Changing Gender Norms and Reducing Intimate Partner Violence: Results From a Quasi-Experimental Intervention Study With Young Men in Ethiopia

Abstract

Objectives. We assessed the effects of a community-based project in Ethiopia that worked with young men to promote gender-equitable norms and reductions in intimate partner violence (IPV).

Methods. A quasi-experimental design was used to assign young Ethiopian men 15 to 24 years of age (809 participants were surveyed at baseline in 2008) to an intervention involving community engagement (CE) activities in combination with interactive group education (GE) sessions promoting gender-equitable norms and violence prevention, an intervention involving CE activities alone, or a comparison group.

Results. Participants in the GE + CE intervention were twice as likely (P < .01) as those in the comparison group to show increased support for gender-equitable norms between the baseline and end-line points. Also, the percentage of GE + CE participants who reported IPV toward their partner in the preceding 6 months decreased from 53% to 38% between baseline and end line, and the percentage in the CE-only group decreased from 60% to 37%; changes were negligible in the comparison group.

Conclusions. Promoting gender equity is an important strategy to reduce IPV.

In addition to being a human rights violation, gender-based violence (GBV) is a widespread public health issue with numerous negative health consequences. GBV, including intimate partner violence (IPV), causes injury and death and decreases survivors’ use of health services.^1,2 For example, violence and fear of violence can be barriers to effective HIV prevention, care, and treatment programs. IPV has also been associated with a lack of contraception use and HIV acquisition.^3,4 Consequently, addressing and reducing GBV is an explicit component of the President’s Emergency Plan for AIDS Relief.⁵

The number of women and girls affected by IPV is enormous. A seminal study conducted in 10 different countries demonstrated that between 15% and 71% of women of reproductive age had experienced physical or sexual IPV (and estimated that 30 % of women globally had experienced IPV).^6,7 The highest rates of IPV were those among Ethiopian women: 53.7% had experienced IPV within the 12 months preceding the interview and 70.9 % over their lifetime. In addition, HIV prevalence in Ethiopia is higher among women (1.9 %) than men (1.0 %).⁸

Certain gender norms—or social expectations about men’s and women’s appropriate roles, rights, and responsibilities—have been shown to be associated with the risk of IPV as well as the risk of HIV and other sexually transmitted infections (STIs).^9–12 For example, 45% of men and 68% of women taking part in the 2011 Ethiopia Demographic and Health Survey reported that wife beating is justified for at least 1 reason.⁸

Over the past decade, a series of programs attempting to address inequitable gender norms have been implemented across the globe; many have emphasized the importance of engaging boys and men in this process.^13,14 A growing body of work has demonstrated that these gender-focused interventions can lead to reductions in violence and to other positive health outcomes (e.g., increased contraception or condom use).^9,15–19 However, rigorous evaluations of only a small number of these programs have been documented in the scientific literature, and recent literature reviews have revealed that existing evaluations have various limitations, including a lack of comparison groups and standardized or validated measures, no exploration of effects across types of IPV, inadequate follow-up rates, and limited use of theoretical frameworks.^16,17

To help address these gaps, we describe the results of a theoretically grounded, quasi-experimental intervention study of a community-based project in Ethiopia, the Male Norms Initiative, that worked with young men to promote gender-equitable norms and reductions in IPV. Our goal was to assess the effects of the intervention using standardized measures of violence and gender norms.

The 2 main intervention components were interactive group education and community mobilization and engagement activities aimed at raising awareness and promoting community dialogue. The interventions focused on promoting critical reflection regarding common gender norms that might increase the risk of violence or HIV and other STIs (e.g., support for multiple sexual partners and acceptance of partner violence). Through this reflection, the participants were able to identify the potential negative outcomes of enacting these norms and the potential positive aspects of more gender-equitable behavior.

In addition, the activities engaged the wider community in supporting a shift in specific harmful norms. Engaging Boys and Men in Gender Transformation, a manual based on EngenderHealth and Promundo’s gender-transformative programming, was used to facilitate this process.²⁰ Hiwot Ethiopia, a nongovernmental organization, led the implementation of the project with technical support from EngenderHealth. PATH led the evaluation in collaboration with Miz-Hasab, an Ethiopia-based research institute.

Both intervention groups participated in community engagement (CE) activities, which took place over a 6-month period from June to November 2008. Beginning with a march on International Father’s Day, these activities involved entire communities and included distribution of monthly newsletters and leaflets (approximately 15 000 in total), music and drama skits reaching 8700 people, monthly community workshop meetings, and distribution of more than 1000 condoms.

A second component of the intervention—group education (GE) activities—took place over 4 months at youth centers during regularly scheduled youth group hours, usually on weekends. The activities included role plays, group discussions, and personal reflection. Sessions enrolling about 20 participants were facilitated by 2 or 3 peer educators each, with oversight from a master trainer. In total, 8 sessions 2 or 3 hours in duration were conducted, drawing on 19 activities from the manual.

