The baseline to assess impact of a mass education-entertainment programme offered an opportunity to identify risk factors for domestic physical violence.
In 2002, cross-sectional household surveys in a stratified urban/rural last-stage random sample of enumeration areas, based on latest national census in Botswana, Lesotho, Malawi, Mozambique, Namibia, Swaziland, Zambia and Zimbabwe. Working door to door, interviewers contacted all adults aged 16–60 years present on the day of the visit, without sub-sampling. 20,639 adults were interviewed. The questionnaire in 29 languages measured domestic physical violence by the question "In the last year, have you and your partner had violent arguments where your partner beat, kicked or slapped you?" There was no measure of severity or frequency of physical violence.
14% of men (weighted based on 1,294/8,113) and 18% of women (weighted based on 2,032/11,063) reported being a victim of partner physical violence in the last year. There was no convincing association with age, income, education, household size and remunerated occupation. Having multiple partners was strongly associated with partner physical violence. Other associations included the income gap within households, negative attitudes about sexuality (for example, men have the right to sex with their girlfriends if they buy them gifts) and negative attitudes about sexual violence (for example, forcing your partner to have sex is not rape). Particularly among men, experience of partner physical violence was associated with potentially dangerous attitudes to HIV infection.
Having multiple partners was the most consistent risk factor for domestic physical violence across all countries. This could be relevant to domestic violence prevention strategies.
Domestic violence – also known as intimate partner abuse, family violence, wife beating, battering, marital abuse, and partner abuse – is an international problem[1,2]. Domestic violence is not a single behaviour but a mix of assaulting and coercive physical, sexual, and psychological behaviours designed to manipulate and dominate the partner to achieve compliance and dependence. Women are more likely to experience physical injuries or psychological consequences[3,4].
Domestic violence is well documented in several African countries. In eastern Nigeria, a clinic-based survey of 300 women reported 40% had experienced violence in the previous year. In one district of Uganda, 30% of 5,109 women attending a clinic had received threats or physical abuse. The majority of respondents viewed wife beating as justifiable in some circumstances. In Durban, South Africa, more than one third of women from a low-income community had experienced domestic violence at some stage. A South African study reported domestic violence associated with violence in childhood, education and multiple partners[8,9]. In southern Africa domestic violence is particularly important because of the multiple links between violence and HIV infection. Links between domestic violence and HIV have been reported in Botswana, Ghana, Malawi, South Africa, Tanzania, Uganda[16,17], Democratic Republic of Congo and Zambia.
This is a baseline assessment of attitudes and practices, from which we intend to measure the impact of mass media campaigns, launched since the baseline by Soul City. The survey content was thus geared to measure the impact of education-entertainment messages, rather than as a specific research hypothesis. One section of the questionnaire dealt with domestic violence – attitudes and subjective norms, collective efficacy, discussion of the issue and experience of physical domestic violence in the last year – and the results are reported here as a cross-sectional survey.
In Botswana, Lesotho, Swaziland, Malawi, Mozambique, Namibia, Zambia and Zimbabwe we stratified the most recent available census into rural, urban (not within the capital region), and urban capital sites. In each country, we drew a last stage random selection of enumeration areas, with probability proportional to the national population (Table 1).
Table 1. Sample weights in each country
Training and fieldwork
After training, coordinators translated, back-translated and piloted the common instruments in 29 languages: Afrikaans, Bemba, Changana, Chichewa, Chindali, Chitimbuka, Chona/Shona, Chope, English, Herero, Kalanga, Kaonde, Kwangali, Lozi, Luvale, Mucua, Ndau, Ndebele, Nyanja, Oshiwambo, Portuguese, Ronga, Sena, Sesotho, Seswati, Setswana, Shangaan, Xitshwa and Xitsonga. Each field team of seven or eight interviewers visited approximately 10 communities, one per day. Interviewers tried to cover all households in each enumeration area, without sub-sampling. In each household, they interviewed all adults aged 16–60 years present at the time of the visit.
An accredited international ethical review board evaluated the proposal, noting concerns that disclosure might place the respondent at risk and that the questions about sexuality probed confidential issues. Interviewers informed each respondent of their right to refuse to participate, and of their right to refuse to answer any question. Before starting the questionnaire, the interviewers requested verbal consent to proceed. They did not record names or other identifying feature, and took precautions that the interview was out of hearing of others.
