not significantly associated with women's risk in the final multivariate model. At the relational-level, men's age difference of 10 or more years; and any employment (compared to none/unpaid) were associated with lower risk. When considering attitudes tolerant towards wife abuse, the strongest association with risk of violence was when both partners held tolerant views. Conclusion: The findings support the assertions of violence being associated with women's prior/additional experiences of abuse and with men's harmful expressions of masculinity. In addition to interventions that focus on transforming gender norms and attitudes (at the individual and community levels), addressing economic, legal and political structural barriers are also required.