Intimate Partner Violence Attitudes and Experience among Women and Men in Uganda

Abstract

This study examines intimate partner violence (IPV) attitudes and experience among women and men in Uganda to inform IPV prevention programs in the region. Nationally representative population-based data from women ages 15–49 and from men ages 15–54 were collected between May and October, 2006 as part of the Uganda Demographic and Health Survey. The survey included questions on women’s and men’s attitudes toward wife beating and information on IPV victimization (women) and perpetration (men). More than half of men and nearly three-quarters of women have attitudes supportive of wife beating in Uganda. More than half of married women report IPV victimization and 40% of married men report perpetration. Women and men who reported witnessing their fathers beating their mothers were more likely to report IPV victimization (perpetration for men). Witnessing violence was also associated with positive attitudes towards wife beating among men. IPV prevention programs need to address the important role of having witnessed wife beating between the mother and the father on men’s subsequent attitudes and behaviors. Women who witnessed wife beating are also the most likely to have supportive attitudes and IPV experience, possibly indicating that their relationship expectations are different than women who did not witness violence. Community-based prevention programs targeting men and women are needed in Uganda and elsewhere in sub-Saharan Africa where gender norms that justify intimate partner violence prevail.

Introduction

Prevention of intimate partner violence (IPV) requires an understanding of the factors that make IPV acceptable and factors associated with IPV victimization and perpetration. To date, there are a small number of studies from sub-Saharan Africa on IPV attitudes and correlates of positive attitudes towards wife beating (Hinden, 2003; Ilika, 2005; Oyediran & Isiugo-Abanihe, 2005; Rani, Bonu, & Diop-Sidibe, 2004). One study among married or cohabiting women from Zimbabwe demonstrates that urban women, women from higher socioeconomic status households, and women who are older are less likely to believe that wife beating is justified (Hinden, 2003). This same study finds that women who have partners who reside with them and women who have partners who make more of the household decisions are more likely to have supportive wife beating attitudes than all others (Hinden, 2003). Similar demographic findings were found in a study from Nigeria (Oyediran & Isiugo-Abanihe, 2005) and in a study that included women from seven sub-Saharan African countries (Rani et al., 2004). The multi-country study also included men and examined factors associated with men’s attitudes toward IPV in six of the sub-Saharan African countries with available data (Rani et al., 2004). Across the multiple countries, factors generally associated with less positive attitudes toward wife beating among men included having more household assets, higher education, and older age. The authors also demonstrated that women were consistently more likely to justify wife beating than were men and the authors concluded that cultural norms create an environment where traditional gender roles make wife beating seem normative and prevention requires changing these traditional gender norms (Rani et al., 2004). Higher wife beating attitudes among women were also found in a descriptive study from rural Uganda (Koenig et al., 2003).

Studies of IPV experience among women also find a set of somewhat consistent factors associated with IPV. These include having been in union two or more times, having had three or more children ever born, working for money, husband drunkenness, and lower education level (Karamagi, Tumwine, Tylleskar & Heggenhougen, 2006; Kishor & Johnson, 2004; Koenig et al., 2003). In a multi-country study of women, in five of the nine countries included, women who were married for a longer period were significantly more likely to have ever experienced IPV than women who were only recently married or in union (Kishor and Johnson, 2004); this may represent a longer exposure time to IPV. Only a few studies from developing countries examine perpetration of IPV as reported by men (Luke, Schuler, Mai, Thien, & Minh, 2007). One study among a random sample of working men from three Cape Town municipalities in South Africa indicated that men who used violence to solve problems in other settings, had more than one current partner, abused alcohol, believed that hitting a woman was justified, and witnessed their mother being abused were more likely to perpetrate sexual violence than all others (Abrahams, Jewkes, Hoffman, & Laubsher, 2004). The authors conclude that prevention of IPV will require addressing all violence norms and behaviors among men, not just IPV.