Of the 17,377 households in 213 randomly selected enumeration areas, 20,639 adults participated from 16,707 households (96% initial acceptance) where 85,114 people lived. 58% (11,872/20,639) were female; 63% (13,017) were rural residents, 22.1% (4,563) urban and 14.8% (3,059) lived in the capital/metro area (Table 2).
Table 2. Characteristics of the sample population
We defined domestic physical violence by responses to the question: "In the last year, have you and your partner had violent arguments where your partner beat, kicked or slapped you?" To facilitate disclosure, interviewers asked this with the respondent alone. If this was not possible, they noted presence of a listener. Interviewers read questions without additional explanations, and recorded answers verbatim. Wherever possible, female researchers interviewed women and male researchers interviewed men. With the exception of one question about pregnancy, interviewers administered the same instrument to men and women.
We limited domestic violence to reports of physical abuse, and we had no measure of severity of the violence. We included items on attitudes to and subjective norms of domestic violence, collective efficacy to reduce domestic violence (Can your community do anything about violence against women?) and discussion of domestic violence (In the last year, how often did you talk with anyone about domestic violence? To whom did you speak most often about domestic violence?). In designing the evaluation of the impact of mass media, we anticipated that some effect might be measured in these intermediate outcomes before changing the actual occurrence of domestic physical violence.
The relevance of partner physical violence to HIV/AIDS risk came from answers to the questions "Do you think you are at risk of getting HIV?" and "If you found you were HIV positive, how would you change your sex life", considering "always use a condom" and "abstain from sex" as positive values. Negative values included "no change", "spread it intentionally", "same partner" and "sleep with virgin to cure".
Data technicians manually digitised questionnaire data twice and eliminated keystroke errors by verifying discordant entries with the original questionnaires. We weighted final estimates in line with the national populations and the eight-country estimates weighted national indicators by the population of each country (Table 1). In a univariate analysis, we stratified each association between partner physical violence and potential risk factors by each of the others in turn (List 1, see Appendix), initially ignoring multiple influences[21,22]. We adjusted for the multiple comparisons by requiring 99% confidence.
For risk factors not explained by any stratifying variable and those with multiple influences, a step down logistic regression model tested the effect of country, age, sex, education, income, food security, household size, occupation, and the factors in List 1 (see Appendix). The several items on attitudes to sexuality and violence showed co-linearity, with no single variable attaining statistical significance in the preliminary logistic regression model. We included the variable from each group that showed the strongest association with the outcome in the model.
Some 16% of men (weighted value based on 1,294/8,113) and 18% of women (weighted value based on 2,032/11,063) reported partner physical violence in the last year; 6.8% (809/11,872) of female respondents and 6.0% (521/8,634) of males declined to answer this question. The lowest rates of partner physical violence came from Mozambique (9%) and Malawi (9%) and the highest from Zambia (32%) (Tables 3 and 4). The 7.1% with someone else present at the time of the interview were more likely to report a violent altercation (OR 1.18, 95%CI 1.02–1.35; 285/1,459 compared with 2,974/17,381 alone at the time).
Table 3. MALE Experience of physical violence in the last year (beat, kicked or slapped), discussion about gender violence and participation in community action about violence against women
Table 4. FEMALE Experience of physical violence in the last year (beat, kicked or slapped), discussion about gender violence and participation in community action about violence against women
Personal and household factors
The gender gap in reported domestic physical was negligible in Botswana, Lesotho, Namibia, Swaziland and Zimbabwe. Elsewhere, female respondents reported being the subjects of partner physical violence more frequently than did male respondents: in Malawi, the population weighted rates were 7% and 11% for males and females respectively (based on 72/1,109 and 176/1,586); in Mozambique, 7% and 11% respectively (based on 70/930 and 148/1,374) and in Zambia, 27% and 36% (based on 337/1,261 and 538/1,509).
Respondents aged 30–39 years reported violent altercations more commonly (20.4% unweighted, based on 908/4,478), with lower rates among older and younger respondents (16–19 years 11.4% 365/3,211; 20–29 years 19.3% 1,518/7,931; 40–49 years, 17.3% 376/2,196; 50–59 years 12.1% 135/1,118; and 60–66 years, 11.0% 26/235).