This intervention and evaluation design was informed by the theory of gender and power, a social structural theory that addresses environmental and social issues relating to gender dynamics, particularly sexual division of labor, sexual division of power, and the structure of cathexis (which is similar to the concept of gender norms).²¹ According to this theory, various negative health and other outcomes stem from the socialization of women to be sexually passive, women’s economic reliance on men, and abusive partnerships. The theory affirms—as does the empirical evidence provided in the introduction—that addressing gender norms is a core factor in reducing both IPV and related health risks such as HIV and other STIs.²²

METHODS

Set in 3 low-income subcities of Addis Ababa, Ethiopia, this quasi-experimental study compared the impact of the previously mentioned program activities among 3 groups of young men 15 to 24 years of age, recruited from 11 youth groups. The 3 subcities (Gulele, Kirkos, and Bole) were comparable in terms of population size, ethnic makeup, and number of municipalities, and they were distant enough from each other to minimize contamination risk. Each subcity was randomly assigned an intervention (or interventions): Gulele was assigned both group education and community engagement activities (GE + CE arm), Kirkos was assigned only community engagement activities (CE-only arm), and Bole, the comparison site, was assigned a delayed intervention after the study period.

The impact of each study arm was tested, along with the differences in impact between the 2 intervention arms and the comparison group. We hypothesized that the 2 intervention groups would report more changes in gender-related attitudes and behaviors than the comparison group and that the combined intervention would have a greater impact than the other interventions.

Study Sample

Participants in all 3 study arms tended to be young, not married, and relatively highly educated (Table 1). On average, almost 70% were currently attending school, and close to 90% were living with their families. At baseline, an average of about 30% of the participants reported having a primary partner (defined as “someone you have had a regular romantic or sexual relationship with over time, e.g., wife, girlfriend, or someone that you live with”) within the 6 months preceding the survey, with the CE-only group (Kirkos) reporting more primary partners than the other 2 groups. Overall, more than half (53%) of the participants reported having committed violence against a primary partner over the 6 months preceding baseline, with the comparison group (Bole) reporting less partner violence than the other 2 groups.

TABLE 1—

Baseline Characteristics of Participants, by Study Arm: Ethiopian Male Norms Initiative, 2008

Characteristic	GE + CE Arm (n = 235)	CE-Only Arm (n = 251)	Comparison Arm (n = 159)
Age, y, %
15–19	57.4	55.8	65.4
20–24	42.6	44.2	34.6
Age, y, median	19.0	19.0	19.0
Marital status, %
Single	97.5	98.8	97.5
Ever married	2.5	1.2	2.5
Education, %*
Primary	50.2	36.2	42.8
Secondary	38.7	53.8	48.4
> secondary	11.1	10.0	8.8
Education, y, median*	8.0	10.0	9.0
Currently in school, %	63.4	68.5	71.7
Living arrangements, %
Lives with family	93.6	89.6	89.9
Lives with friends/partner	2.6	7.6	5.0
Lives alone	3.8	2.8	5.1
GEM Scale score, mean	59.8	58.5	59.9
Had primary partner in past 6 mo,a %*	25.1	38.6	28.3
Committed any type of violence against primary partner in past 6 mo, %*	52.5	59.8	37.8

Note. CE = community engagement; GE = group education; GEM = gender-equitable men.

^aFor this variable, sample sizes were as follows: GE + CE, n = 59; CE only, n = 97; and comparison, n = 45.

*P < .05 (difference between study arms).

Quantitative Data Collection

Interviewer-administered surveys were conducted with 729 baseline and 645 end-line participants (young men from community-based youth groups in the 3 subcities). All members of the 11 groups were invited to participate at baseline and were followed up at end line; the overall response rate was 89%. Surveys were administered in May and June 2008, prior to any intervention activities (sample sizes were 244 in Gulele, 287 in Kirkos, and 198 in Bole), and 6 months later in December 2008 (with sample sizes of 235 in Gulele, 251 in Kirkos, and 159 in Bole).

Key Measures

Reflecting the primary constructs of the theory of gender and power, indicators used to assess program impact included support for equitable (or inequitable) gender norms and gender-related behaviors (e.g., IPV, HIV risk behavior). Information was also collected about exposure to the intervention (e.g., number of activities attended). We examined 2 primary outcomes in our analysis: views toward gender norms, as measured with the Gender-Equitable Men (GEM) Scale, and IPV, measured according to physical or sexual violence committed during the preceding 6 months and any type of violence (physical, sexual, or psychological) committed during the preceding 6 months.