A common factor found to be associated with IPV victimization (and perpetration) in studies of women (and men) is having ever witnessed your mother being beaten by your father (witnessing IPV) (Abrahams et al., 2004; Kishor, & Johnson, 2004; Koenig, Stephenson, Ahmed, Jejeebhoy, & Campbell, 2006). In these studies, witnessing IPV may reflect relationship norms (and expectations) that increase a woman’s risk of experiencing IPV (or alternatively men’s risk of perpetrating IPV). Thus, witnessing IPV may be associated with supportive attitudes toward IPV as well as IPV experience. A multi-country study using Demographic and Health Survey data recently demonstrated that women who approve of wife beating under varying scenarios are at increased risk of having experienced IPV than women who do not approve of wife beating (Kishor and Johnson, 2004). A challenge for studying how witnessing wife beating and wife beating attitudes are jointly associated with IPV is the potential for multiple directions of causality between attitudes toward wife beating and IPV. In particular, while it is programmatically convenient to say that women with supportive attitudes toward wife beating are at greater risk of IPV, and thus programs should change these attitudes, the converse is also possible. In particular, women who experience IPV may rationalize their experiences and report attitudes favorable of IPV, suggesting that IPV experience leads to supportive attitudes rather than the opposite (Koenig et al., 2003). In this last scenario, witnessing IPV may be unrelated to IPV attitudes and experience.

This study contributes to earlier work by examining IPV attitudes, IPV experience, and the role of having witnessed IPV on attitudes and experiences, using recently collected, nationally representative population-based data from Uganda that included women and men. The study teases out the differential effects of witnessing IPV on women’s and men’s attitudes and on their experience of IPV (as victims or perpetrators). The results are used to recommend future violence prevention strategies for Uganda and elsewhere in sub-Saharan Africa where attitudes supportive of IPV prevail.

Method

Context

Uganda is a land-locked country in East Africa with a 2008 estimated mid-year population of about 29 million inhabitants (PRB, 2008). More than 80% of Uganda’s population lives in rural areas and about half of the urban population is based in Kampala, the capital city (UBOS, & Macro International Inc., 2007). In 2007, the total fertility rate (TFR) in Uganda was estimated to be 6.7 births per woman (UBOS, & Macro International Inc., 2007). The TFR varied widely between urban (TFR=4.4) and rural areas (TFR=7.1). The prevalence of HIV in Uganda was estimated in 2004 to be 6% among adults aged 15–49 with a higher prevalence among women (8%) than among men (5%) (MOH, & ORC Macro, 2006). The prevalence of IPV from previous studies in Uganda ranges from 30% in rural South Central Uganda (Koenig et al., 2003) to 58% of rural women and 45% of urban women in Eastern Uganda (Karamagi et al., 2006).

Over the last two decades, parts of Northern Uganda have been characterized by extreme violence led by the Lords Resistance Army (LRA). The LRA is notable for extreme mutilations and abductions, which has led to widespread fear in the areas affected (IRIN, 2007). The war and brutal attacks led by the LRA, often targeting women and girls, has led to large numbers of internally displaced populations in Northern Uganda. In the 2007 Uganda Demographic and Health Survey, It was estimated that about 6% of the population resided in internally displaced camps in the North (UBOS, & Macro International Inc., 2007). These persons may have IPV attitudes and experiences that differ greatly from their counterparts from less affected areas of Uganda.

Data and Variables

This study uses the 2006 Uganda Demographic and Health Survey (UDHS) data, collected between May and October 2006 (UBOS, & Macro International Inc., 2007). The survey includes a nationally representative sample of 8,531 women ages 15–49 years and 2,503 men ages 15–54 years. The study employed multi-stage sampling to first select a random sample of enumeration areas and then systematically select a random sample of households in each enumeration area. All eligible women were approached and asked to participate in the interview. In one third of the households, men were asked to be interviewed. Of every three households, one was pre-selected and one woman (or one man) was asked questions on domestic violence. A total of 2,087 women and 1,844 men responded to the domestic violence questions. Only women and men who were ever married or in union were asked the IPV questions. Therefore our sample is reduced to include 1,749 ever-married or in union women (weighted sample size is 1,598) and 1,318 ever-married or in union men (weighted sample size is 1,158).