We found high reported rates of domestic physical violence in four of 29 interview languages. No less than 54% (82/152) of Lozi speakers (Zambia) reported partner physical violence in the last year. From the same country, 46% (99/197) of Tonga, 34% (339/995) of Bemba and 28% (206/744) Nyanja responders reported partner physical violence.
Some 31% (6,248/19,895) of the respondents had completed primary school; 3.5% (744/20,639) declined to answer this question. At first glance, the average person who had not completed primary school seemed more likely to report partner physical violence: OR 1.18 99%CI 1.05–1.32 (2,350/12,016 among those who had not completed primary education compared with 931/5,933 who had done so reported a violent altercation with a partner). This effect disappears entirely when stratifying by country; the levels of education combined with quite different rates of violent altercation seem to confound the measurement. In Zambia, the only country where education was associated with violent altercations, the average person who had not completed primary school was less likely to report a violent argument with a partner: argument with a partner: OR 0.82 95%CI 0.69–0.98 (600/1,979) among those who had not completed primary education compared with 266/768 who had done so reported a violent altercation with a partner).
We could find no obvious trend of violent altercation with increasing household size; missing data 6.6% (1,360/20,639). The average person living in a household with more than five members was less likely to report a violent altercation than one living in a household of 1–5 people (OR 0.88 99%CI 0.63–0.98; 1,295/7,887 in higher occupancy households compared with 2,049/11,383 in lower occupancy households reported a violent altercation).
Most respondents lived in rural areas (63.1% or 13,017/20,639); a further 22.1% were urban (4,563/20,639) and 14.8% lived in the capital city (3,059/20,639). There was very little difference in partner physical violence: rural 17.8% (2,164/12,160), urban 17.2% (736/4,287) and capital 15.8% (447/2,837).
Total household income
One in every ten (1,940/18,370) reported no income in the last month (11% or 2,269/22,630 declined to answer this question). Stratifying by country, there was no convincing association of domestic physical violence with income (OR adjusted 1.08, 99%CI 0.85–1.53; 346/1,757 of those with no income and 27,017/15,458 of those with an income). There was no detectable gender difference in this effect.
One in every ten did not register an occupation (3.7% 751/20,639 missing data). Housewives were most likely to report partner physical violence (25.6% based on 443/1,730), followed by those who described themselves as unemployed (19.5% based on 812/4,169). There was also no convincing association between remunerated occupation and partner physical violence (OR 0.95, 99%CI 0.8–1.1). We constructed a new variable to reflect the "income gap" between personal employment and total household income: overall, unemployed individuals in households with some income were more likely to report domestic physical violence (OR 1.43 99%CI 1.27–1.60; 901/4,111 with the income gap and 2,091/12,722 without it reported physical violence). On stratification by sex of respondent and country, however, it turned out that this association is ascribed mostly to women in Namibia and Zambia.
One in every three respondents reported having insufficient food in the last week (34.5% unweighted, 7,070/20,475); 0.8% (164/20,639) declined to respond. As with personal income, the average person reporting insufficient food was slightly more likely to report partner physical violence (OR 1.22 99%CI 1.10–1.35; 1,271/2,679 with insufficient food reported, compared with 2,052/12,536 with sufficient food). We could not explain this effect by urban/rural residence, country, attitudes to sexuality or sexual violence or any the personal factors we documented.
Attitudes about sexuality and sexual violence
Tables 5, 6, 7, 8, 9, 10 show the variation from country to country in attitudes about sexuality and sexual violence. Several of these beliefs were associated with partner physical violence (Tables 11 and 12): the belief that men have the right to have sex with girlfriends if they buy them presents (OR 1.42 99%CI 1.25–1.60), it is okay for an older man to have sex with teenagers (OR1.38 99%CI 1.20–1.59), women do not have the right to refuse sex with husbands and boyfriends (OR1.18 99%CI 1.05–1.30) and a person has to have sex to show love (OR 1.44 99%CI 1.38–1.59). Beliefs about gender violence were also associated with violent altercations: forcing one's partner to have sex is not rape (OR 1.23 99%CI 1.10–1.37) and women sometimes deserve to be beaten (OR1.56 99%CI 1.4–1.72). These associations were not explained by country, education, sex, remunerated occupation, income, multiple partners, household factors (like crowding, language, food security), or other attitudes and beliefs about sexuality or sexual violence.