The GEM Scale was originally developed in response to the relative lack of tools to quantitatively measure changes resulting from gender-focused interventions²³; it has since been applied in multiple cultural contexts. GEM Scale items address norms related to violence (e.g., “A woman should tolerate violence in order to keep her family together”), reproductive health and disease prevention (e.g., “Women who carry condoms on them are easy”), sexuality (e.g., “A woman who has sex before she marries does not deserve respect”), and household decision-making (e.g., “A woman should obey her husband in all things”). We used a 24-item version of the scale adapted for the Ethiopian context (Cronbach α = 0.88).¹⁰ Each GEM Scale item includes 3 response categories: agree, partially agree, and do not agree. IPV measures were derived from the World Health Organization’s Multi-Country Study on Women’s Health and Domestic Violence Against Women.²⁴

Our measure of physical or sexual violence included 7 items (e.g., slapping or hitting with hand or fist, choking, and forced sexual intercourse). In addition to these 7 items, our measure of any type of violence included 4 items focusing on psychological violence (e.g., belittling or humiliating one’s partner in front of others, or purposefully scaring or intimidating her).

Data Analysis

Our statistical analysis was limited to the 89% of young men who were surveyed at both time points (determined by individual identification numbers). Sensitivity analyses confirmed that there were no major sociodemographic differences between young men who were and were not followed up. The GEM Scale total score was created by summing scores from the 24 individual items; 1 point was given for the least gender-equitable responses, 2 points were given for moderately gender-equitable responses, and 3 points were given for the most gender-equitable responses.

Bivariate and multivariate logistic regression analyses modeled change in gender norms as a dichotomous variable to take into account linked baseline and end-line data. Participants were dichotomized into 2 groups according to whether the change in their GEM Scale score was (or was not) higher than the mean change. The models controlled for GEM Scale scores (continuous) at baseline as well as other participant characteristics. GEM Scale scores were also trichotomized (based on the range of scores) into low-, moderate-, and high-equity categories to allow a visual depiction of shifts in responses (Figure 1). The McNemar tests (for paired data) was used to demonstrate changes in individual GEM Scale item scores.

FIGURE 1—

Changes in GEM Scale score categories from baseline to end line in each study arm: Ethiopian Male Norms Initiative, 2008.

Note. CE = community engagement; GE = group education; GEM = gender-equitable men.

We used generalized estimating equations in assessing changes in violence to take into account paired data and to adjust for intrarespondent clustering among young men with primary partners at both the baseline and end-line points. We conducted the χ² test to assess differences in cross-sectional subsamples of participants with primary partners 6 months prior to the survey (which varied between the baseline and end-line points). Multivariate analyses controlled for age, GEM Scale score, intervention group, and time and included a time by intervention group interaction term.

RESULTS

Our description of our findings focuses on participants’ attitudes toward gender norms and acts of IPV.

Support for Gender Norms

At baseline, participants across all 3 arms expressed substantial support for a variety of inequitable norms. For example, most respondents (58%) agreed that a woman should tolerate violence to keep her family together. There was also support for male dominance in daily and household decision-making (e.g., 50% of the participants agreed that a woman should obey her husband in all things). Support for male norms regarding risk behaviors associated with HIV and other STIs was common; for example, 35% of the participants agreed that a man needs other women even if things with his wife are fine. However, condom use was considered relatively acceptable, with only 12% of the participants agreeing that a man should be outraged if his wife asks him to use a condom.

Overall, when results for all 24 items were combined to determine the overall GEM Scale score, respondents in the intervention groups exhibited a positive, significant shift toward support for gender-equitable norms between the baseline and end-line points. In the bivariate analysis, increases in GEM Scale scores were more likely among participants in the 2 intervention groups (GE + CE and CE only) than among those in the comparison group. The multivariate logistic regression analysis showed that there was a significant difference between participants in the GE + CE and comparison groups, with increased scores twice as likely among intervention participants as comparison participants (P < .01). About one third more young men in the CE-only group than in the comparison group showed score increases, although this comparison was not significant.

With respect to specific GEM Scale items, the GE + CE group exhibited a positive shift in 7 items, and the CE-only group exhibited a positive shift in 5 items; there were no positive shifts in the comparison group. These items addressed the entire range of domains covered by the GEM Scale, including partner violence, daily decision-making, condom use, and roles in sexual relationships. Scores on one item (“It disgusts me when I see a man acting like a woman”) shifted in the negative direction in the 2 intervention groups.

When scores were grouped into low-, moderate-, and high-equity categories (Figure 1), shifts in responses among members of the 2 intervention groups between the baseline and end-line points tended to be from low or moderate equity to high equity. Conversely, in the comparison group, shifts tended to be from the moderate- to the low-equity category.

Gender-Based Violence

At baseline, most participants with a primary partner over the preceding 6 months (62%) reported that they had ever been violent toward that partner (e.g., 28% had slapped their partner). Fifty-three percent reported some form of violence against their primary partner in the previous 6 months (24% had slapped their partner, 15% had pushed or shoved her or pulled her hair, and 29% had purposely scared or intimidated her).