The dependent variables of interest are (a) whether the respondent has any attitudes supportive of wife beating and (b) if the respondent has ever been involved in incidents of IPV. All respondents were asked if it was acceptable for a man to beat his wife in five scenarios (wife burns food, refuses sex, neglects children, goes out without informing her husband and argues with husband). Respondents who said “yes” to any of the scenarios are coded one to reflect attitudes supportive of wife beating; respondents who reported “no” to all scenarios are coded zero. For IPV victimization, women were asked a series of questions on physical and sexual violence based on a modified version of the revised Conflict Tactics Scale (Ellsberg, & Heise, 2005; Straus, Hamby, Boney-McCoy, & Sugarman, 1996). Women who were ever married or in union and reported that a husband or partner ever pushed, shook or threw something at them, slapped them, twisted their arms, pulled their hair or punched them; kicked, dragged or beat them; tried to choke or burn them; threatened or attacked with a knife, gun or another weapon; or ever forced them to have sex or perform sexual acts are coded as one. All other women are coded zero. Men who were ever married or in union were asked a single question: “Have you ever hit, slapped, kicked or done anything else to physically hurt your (last) wife/partner at times when she was not already beating or physically hurting you?” Men who responded affirmatively are coded one on the IPV perpetration variable and zero otherwise. Finally, the attitude and IPV variables were used to create a joint outcome with four categories: 1) no attitudes supportive of wife beating and no IPV victimization (perpetration); 2) attitudes supportive of wife beating and no IPV victimization (perpetration); 3) no attitudes supportive of wife beating and ever IPV victimization (perpetration); and 4) both attitudes supportive of wife beating and IPV victimization (perpetration).

Background characteristics were selected for inclusion in the analysis based on their significance in previous studies of violence attitudes and behaviors (Hinden, 2003; Karamagi et al., 2006; Rani et al., 2004). The background variables are listed in Table 1 with their respective coding schemes. To examine the role of witnessing IPV, we included a dichotomous measure of whether the respondent reported ever witnessing his or her father beating his or her mother (yes or no).

Table 1.

Demographic Characteristics and IPV Witnessing among Married or In Union Women and Men, Uganda DHS

	Women, N=1,749	Men, N=1,318
	%	%
Region
Rural	85.8	86.5
Other urban	7.5	6.7
Kampala	6.8	6.9
Age group
<25	26.7	12.7
25–39	53.2	56.7
40+	20.1	30.6
Number of children
None-2	34.5	32.4
3–4	27.3	24.0
5+	38.2	45.6
Education level
None	23.3	7.7
Primary	62.8	66.1
Secondary +	13.9	26.2
Religion
Catholic	43.3	42.6
Protestant	32.7	39.8
Muslim	10.2	9.9
None/other	13.8	7.7
Wealth quintile
Lowest	18.8	16.8
Low	19.8	23.4
Medium	20.9	19.3
High	20.6	20.4
Highest	19.9	20.1
Father beat mother
No	52.2	40.9
Yes	47.8	59.1

Analyses

Logistic regression techniques were employed to examine factors associated with attitudes supportive of wife beating and factors associated with IPV victimization or perpetration. In the last analysis, multinomial logistic regression methods were employed to determine which factors are associated with the joint four-category outcome variable. All analyses were adjusted for the multistage sampling design and utilized the domestic violence sampling weights. All analyses were performed using Stata version 9.2. This study using anonymous public-release data was reviewed by the University of North Carolina at Chapel Hill Institutional Review Board and considered exempt from human subjects research.

Results

Characteristics of Sample

Table 1 presents the characteristics of the sample of ever-married or in union women and men who were asked the domestic violence questions in the UDHS. About two-thirds of women in the sample have primary education and a quarter has no education. About two fifths of women are Catholic and a third is Protestant; ten percent of women are Muslim. The male sample is similar to the female sample in terms of residence, religion, and wealth group. Generally speaking, the male sample is older and more educated than the female sample. Finally, Table 1 indicates that nearly half of women report that their father beat their mother (48%). The percentage of men giving this response is higher at 59%.