Table 5. Male attitudes about sex
Table 6. Female attitudes about sex
Table 7. Male attitudes about violence
Table 8. Male attitudes about violence
Table 9. Male attitudes and subjective norms about sexual violence
Table 10. Female attitudes and subjective norms about sexual violence
Table 11. Male respondents: Associations with domestic physical violence (number of responses, Odds Ratio and 99%confidence interval)
Table 12. Female respondents: Associations with domestic physical violence (number of responses, Odds Ratio and 99%confidence interval)
One in every four respondents (4,468/17,948) who answered the question reported having two or more sexual partners in the last year; 15.9% (3,276/20,639) declined to answer. The proportion reporting multiple partners, out of those who had partners in the last year, varied somewhat by country: Botswana 32.1% (566/1,760), Lesotho 43.9% (780/1,760), Malawi 12.5% (274/2,195), Mozambique 31.6% (706/2,212), Namibia 21.0% (440/2,062), Swaziland 35.1% (517/1,465), Zambia 26.0% (600/2,316) and Zimbabwe 26.8% (585/2,175).
Using two or more partners in the last 12 months as a definition of multiple partners, there was a strong association with partner physical violence: female respondents OR 1.87 99%CI 1.46–2.41 (450/1564 of those with two or more partners compared with 1479/8332 among those with one on no partners) and male respondents OR 2.00 99%CI 1.47–2.66 (627/2755 among those with two or more partners compared with 592/4616 among those with one or no partners).
In all age groups in all countries, having multiple partners was a risk factor for violent altercations. A logistic model taking into account country, food security, sex of respondent, income, education and employment accentuated the risk of violent altercations for people with multiple partners (unadjusted OR 1.75, adjusted OR 2.03 99%CI 1.65–2.42, indicating underestimation of the unadjusted estimate).
Partner physical violence increased progressively with number of partners in the last 12 months: 234/1689 (13.9%) with no partners, 16.3% (1849/11324) with one partner, 22.7% (516/2269) with two partners, 25.4% (253/1034) with three partners, 29.2% (118/405) with four and 29.2% (185/633) with five or more partners reported domestic physical violence in the last year (χ2 199.8, 5 df).
Community dynamics and collective efficacy
A large proportion of the sample (65%, 12760/19626) said that domestic violence was considered a serious issue in their community (4.9% missing data, 1004/20639). Yet two thirds (9944/15880) of those who did not report physical violence and one half of those reporting partner physical violence in the last year (1654/3336) had never spoken about it. Those who spoke about it did so most frequently with friends (50.0% 3754/7504) and family (24.2%, 1819/7504). One in every ten said they had discussed with a neighbour (720/7504) and another one in ten with a partner or spouse (745/7504). There were no remarkable differences between male and female respondents, or between those who reported violent altercations and those who had not done so.
Over one half of the respondents said that their community could do something about violence against women (unweighted 56.2% based on 10466/18617, missing data 2017/20639 or 9.7%). Male respondents were more likely to express collective efficacy (OR 1.12 99%CI 1.02–1.23, 4529/7828 male and 5879/10685 female respondents felt their communities could do something about violence against women). Collective efficacy was highest in Botswana (75.6% 1715/2268) and Lesotho (62%, 1299/2095) and lowest in Zambia (44.5%, 1215/2732).
Relevance of partner physical violence to HIV risk
People who reported partner physical violence (male or female) were significantly more likely to believe they were at risk of getting HIV (OR 1.51, 99%CI 1.37–1.68; 1615/3075 who reported partner physical violence and 6261/14832 who did not report partner physical violence said they were at risk of HIV infection). This was not explained by country, sex of the respondent or any of the factors we could test in this study.