By end line, the proportion of young men who reported violence toward a primary partner over the preceding 6 months had decreased significantly in both intervention groups, but there was no such change in the comparison group. In the GE + CE group, the percentage of young men who reported physical or sexual violence toward their partner decreased from 36% to 16% (P < .05), and the percentage who reported any type of violence (physical, sexual, or psychological) decreased from 53% to 38% (Figure 2). Similarly, the percentage of young men in the CE-only group who reported physical or sexual violence decreased from 36% to 18% (P < .05), and the percentage of who reported any type of violence decreased from 60% to 37% (P < .05; Figure 3). There were no changes in the comparison group in reports of either of these types of behaviors.

FIGURE 2—

Changes in physical or sexual partner violence committed in the preceding 6 months: Ethiopian Male Norms Initiative, 2008.

Note. CE = community engagement; GE = group education.

*P < .05 (difference between baseline and end line within study arm).

FIGURE 3—

Changes in any type of intimate partner violence (physical, sexual, or psychological) committed in the preceding 6 months: Ethiopian Male Norms Initiative, 2008.

Note. CE = community engagement; GE = group education.

*P < .05 (difference between baseline and end line within study arm).

In the multivariate analysis, the odds of violent behavior were reduced more in the intervention groups than in the comparison group. Over time, young men in the CE-only group were 65% less likely than those in the comparison group to exhibit any type of violence toward their partner (P = .06), and participants in the GE + CE group were 55% less likely to exhibit such behavior. Similar results were found for physical or sexual violence, with participants in both intervention groups being 60% less likely to be physically violent over time than participants in the comparison group. Furthermore, agreement with more equitable gender norms was associated with a trend toward a reduction in physical violence. Specifically, high-equity GEM Scale scores were associated with a 34% reduction in the odds of any type of violence (P = .08).

DISCUSSION

Our findings indicate that the Male Norms Initiative in Ethiopia successfully influenced participants’ attitudes toward both gender norms and acts of IPV. At baseline, levels of IPV among the participating young men were high (consistent with data from a nationally representative survey in which most respondents justified partner violence under various circumstances⁸). Whereas participants in each of the intervention groups reported less violence over time, no similar positive change was observed in the comparison group. This finding is consistent with previous reviews of IPV prevention interventions in which it has been concluded that the most effective interventions include individual-level curricula, community-based activities, and, often, multiple components.¹⁷

In our multivariate IPV analyses, however, only the findings from the community engagement (CE-only) intervention remained marginally significant; these results were likely influenced by the relatively small proportion of young men with primary partners and related effects on our statistical power to detect change. Overall, our findings suggest that both interventions resulted in positive changes, offering the potential for pursuing more than 1 strategy, each with its related pros and cons; for example, group interventions can be resource intensive, whereas community-based activities can be quite diffuse.

Young men in the combined intervention group (GE + CE) but not the comparison or CE-only group reported more support for gender-equitable norms at the end-line point. Evidence from evaluations in other contexts suggests that interactive group education is a key component of this type of intervention and may be necessary to sufficiently influence often deep-seated and complex gender-related norms (including those existing in Ethiopia).^11(p288) Of note, whereas almost all GEM Scale item scores shifted in the more equitable direction, scores on the item focusing on men “acting like” women shifted in the negative direction; further research is needed to understand this result.

In addition to providing new information about successful IPV prevention strategies, our study makes an important contribution by addressing many of the evidence gaps identified in the literature to date.^16,17 The study included a rigorous design with a comparison group, maintained a high follow-up rate, incorporated standardized and validated outcome measures, and examined more than one type of IPV. However, several limitations should also be highlighted. First, participants were selected from existing youth groups, so they may not have been representative of the general population of young people. Second, although the interventions were randomly assigned according to community, the 3 communities in which the interventions were implemented were not randomly selected (because such a random selection would not have been logistically feasible). As a result, even though similar communities were selected and participants had many comparable sociodemographic characteristics and comparable GEM Scale scores, rates of violence among individual participants differed somewhat at baseline.³

Third, there is always the risk of social desirability bias (or people’s tendency to answer questions in a manner perceived to be favorable to others) when self-reported data are used, especially in the case of sensitive issues; however, our use of a comparison group and our inclusion of interviewers who were not involved with the intervention should have helped minimize this risk. Finally, it was not possible to conceal study group assignments from participants or study personnel, and thus performance bias cannot be ruled out.

Our results showed a strong connection between positive shifts in views toward gender norms and reductions in reported violence over time. The public health implications of GBV are far reaching, and our findings suggest that confronting inequitable gender norms is an important element of IPV prevention strategies. Our study contributes to the evidence base regarding which intervention modalities can be most successful in preventing such violence.