In Table 2, the measures that went into the creation of the attitude dependent variable are presented as are the violence experience variable and the joint attitude/experience outcome. Among women, about three-quarters approve of wife beating under any of the five scenarios. The scenarios that are most likely to elicit an affirmative response are “neglects children” (57% agree) and “goes out without informing husband” (55% agree). A significantly smaller proportion of men report that wife beating is acceptable under any of the scenarios (57%). Among men, the scenarios that are most likely to elicit a positive response are “neglects children” (43% agree), closely followed by “goes out without informing husband” (39% agree) and “argues with husband” (35% agree). More than half of women (57%) report that they have ever experienced sexual or physical violence. A significantly smaller percentage of men report ever perpetrating IPV (40%). The distinction between reported IPV victimization and perpetration may reflect distinctions in the way the questions were asked (women were asked about specific acts of violence and men were just asked if they ever perpetrated violence) and possible differential-reporting of men who, for example, might not perceive pushing or shoving (or another form of violence) as a violent act.

Table 2.

Attitudes Toward Wife Beating and IPV Experience among Married or In Union Women and Men, Uganda DHS

	Women, N=1,749	Men, N=1,318
	%	%
Agrees with wife beating by scenario (% agree)
Wife goes out without informing husbanda	54.7	38.6
Neglects childrena	57.4	42.6
Argues with husbanda	42.5	34.5
Refuses to have sex with husbanda	34.4	19.4
Burns the fooda	24.4	12.9
Agree with wife beatinga
Never	26.6	43.1
Any of five scenarios	73.4	56.9
Any violence experiencea (victimization/perpetration)
Never	40.8	59.9
Ever	56.6	40.1
Attitude/Behaviora
Disagree, no experience	14.1	31.3
Agree, no experience	29.3	28.6
Disagree, experience	12.6	11.8
Agree and experience	44.0	28.4

Note: Unweighted sample sizes shown, weighted sample sizes are: 1,598 and 1,158, respectively.

^aDifference between males and females are significant at p=.001.

Factors Associated with Positive Attitudes toward Wife Beating

In Table 3, multivariate logistic regression results are presented to examine which women (and men) have positive attitudes toward wife beating. Among women married or in union, those who are from rural areas or from an urban area other than Kampala are significantly more likely to have positive attitudes toward wife beating than women from Kampala. Women with less than five children and women with no religion or another religion are significantly less likely to have positive attitudes than women with five or more children and women who are Catholic.

Table 3.

Multivariate Logistic Regression Odds Ratios (and Confidence Intervals) of Factors Associated with Reporting that Wife Beating is Acceptable under at Least One Circumstance, Uganda DHS, 2006

	Female Married/In Union	Male Married/In Union
	N=1,746	N=1,315
Region:
Rural	2.84 (1.61–5.02)***	2.00 (1.06–3.78)*
Other urban	3.94 (2.06–7.56)***	1.54 (0.69–3.44)
Kampala (ref)	1.0	1.0
Age group:
<25	1.40 (0.94–2.08)	2.46 (1.50–4.06)***
25–39 (ref)	1.0	1.0
40+	0.88 (0.59–1.31)	1.10 (0.78–1.54)
Number of children
None-2	0.52 (0.34–0.78)**	1.13 (0.80–1.60)
3–4	0.65 (0.46–0.92)*	0.85 (0.60–1.22)
5+	1.0	1.0
Education level:
None (ref)	1.0	1.0
Primary	1.17 (0.84–1.64)	1.51 (0.89–2.56)
Secondary +	0.75 (0.48–1.17)	0.85 (0.46–1.57)
Religion:
Catholic (ref)	1.0	1.0
Protestant	0.98 (0.73–1.36)	1.03 (0.76–1.39)
Muslim	1.19 (0.72–1.97)	0.94 (0.55–1.61)
None/other	0.62 (0.40–0.97)*	0.69 (0.38–1.22)
Wealth quintile:
Lowest (ref)	1.0	1.0
Low	1.10 (0.72–1.68)	1.21 (0.78–1.86)
Medium	1.19 (0.74–1.92)	0.79 (0.50–1.25)
High	0.67 (0.42–1.06)	0.79 (0.49–1.28)
Highest	0.60 (0.36–1.00)	0.86 (0.51–1.44)
Father beat mother:
No (ref)	1.0	1.0
Yes	1.29 (0.98–1.70)	1.82 (1.39–2.38)***

