Unequal power relations and partner violence against women in Tanzania: a cross-sectional analysis

Violence against women is widely accepted as a human rights violation and public health concern [1, 2]. The most common form of violence against women is that perpetrated by men towards their female partners, and prevalence estimates suggest that globally, one in three women have experienced physical or sexual violence by an intimate partner (generally defined as a current or former spouse or cohabiting partner) in their lifetime [3].

Partner violence against women is more prevalent in patriarchal societies and research on associated risk factors is commonly framed within the context of unequal power relations that emphasise men’s and women’s roles, and assert men’s dominance over women [4, 5]. These gendered inequalities are theorised to be products of broader structural systems—political (e.g. lack of gender responsive policy making), legal (e.g. inadequate provision of legal and social services), and economic (e.g. unequal access to education, economic resources and employment opportunities)—that reinforce the disadvantaged status of women at both the community and the individual levels [6, 7].

Since the World Health Organization’s (WHO) 2000–2003 multi-country study on domestic violence and women’s health (WHO study) [1, 8], there has been an expansion of studies that have explored, besides prevalence and patterns of violence, male and female factors and their associations with partner violence against women. These studies have advanced understanding about the role of, in particular, individual level factors on gender relations and the mechanisms through which these factors shape women’s risk of partner violence.

Poverty or low household socio-economic status (SES) has been consistently found to be associated with higher rates of partner violence against women in low- and middle-income countries (LMIC) and in high income settings [9–12]. In a systematic review of published evidence from LMIC, in fifteen out of sixteen settings, household wealth (measured by ownership of durable assets) was found to have a protective association with women’s risk of past year physical or sexual violence, although, this association was significant in only eight settings and highlights that household wealth is not universally protective [9]. Early theories, family stress theory, argued that the inherent stress of poverty is the mediating factor that leads men to be violent towards their wives and female partners [11]. Poverty stress is further intensified in settings where ideals of successful manhood firmly place men to be the household’s main provider [12]. In such settings, limited or poor employment options for men may then lead to feelings of anxiety and despair and a crisis of male identity ensues. In some LMIC, including in Tanzania, it has been observed that feelings of economic disempowerment among men has resulted in a reshaping of masculine ideals that involve the excessive use of alcohol and relationships with other women, both of which have been found to significantly increase women’s risk of violence [5, 13–17].

The concept of successful manhood brings to the fore the tandem notion of “successful womanhood” that traditionally lay in reproductive responsibilities such as bearing children and especially sons, as well as in maintaining family values and family harmony [18, 19]. Transgressions of good or appropriate wifely behaviour include women’s use of alcohol, relationships with other men and displays of autonomy. While aspects of women’s empowerment such as education, economic independence and ownership of capital assets have been found to be protective in some settings, it has been found to have a risk association in others [9, 20]. Within the context of poverty, women’s financial contributions can ease financial stresses within households. A competing view, relative resource theory, however, asserts that economic (e.g. employment or income) or status (e.g. educational attainment) differentials that favour women over men increases a woman’s risk of violence because of challenges to established gender norms [21, 22]. So if a woman is working when her husband or male partner is not, then this may confer a risk onto women if this unequal status fuels men’s feelings of inadequacy [5, 21].

Another factor that has the effect to disadvantage women is early or other forms of abuse. Women’s early experience of violence (either childhood violence or witnessing their mother being beaten) may reinforce notions of inferiority and acceptance of abuse by a partner [5]. By contrast, men who witness violence towards their mothers or who were beaten themselves as children are more likely to become perpetrators of violence [17, 23, 24].

This study used the 2015 DHS data for “matched” couples in Tanzania. The DHS uses a multistage sampling method to select a nationally representative sample of households [36]. In the first stage 608 enumeration areas or “clusters” were selected with a probability proportional to size, from all 30 regions of the country (25 from mainland Tanzania and 5 from Zanzibar). Within each cluster, 22 households were randomly selected. A household survey, which included a listing of the names, ages and sex of all resident individuals, was administered and completed in 12,563 (of 13,360) households. In each selected household, a Woman’s questionnaire was administered to all eligible women (ages 15–49 and resident) from which one randomly selected woman received the module on domestic violence. In one-third of the selected households, all eligible men (ages 15–49 and resident) were interviewed using the Man’s questionnaire. In both questionnaires (Woman’s and Man’s), respondents were asked if they are married or living with someone as if married and if yes, the name of their partner. The respondent’s spouse/partner was then identified from the household listing and their unique household line number recorded. The DHS “couples” dataset links the two data files (women’s and men’s surveys) based on whom the respondents name as their partner.

Data were used for the 1278 matched couples where the domestic violence module had been administered to women (see Fig. 1). Verbal informed consent was obtained from all individual respondents included in the study who were able to accept or decline to participate.

Almost 40% of women in this “matched” sample reported that they had experienced physical or sexual violence by an intimate partner in their lifetime and past year prevalence was 33%. These prevalence estimates have not notably changed since the first population-based prevalence estimates documented by the WHO (in 2002) when 40% of currently partnered women in Dar es Salaam and 55.6% in Mbeya reported lifetime experience of physical or sexual partner violence and 20.8 and 31.0% reported partner violence in the past year [39].

Several key factors were found to influence women’s risk of experiencing violence at the hands of their partners. Among factors relating to the woman, a significant risk association was found with exposure to violence in childhood (witnessing mother being hit). Intergenerational exposure to violence has been consistently found to increase women’s risk of partner violence in studies from Tanzania and in studies across cultures [16, 40]. This highlights the cyclical nature of violence that serve to disadvantage women by conferring vulnerability in later life. Tolerant attitudes towards wife beating, an indicator of women’s low social value and the extent of male hierarchy that exists within society, measured from the perspectives of the woman and the man, were found to increase women’s odds of experiencing partner violence, and in the relational model, the strongest association with violence was found when both partners held tolerant views. Women’s young age at sex initiation (17 years or lower) was also found to elevate women’s risk of past year partner violence. A study in Rakai District, Uganda, documented the same result and argued either early sexual onset has a disempowering effect on women who are then less able to protect themselves against violence later in life, or that women who become sexually active early are self-selected for subsequent abusive relationships [41].

Having given birth to one or more children elevated women’s risk of past year partner violence compared to not having given birth, although interestingly, the strong increase with higher parity that was seen in the bivariate analysis does not remain in the individual or relational models. A study in Viet Nam documented similar findings in terms of having children or not [17]. It could be hypothesized that there is a potential link between increasing number of children and a decrease in potential for women to engage in employment, and that the observed association is a reinforcement of the structural norms which serve to keep women dependent on their partners. Some studies, however, have suggested that high parity and unintended pregnancy is a consequence of violence, rather than a risk factor for violence, and is related to women’s lack of ability to control their fertility [42, 43].

Although marginally insignificant in the multivariate model, the risk association with women’s sole ownership of land and/or housing, which was significant at the bivariate level, is counter to the supposition from economic theory that ownership of capital assets empowers women to negotiate less violence—as has been found in Kerala, India [44]. Women’s independent wealth is hypothesised to raise women’s bargaining power or to facilitate women’s ability to leave a violent relationship by lowering their “threat point”—the threshold at which a woman evaluates that her welfare is better outside of the household [45]. Analysis of the WHO study data from Tanzania found that ownership of capital assets did not have the empowering effect to enable abused women to leave a violent relationship [46]. Exactly why ownership of a capital asset increases women’s risk of partner violence is not clear. It may be that women who own assets are more likely to be confrontational or that it is a transgression of gender norms and men use violence to reassert their dominance within the household. Alternatively, it may be that abused women who are able to, invest in capital assets with the view to leave a violent relationship at some point in the future. Greater household SES, usually measured by ownership of assets in LMIC, has generally been found to be protective (although not always significantly) [9, 16]. In this study, a significant protective association with asset wealth was found only at the bivariate level and with only the highest asset quintile. This could reflect a greater difference in wealth between the top one-fifth of households and the rest.

Among men’s characteristics, polygamy and alcohol use—expressions of masculinity that have increased with the structural economic reforms and work insecurity—were both significantly and positively correlated with women’s risk of partner violence, and both factors have been consistently found to be associated with women’s risk of violence in Tanzania and elsewhere [14, 17, 39, 47]. Reasons for the finding with polygamy could be that women have less power and are more likely to be dependent on their husbands, thus raising their vulnerability, or because women are likely to chafe at this expression of traditional privileges, as found in Moshi, Tanzania [50]. While several theories have been advanced to explain the relationship between men’s alcohol use and partner violence against women, an in-depth analysis from fourteen sub-Saharan Africa countries concluded that the most likely causes for the significant correlation (in all fourteen countries) were behavioural disinhibition—that alcohol use impairs cognitive functioning and increases aggression—and relationship dissatisfaction [48]. Although an independent risk factor by itself, men’s excessive alcohol use in Tanzania has been argued to have arisen in part because of men’s increasing frustrations at not being able to meet their gendered role expectations as the household’s main breadwinner [12].

In this study, compared to not working, the odds of partner violence against women was significantly lower when the man was working and a lower odds ratio (albeit slightly) was observed among men in stable paid work. Interestingly, and in line with family stress theory, in the relational model both the woman and the man working in the same category of employment (either both stable paid or both seasonal paid) and when the man is in stable paid and the woman in seasonal employment (i.e. the man has higher employment status), reduced the odds of women experiencing violence.

Several limitations underlie this study which must be borne in mind when interpreting the results. The first is that the cross-sectional design of the study means that it is not possible to determine the directionality of relationships between many of the variables and partner violence. A second limitation of this study is that the analysis is limited by the variables provided which have not necessarily been collected with this analysis in mind. For example, some studies highlight an even higher risk associated with the male partner’s witnessing or experiencing violence as a child (compared to the woman) [17, 23, 24], but our dataset did not include variables that enabled looking at this. Further limitations to highlight include that the DHS data collection may not train researchers to collect sensitive data on partner violence in the same robust way as studies like the WHO study; or that couples who took part in the DHS surveys are different from those where one or both partner declined to take part; and finally, the analysis does not use “dyadic analytical techniques” to take into consideration that individuals are linked, but rather constructed simplistic couple-level indicators.

Over the last few years, Tanzania society has undergone structural economic reforms, with an increased number of women working and an increased number of men experiencing a crisis of masculinities. At the same time, the government of Tanzania has introduced measures to address violence against women in the country with developments of guidelines for clinical management and law enforcement [49]. Against this backdrop, prevalence of partner violence against women remains high. In other LMIC, targeted interventions to individuals, couples and the wider community have been found to reduce rates of partner violence [50]. Greater efforts, however, are required to address the structural constraints that perpetuate gender inequalities and that will continue to put women at risk. As efforts to address violence against women go on in Tanzania, this study provides an invaluable benchmark for the continuous monitoring of the effects these scale-up attempts have on reducing incidence of violence against women.