The average male respondent who reported partner physical violence was significantly more likely to anticipate a negative reaction to knowing he was HIV positive (no change, spread intentionally, sleep with virgin, etc) compared with one who had not suffered violence in the last year (OR 1.51, 99%CI 1.23–1.83, 286/1163 among those reporting and 1089/6142 not reporting partner physical violence). This association did not hold for female respondents, and among men it was not explained by country or any of the other variables we could test (List 1, see appendix).
High rates of domestic physical violence in all eight countries were conspicuously independent of education, household size, household income and remunerated employment. After taking into account age, sex, country and other factors, domestic physical violence was strongly associated with income gradients (being unemployed in the context of some household income) and home language in one country, and with multiple partners in the last year in all countries. Victims of partner physical violence were more likely to feel at risk of HIV infection and more likely to anticipate antisocial behaviour if they found they were HIV positive.
This is a cross-sectional household survey based on face-to-face interviews. This design limits conclusions about causality of, for example, multiple partners leading to physical violence or being the consequence of physical violence. It is likely that some respondents held back from expressing their true belief or experience. Even with the best field practices – including independent translation and back-translation of questionnaires, standardised training of local interviewers, in-country piloting and consultation with local community representatives, double-data entry and verification – measurement error is possible. The sample makes the results relevant to the eight countries, but not necessarily to other countries.
A major limitation is that we only considered domestic physical violence. This almost certainly underestimates the level of domestic violence. Other forms (verbal, sexual, economic and psychological) were beyond the scope of the study. In all countries we asked the same questions of men and women. We were able to examine several intermediate outcomes related to domestic violence – including attitudes, subjective norms, collective efficacy and discussion/socialisation – but most of these could be addressed only superficially through one or two items in the questionnaire.
We had no measure of severity or frequency of physical domestic violence, making it difficult to interpret the proportion of men and women who reported partner violence in the last year. Large studies in the UK and USA have reported similar proportions of partner violence for males and females, but found male on female violence to be more severe than female on male violence[23,24]. It is quite possible that the same is true for southern Africa. The men we interviewed were at home during working hours and, in this respect at least, they may not be typical of all men in the eight countries. We also did not ask who initiated the altercation, so it is also possible these reports reflect women defending themselves from male-initiated violence. Even so, the finding is compatible with a degree of female agency in domestic physical violence and supports our conclusions from South Africa that initiatives against sexual violence should look beyond gender stereotypes of victims and villains.
There was no recognisable pattern of poverty and domestic violence between countries (Mozambique, the poorest country, reported the lowest rates while Zambia reported the highest). We also did not find significant associations between victims and their individual education or employment, and we could only address the income gradient between partners through a proxy variable. It is possible that in-household inequality in education and income could be more relevant to domestic violence than we were able to measure in this study. There was no interpretable association between the Gini coefficient (measuring inequality in the country) and male or female reports of violence (Tables 3 and 4). The Gini coefficient used for Botswana and Lesotho was 0.63, Malawi 0.50, Mozambique 0.40, Namibia 0.74, Swaziland 0.61, Zambia 0.42 and Zimbabwe 0.61.
The occurrence of domestic physical violence in some parts of Zambia raises the question of something being done differently there, despite efforts to reproduce exactly the same survey in all countries. Whatever the reason for the higher rates of domestic physical violence detected in Zambia, it seems unlikely the same error lies behind the inability to demonstrate an association between violent altercations and education, overcrowding, income and age – consistent across all the countries.
If there is good news from this study, it is that multiple partners, attitudes and subjective norms are more in the control of most individuals than are poverty, overcrowding and education – without detracting from the need for massive investment in these sectors.
An unanswered question is how to modify attitudes or multiple partners. There is also no guarantee that changing attitudes will, on its own, impact on behaviour. The study confirms the importance of moving beyond gender stereotypes of victims and villains. Men also report suffering partner physical violence, although our inability to measure severity could mask an important gender difference. The solutions to domestic violence lie with both men and women, and both have agency in this regard. There was also a prominent sense of collective efficacy, the majority expressing they could do something about domestic violence.
Although many thought their community could deal with violence against women, few victims and still fewer of the non-victims said they had discussed violence against women with anyone. Stimulating discussions about violence against women offers one direction for initiatives against partner physical violence. Wider discussion could influence social norms, in addition to targeting individual attitudes and supportive public policy.