Note: Unweighted N’s shown, weighted values are: 8,527, 1,596, 2,503 and 1,156, respectively.

^*p ≤ 0.05;

^**p ≤ 0.01;

^***p ≤ 0.001.

Among men, those who are from rural areas are two times more likely to report positive wife beating attitudes than are men from Kampala. Men who are under 25 are also two times more likely to report positive attitudes than men ages 25–39. Finally, men who witnessed violence between their father and mother are nearly two times more likely to have positive attitudes toward wife beating than men who did not witness such violence.

Factors Associated with IPV Experience and Perpetration

Table 4 provides the multivariate logistic regression analysis results of IPV victimization (women) and IPV perpetration (men). Women ages 40+ years are significantly less likely to have been victims of IPV than women ages 25–39. Women who ever witnessed their fathers beating their mothers were significantly more likely to have ever been victims of IPV than women who did not witness violence, controlling for all other factors.

Table 4.

Multivariate Logistic Regression Odds Ratios (and Confidence Intervals) of Factors Associated With Reporting Ever Victimization (Women) or Perpetration (Men) of Physical or Sexual IPV, Uganda DHS, 2006

	Female – Married/In Union	Male – Married/In Union
	N=1,746	N=1,314
Region:
Rural	1.58 (0.92–2.72)	0.73 (0.35–1.52)
Other urban	1.08 (0.53–2.20)	0.81 (0.34–1.96)
Kampala (ref)	1.0	1.0
Age group
<25	0.74 (0.53–1.04)	0.87 (0.51–1.47)
25–39 (ref)	1.0	1.0
40+	0.58 (0.41–0.82)**	1.08 (0.78–1.49)
Number of children
None-2	0.89 (0.62–1.28)	0.46 (0.32–0.65)***
3–4	1.06 (0.78–1.44)	0.81 (0.58–1.13)
5+	1.0	1.0
Education level
None (ref)	1.0	1.0
Primary	1.18 (0.86–1.62)	0.95 (0.55–1.65)
Secondary +	0.78 (0.50–1.23)	0.84 (0.47–1.51)
Religion
Catholic (ref)	1.0	1.0
Protestant	1.02 (0.77–1.36)	1.00 (0.74–1.37)
Muslim	0.81 (0.51–1.27)	0.90 (0.55–1.49)
None/other	0.91 (0.61–1.35)	0.77 (0.46–1.29)
Wealth quintile
Lowest (ref)	1.0	1.0
Low	1.23 (0.87–1.75)	1.20 (0.78–1.86)
Medium	0.99 (0.64–1.53)	0.76 (0.47–1.22)
High	1.06 (0.68–1.64)	0.79 (0.49–1.28)
Highest	0.78 (0.49–1.24)	0.63 (0.36–1.12)
Father beat mother
No (ref)	1.0	1.0
Yes	1.62 (1.28–2.04)***	1.84 (1.39–2.44)***

Unweighted N’s shown, weighted values are: 1,596 and 1,155, respectively.

^**p ≤ 0.01;

^***p ≤ 0.001.

In the second column in Table 4, the factors associated with IPV perpetration among men are presented. Men with two or fewer children were significantly less likely to have perpetrated violence than men with five or more children. As with women, men who witnessed their fathers beating their mothers were significantly more likely to have perpetrated IPVthan men who did not.

Simultaneous Examination of IPV Attitudes and Experiences

Table 5 presents the multinomial logistic regression coefficients from the analysis that examines women’s attitudes toward wife beating and IPV victimization simultaneously. The dependent variable has four categories: not supportive of wife beating/no IPV victimization; supportive of wife beating/no IPV victimization; not supportive of wife beating/IPV victimization; and supportive of wife beating/IPV victimization (referred to as neither, attitude-only, victimization only, and both, respectively). In the comparison of the attitude-only group to the neither group, residence, number of children, and religion show similar results to the previous analysis. In this comparison, women with secondary education are significantly less likely to be in the attitude-only group (more likely to be in the neither group). Notably, witnessing violence is not significantly associated with having the attitude only in this model.

Table 5.

Multinomial Logistic Regression Odds Ratios (Standard Errors) for Factors Associated with Violence Attitudes and Violence Experience among Women Married or In Union, Uganda DHS, 2006

	Violence attitude/no experience Vs. No attitude, no experience	No attitude/IPV experience Vs. No attitude, no experience	Both attitude and IPV experience Vs. No attitude, no experience
Region
Rural	3.32 (1.31–8.33)*	1.84 (0.75–4.51)	4.03 (2.09–7.79)***
Other urban	5.45 (2.16–13.76)***	1.59 (0.47–5.39)	4.12 (1.79–9.51)***
Kampala (ref)	---	---	---
Age group
<25	1.36 (0.87–2.12)	0.67 (0.37–1.23)	1.02 (0.66–1.59)
25–39 (ref)	---	---	---
40+	0.77 (0.45–1.32)	0.43 (0.23–0.83)*	0.49 (0.29–0.84)**
Number of children
None-2	0.57 (0.34–0.94)*	1.04 (0.52–2.07)	0.50 (0.31–0.82)**
3–4	0.78 (0.47–1.29)	1.42 (0.78–2.56)	0.78 (0.48–1.27)
5+	---	---	---
Education level: None (ref)
Primary	0.78 (0.49–1.23)	0.67 (0.37–1.22)	1.09 (0.70–1.70)
Secondary +	0.52 (0.28–0.98)*	0.47 (0.22–1.00)	0.50 (0.27–0.91)*
Religion
Catholic (ref)	---	---	---
Protestant	0.85 (0.55–1.33)	0.83 (0.48–1.42)	0.94 (0.60–1.46)
Muslim	1.16 (0.60–2.27)	0.76 (0.36–1.60)	0.95 (0.49–1.84)
None/other	0.55 (0.31–0.96)*	0.79 (0.40–1.57)	0.56 (0.32–0.98)*
Wealth quintile
Lowest (ref)	---	---	---
Low	1.11 (0.61–2.00)	1.28 (0.63–2.60)	1.33 (0.77–0.91)
Medium	1.53 (0.78–3.00)	1.42 (0.66–3.03)	1.36 (0.77–2.31)
High	0.76 (0.40–1.44)	1.32 (0.62–2.82)	0.78 (0.43–1.43)
Highest	0.80 (0.42–1.55)	1.19 (0.69–1.60)	0.56 (0.29–1.08)
Father beat mother
No (ref)	---	---	---
Yes	0.92 (0.63–1.35)	1.05 (0.69–1.60)	1.70 (1.19–2.44)**

Unweighted N is 1,746; weighted N is 1,596.

^*p ≤ 0.05;

^**p ≤ 0.01;

^***p ≤ 0.001.

Column 2 presents factors associated with IPV victimization only vs. neither. Controlling for all factors, women over age 40 are significantly less likely to be in the IPV-only group (more likely to be in the neither group). It is worth nothing that in the comparison between women who experienced IPV only and the neither group, the witnessing violence variable is not significant. Finally, the third column shows the factors associated with having both supportive violence attitudes and IPV victimization as compared to neither the attitudes or victimization. Women from rural and other urban areas are significantly more likely to have both the attitude and IPV victimization than women from Kampala. Women who are over 40, women with two or fewer children, women with secondary education, and women who report no religion are significantly less likely to have attitudes supportive of wife beating and IPV victimization experience and are more likely to be in the neither group. Furthermore, women who witnessed violence are significantly more likely to have supportive attitudes and have been victims of IPV than to be in the neither group as compared to women who did not witness violence.

The same multinomial logistic regression analysis was undertaken with the male data (see Table 6). Across all three comparisons, men who witnessed violence were significantly more likely to have attitudes supportive of IPV, to have perpetrated IPV, and to have both the positive attitudes and to have been a perpetrator as compared to men who did not witness violence. Other factors associated with one or another comparison were similar to the logistic regression findings presented in Tables 3 and 4.

Table 6.

Multinomial Logistic Regression Odds Ratios (Standard Errors) for Factors Associated with Violence Norms and Violence Perpetration among Men Married or In Union, Uganda, DHS, 2006

	Violence attitude/no experience Vs. No attitude, no experience	No attitude/IPV experience Vs. No attitude, no experience	Both attitude and IPV experience Vs. No attitude, no experience
Region
Rural	2.52 (0.97–6.57)	0.78 (0.28–2.14)	1.40 (0.57–3.45)
Other urban	1.38 (0.43–4.04)	0.59 (0.15–2.38)	1.28 (0.48–3.46)
Kampala (ref)	---	---	---
Age group
<25	2.37 (1.34–4.17)**	0.47 (0.16–1.37)	1.88 (0.99–3.60)
25–39 (ref)	---	---	---
40+	1.06 (0.70–1.60)	1.03 (0.56–1.90)	1.15 (0.76–1.73)
Number of children
None-2	1.41 (0.91–2.17)	0.49 (0.27–0.88)*	0.58 (0.37–0.90)*
3–4	0.84 (0.54–1.32)	0.77 (0.44–1.35)	0.74 (0.48–1.14)
5+	---	---	---
Education level: None (ref)
Primary	1.96 (0.97–3.94)	1.22 (0.43–3.49)	1.33 (0.69–2.55)
Secondary +	1.21 (0.53–2.79)	1.43 (0.49–4.18)	0.73 (0.36–1.51)
Religion
Catholic (ref)	---	---	---
Protestant	1.16 (0.80–1.69)	1.25 (0.77–2.02)	1.03 (0.69–1.53)
Muslim	1.14 (0.61–2.16)	1.25 (0.51–3.05)	0.88 (0.46–1.70)
None/other	0.73 (0.37–1.44)	0.94 (0.46–1.93)	0.58 (0.28–1.21)
Wealth quintile
Lowest (ref)	---	---	---
Low	1.39 (0.82–2.35)	1.55 (0.75–3.23)	1.40 (0.84–2.35)
Medium	0.85 (0.48–1.52)	0.82 (0.38–1.76)	0.68 (0.37–1.25)
High	0.93 (0.51–1.67)	1.03 (0.48–2.19)	0.69 (0.38–1.27)
Highest	1.17 (0.60–2.28)	0.93 (0.42–2.04)	0.61 (0.30–1.21)
Father beat mother
No (ref)	---	---	---
Yes	2.13 (1.50–3.02)***	2.68 (1.70–4.21)***	2.58 (1.81–3.68)***

Unweighted N is 1,314; weighted N is 1,155.

^*p ≤ 0.05;

^**p ≤ 0.01;

^***p ≤ 0.001.

Discussion

Do violence attitudes and behaviors vary among women and men in Uganda and does this have implications for prevention programming? This study sought to answer this question, using recently collected nationally representative data from women and men in Uganda. We demonstrated that a high percentage of women and men have attitudes supportive of wife beating and women are significantly more likely to report supportive attitudes than men. The study also demonstrates that women with attitudes supportive of wife beating and who have been victims of IPV are different than those women who only have supportive attitudes or only experienced IPV. In particular, women who reported that they witnessed their father beating their mother were more likely to report supportive wife beating attitudes and IPV victimization; however, witnessing violence was not associated with having the attitude only or IPV victimization only.

This study confirms findings from an earlier study of wife beating attitudes among men in six sub-Saharan African countries (Rani et al., 2004)by demonstrating that younger men are more likely to have attitudes supportive of wife beating. The findings from this study from Uganda, however, permit greater depth in analysis and demonstrate that younger men were not more likely to have acted out their attitudes and have ever been perpetrators. The main factor associated with supportive attitudes and/or IPV perpetration among men was witnessing the father beat the mother. This indicates that among men, the intergenerational transmission of IPV is an important problem in Uganda that needs to be considered in any IPV prevention programming.

This study sought to confirm the findings of earlier research and to further research in the field of IPV in several ways. First, this study is unique because it uses data from a nationally representative sample to examine wife beating attitudes and IPV perpetration from a male perspective. Most studies that include measurement of male perpetration are from developed countries and concentrate on specific high-risk populations such as men convicted of IPV or men in the military (Bergen, & Bukovec, 2006; Tilley, & Brackley, 2005). One study among men in South Africa focused specifically on working men from three municipalities, finding as we did, that men who witnessed violence were more likely to be perpetrators (Abrahams et al., 2004). Our research also included similar measures for women and men on attitudes, demographics and witnessing violence as a child; this permits the comparison of models across genders.

This study demonstrated that a greater percentage of women than men reported positive attitudes toward wife beating and IPV experience. This may reflect a cultural context where women accept and expect that wife beating is part of a normal union. Greater depth of information is needed from women (and men) to understand why wife beating is considered acceptable; this type of information is crucial for the development of programs to change social norms that put women at risk of IPV and to change men’s perpetration behaviors.

This study has a number of limitations. Because the data used are cross-sectional, the direction of causality could not be assessed. In particular, while it is tempting to say that persons with attitudes supportive of IPV are more likely to perpetrate or be victims of IPV, we acknowledge that the direction could be the opposite. For this reason, we undertake a multinomial logistic regression model that examines the joint outcome of attitudes and behaviors. Second, only a sub-sample of women and men in the Uganda DHS was administered the IPV module; this reduces the sample size for the models of IPV. While the reduced sample remains representative of Uganda, having smaller sample sizes reduces our ability to examine some of the regional differences associated with IPV attitudes and experience. For example, it was not possible to examine those who live in internally displaced camps compared to those who live in other parts of Northern Uganda. Third, it is common in surveys of this type that violence victimization and perpetration are under-reported because of recall and reporting biases that lead to non-disclosure of violence behaviors (Garcia-Moreno et al., 2006). The survey methodology was designed to reduce the reporting bias, but could not eliminate it entirely. Fourth, there is the possibility for under-reporting of IPV if women experienced another type of abuse not included in the survey and thus were coded as never experiencing IPV. Moreover, in this study, the measure of IPV used is ever experience of IPV. This may be under-reported by older women who may have forgotten IPV that they experienced many years ago. Finally, the question about perpetration asked of men is much less comprehensive than the series of questions women received; this may have led to men underreporting IPV perpetration.

These findings provide insights into areas to consider in future formative work for designing IPV prevention programs. Given that a high percentage of women consider violence acceptable, prevention programs will need to examine and address the root causes of these attitudes and women’s and men’s expectations for household dynamics and gender roles. Priority also needs to be given to stopping the intergenerational transmission of violence. This means incorporating programs on healthy relationships into schools to change gender attitudes and norms at an early age, undertaking pre-marital counseling with instruction about healthy relationships and negotiation skills, and working with community leaders to address the acceptability of IPV with the objective of changing unhealthy norms. Finally, screening women and men in health facilities to learn about their attitudes toward IPV and their exposure to domestic violence during their childhoods may help to identify those women and men most in need of one-on-one counseling and services to reduce their risk of subsequent IPV victimization or perpetration. These programs and studies are important for supporting the design and implementation of new and improved programs that are aimed at preventing IPV and its associated consequences in Uganda and elsewhere in sub-Saharan Africa.