List 1. Variables tested sequentially, from which independent associations were included in logistic regression model
Individual and household characteristics
How many people live in the household
Age and sex of each one
Language spoken at home most of the time
Last grade of education respondent completed
Main occupation of respondent
Total household income per month
Did household have enough food in the last week
Was the respondent alone or was someone listening
Do you think you are at risk of getting HIV
If you found you were HIV positive, how would you change your sex life
If a woman gets raped its her own fault.
Forcing sex with someone you know is not rape.
Forcing your partner to have sex is rape.
Subjective norms about sexual violence
Do most people in your community feel forcing your partner to have sex is rape?
Do most people in your community feel women have a right to refuse sex with their partners?
Is violence against women considered a serious problem in this community?
Collective efficacy about sexual violence
Can your community do anything about violence against women?
Attitudes to domestic violence
Women have the right to refuse to have sex with partner
Violence between a man and a woman is a private matter Women sometimes deserve to be beaten.
Subjective norms about domestic violence
Do most people in your community feel women sometimes deserve to be beaten?
Discussion about domestic violence
In the last year, how often did you talk with anyone about domestic violence? [never, seldom or often]
To whom did you speak most often?
Practices relating to domestic violence
What community activity about violence against women have you participated in?
In the last year, have you and your partner had violent arguments where someone was physically hurt?
Men have the right to have sex with their girlfriends if they buy them gifts.
Its okay for an older man to have sex with teenagers
A person has to have sex with their boyfriend or girlfriend to show that they love them.
Do most of your friends feel men have the right to sex with their girlfriends if they buy them gifts?
All authors declare that there is no competing interest. Esca Scheepers and Sue Goldstein were employed by Soul City, which subcontracted the national education-entertainment programmes in the eight countries.
NA was involved in study and questionnaire design, statistical analysis, drafting manuscript. AHF was involved in statistical analysis, interpretation and drafting manuscript. SM was involved in study design, acquisition of data, drafting manuscript. ES and SG were involved in study and questionnaire design, analysis and interpretation, drafting manuscript, administration and technical support. NA, AHF, ES and SG had full access to all data and take responsibility for the integrity of the data and accuracy of data analysis. All authors read and approved the final manuscript.
The eight national surveys were funded by a grant from the European Union, made available through Soul City. Lorenzo Monasta trained fieldworkers and coordinated fieldwork in Malawi, as did Charlie Whitaker in Mozambique, Sharmila Mhatre in Zambia, Manuel Pascual Salcedo in Swaziland and Lesotho, and Serge Merhi in Namibia and Zimbabwe. Marietjie Myburg supported fieldwork in Swaziland and Lesotho. Candyce Hamel provided analysis support.
World Health Organization: [http:/ / www.who.int/ violence_injury_prevention/ violence/ world_report/ en/ full_en.pdf] webcite
World report on violence and health. Geneva : WHO; 2002.
(Accessed 26 July 2005).
Retrieved 8 March 2007
S Afr Med J 2003, 93:455-457. PubMed Abstract
Social Science and Medicine 2002, 54:325-337. Publisher Full Text
Medical Anthropology 1997, 17:363-380. PubMed Abstract
Health, Risk & Society 2000, 2(2):189-203. Publisher Full Text
Acquired Immune Defic Syndr 2003, 34(2):203-13. Publisher Full Text
Social Science and Medicine 1993, 37(8):989-994. Publisher Full Text
Social Science and Medicine 2001, 52:509-518. Publisher Full Text
Singhal A, Usdin S, Scheepers E, Goldstein S, Japhet G: Harnessing the Entertainment-Education Strategy in Development Communication by Integrating Program Design, Social Mobilization and Advocacy. In Communication and Development in Africa. Edited by Charles Okigbo, Festus Eribo. Boston, Rowman & Littlefield; 2004.
Walker A, Hershow C, Nicholas S: Crime in England and Wales 2005/6. [http://www.homeoffice.gov.uk/rds/pdfs06/hosb1206.pdf] webcite
Home Office Statistical Bulletin, London; 2006.
U.S. Department of Justice Office of Justice Programs, National Institute of Justice, Washington DC; 2000.
Retrieved on 2 June 2007.
The pre-publication history for this paper can be accessed here: