Gender-norms, violence and adolescence: Exploring how gender norms are associated with experiences of childhood violence among young adolescents in Ethiopia

ABSTRACT

Adolescence is a time of particular risk for violence perpetrated by parents, teachers, peers and intimate partners. Social norms that condone violent discipline, promote masculinities focused on violence, and support gender inequality play an important role in perpetuating violence. However, little is known about the relationship between inequitable gender norms and children’s experiences of violence from parents or other adults in the household. Utilising data from the Gender and Adolescence: Global Evidence (GAGE) study, this paper explores how adolescent and household attitudes and community-level gender norms influence experiences of violence among young adolescents (aged 10–12) in Ethiopia. Our results show that community norms, rather than adolescent and household attitudes, are significantly associated with experiences of household violence. This result holds for boys and girls, and in rural areas. In urban areas, however, adolescent attitudes were more influential than community norms, perhaps indicating less cohesive communities. Overall, these findings suggest that violence prevention programming should prioritise shifting community norms, particularly in rural areas, in order to promote adolescent girls’ and boys’ right to bodily integrity.

KEYWORDS:

• Violence against children
• gender norms
• mixed-methods
• Ethiopia

Introduction

Globally, more than half of all children aged 2–17 years are estimated to have experienced physical, sexual or psychological violence in the past year (Hillis et al., 2016). Children are most at risk of experiencing violence in their own homes, where adult caregivers perpetrate physical, sexual or psychological violence as a means of expressing power or control – often in an effort to discipline their children to promote good behaviour (Devries et al., 2018; UNICEF, 2014).

Violence can have lasting effects and perpetuate a cycle of violence across generations. Girls who experience childhood violence are at greater risk of experiencing gender-based violence later in life, while male perpetrators have often experienced violence during their own childhood (Abramsky et al., 2011; Fulu et al., 2017). Exposure to violence in childhood can also impair brain development and lead to lifelong social and health challenges (Norman et al., 2012).

Social norms define ‘what is considered normal and acceptable behaviour’ for members of a specific group (Cislaghi & Heise, 2018, 2020). They describe what people think others do (descriptive norms) and what people believe others think they should do (injunctive norms) (Rimal, 2008). Both sexes can uphold social norms that condone violent discipline, promote masculinities focused on violence, and support gender inequality.

Gender norms are a subset of social norms. They are social expectations that exist outside of individual attitudes and reflect what is considered appropriate behaviour for women and men, girls and boys within specific groups or societies (Cislaghi & Heise, 2020). The different roles and behaviours of females and males, children as well as adults, are shaped and reinforced by gender norms within society. Gender norms – learned during childhood, and reinforced by family and community interactions – perpetuate power imbalances based on gender (Cislaghi & Heise, 2020; Tenenbaum & Leaper, 2002).

Research has begun to elucidate how gender norms affect behaviours and health, including violence against women and children (see, for example Pulerwitz et al., 2015; VanderEnde et al., 2012; Weber et al., 2019; Yount et al., 2018). Communities with more inequitable gender norms tend to have higher rates of violence against women – most notably intimate partner violence (Gressard et al., 2015; Heise & Kotsadam, 2015; Vyas & Heise, 2016). While much of the evidence demonstrates that inequitable gender norms are associated with increased rates of violence, some research suggests that violence may be lower in the most gender-inequitable societies as other norms effectively control women without the need for violence (Campbell, 1985). In these societies, women who are perceived to challenge these existing patriarchal gender norms may be at greatest risk of violence (Campbell, 1999; Weber et al., 2019).

For boys and girls, adolescence is a time of particular risk for violence perpetrated by parents, teachers, peers and intimate partners. For young adolescents, corporal punishment by parents and caregivers is the most common form of physical violence experienced (UNICEF, 2014). This is true in Ethiopia, where, for example, 90% of children in the Young Lives study reported having experienced violence, with physical violence most prevalent and boys most at risk (Pankhurst et al., 2018; Save the Children, 2011). Furthermore, these adolescents are coming of age in a society where gender-inequitable attitudes are normalised – particularly among women. The 2016 Ethiopia Demographic and Health Survey (EDHS) found that 63% of women and 28% of men agree that a husband is justified in beating his wife in at least one possible scenario (burning food, arguing, going out without permission, neglecting the children, or refusing to have sexual intercourse) (Central Statistical Agency (CSA) and ICF International, 2017).

As with violence against women, childhood violence has its roots in unequal power dynamics and can be affected by inequitable gender norms (Guedes et al., 2016; Maternowska et al., 2016). However, much of the existing research has focused on the connections between social norms that promote the acceptability of corporal punishment by parents and subsequent perpetration of violence, rather than the influence of gender norms specifically (see, for example, Fleckman et al., 2019; Lansford & Dodge, 2008; Taylor et al., 2011), although Leah and Cislaghi (2019) suggest that the same norms that drive violence against women also drive violence against children.

The limited evidence on inequitable gender norms and violence against children has focused on exploring links between the attitudes of female caregivers – particularly around the acceptability of intimate partner violence – and their reported use of corporal punishment, or their adolescent daughters’ self-reported experiences of violence (Falb et al., 2017; Lansford et al., 2014). One study in Pakistan examining the impact of adolescents’ gender attitudes on their likelihood of experiencing violence at home found that agreement with inequitable gender norms was associated with increased experience of violence (Saeed Ali et al., 2017).

While violence against women and violence against children share overlapping risk factors, there is a need for more research on the role of gender norms in experiences of childhood violence. This paper uses mixed methods data from the Gender and Adolescence: Global Evidence (GAGE) study to explore how individual, household and community factors influence experiences of violence within the home among young adolescents (aged 10–12) in Ethiopia.

Conceptual framework

This paper, and the wider GAGE study, focuses on the roles of power, gender and social norms within a context of multiple relationships interacting between differing domains. GAGE’s conceptual framework (2019) builds on a socio-ecological model and explores the factors that affect adolescent health and social outcomes, including the role of parental attitudes and community norms in shaping adolescent capabilities and broader wellbeing.

This paper accounts for these differing levels of influence by exploring adolescent attitudes, caregiver attitudes, and community norms on adolescent experiences of violence within the household. The authors posit that the power to adhere to or deviate from an existing norm is central to our understanding of how norms influence individual behaviour (see also Pulerwitz et al., 2019). We use a gendered approach throughout our framework and analysis, acknowledging that differing gendered exposures (e.g. school access, expectations of the roles of girls, boys, men and women, etc.) affect adolescent health outcomes (Heise et al., 2019). This article seeks to understand the specific influences of gendered attitudes and norms on adolescent experiences of violence within the home.

Methods

GAGE study design and ethics

GAGE follows 20,000 adolescents and their caregivers across six countries over nine years (2015–2024), using qualitative and quantitative research methods to better understand what works to support the development of adolescents’ capabilities. This paper uses GAGE baseline data collected in Ethiopia in 2017 and 2018.

The GAGE baseline research design and tools were approved by the George Washington University Committee on Human Research's Institutional Review Board (071721), the Overseas Development Institute Research Ethics Committee (02438), the Ethiopian Development Research Institute (EDRI/DP/00689/10), the Addis Ababa University College of Health Sciences Institutional Review Board (113/17/Ext), the Human Subjects Committee for Innovations for Poverty Action IRB-USA (14160), and the Afar, Amhara and Oromia regional Bureaus of Health ethics committees.

Data collection

The quantitative dataset includes responses from 5448 young adolescent girls and boys aged 10–12 from three diverse regions of Ethiopia (rural Afar, rural and urban areas of Amhara, and rural and urban areas of Oromia). In each site, a household census was undertaken to create the overall sample frame. In addition to the random sample, the research team purposively sampled out-of-school adolescents and adolescents with disabilities. Sampling weights, reflecting the probability of selection into the study sample, were used to make the results representative of the target population.

The qualitative data included 192 adolescents and 158 caregivers (see Table 1) who were predominantly selected from the quantitative sample based on pre-established criteria to ensure a heterogeneous sample. Where there were insufficient adolescents who were ever married or divorced in a research site, the qualitative team used a snowballing approach to identify and recruit young people into the sample, and these adolescents were also then included in the quantitative survey (see Jones et al., 2018, for details).

Table 1. Qualitative research sample with young adolescent cohort and their caregivers.

	Adolescent girls (10–12 years)	Mothers	Adolescent boys (10–12 years)	Fathers	Total adolescents	Total caregivers
S Gondar, Amhara	43	28	23	22	66	50
East Hararghe, Oromia	40	26	21	21	61	47
Zone 5, Afar	19	19	15	15	34	34
Dire Dawa City Administration	8	8	7	6	15	14
Debre Tabor, Amhara	8	8	8	5	16	13
Total	118	89	74	69	192	158

Consent (written or verbal as appropriate) was obtained for caregivers and married adolescents; written or verbal assent was obtained for all unmarried adolescents under the age of 18. Research instruments were translated into Afaan Oromo, Amharic or Afar Af’ and, in the case of the survey, administered face-to-face with tablet-based interviews by extensively trained local enumerators or, in the case of the qualitative individual and group interviews, by experienced qualitative researchers from the same region, linguistic group and sex as the interviewee. There was also a robust protocol for referral to services, tailored to the different realities of the diverse research sites.

Quantitative measures

Predictor: gender norms and attitudes

The primary independent measures explored in this paper are inequitable gender attitudes (IGAs) and inequitable gender norms (IGNs), measured at the level of the individual and household (attitudes) and the community (norms). For example, respondents were asked if they agreed, partially agreed or disagreed with statements such as ‘girls and boys should share household tasks equally’. The statements were designed to examine both descriptive norms and injunctive norms – for example, ‘most women in my community have the same chance to work outside the home as men’ (see also Baird et al,. 2019).

We summarised a series of 16 attitude questions covering the domains of education, time use, financial inclusion and economic empowerment, relationships and marriage, and sexual and reproductive health into an individual attitude scale. The Cronbach’s alpha for this scale is .65. Household data were collected from the primary female caregiver (PFC) of each child respondent (CR) covering the same domains, as well as additional questions on gender-based violence. This was summarised as a 14-point scale. The Cronbach’s alpha for this scale is .63.

Finally, community-level norms were constructed by taking the mean of individual responses for each community (kebele). Measuring community norms by aggregating individual data were noted in Costenbader et al. (2019) as a typical practice in social norms measurement, and has been used in multiple studies (Kaggwa et al., 2008; Sedlander & Rimal, 2019; Stark et al., 2018). A 12-point scale was created to summarise community-level social norms. The Cronbach’s alpha for this scale is .75. For each scale, agreement or partial agreement with more highly gender-inequitable statements was coded as a ‘1’ while disagreement was coded as a ‘0’ in the final analysis, creating summary scales with larger values reflecting agreement with more inequitable norms or attitudes.

Outcome variables

The primary outcome utilised was experience of household violence in the past 12 months. The household violence variable summarises whether a CR experienced physical violence (being pushed, slapped, hit, beaten or otherwise physically hurt by a parent or other adult in your household) or psychological violence (parent or other adult yelled at you, called you names, or treated you poorly such as withholding food from you when others in the family were fed). Secondary analysis was also undertaken for experiences of physical and psychological violence as separate outcomes.

Data analysis

Quantitative analysis

Ordinary least squares regression (linear probability model [LPM]) was used to explore the associations between IGA and IGN and violence outcomes among young adolescents.¹ The regression equation utilised was:$y_{ic}=\alpha +{\text{β}}_{1}G{N}_c+({\text{β}}_{1}G{N}_c{)}^2+{\text{β}}_{2}S{A}_H+{\text{β}}_{3}S{A}_I+X_{ic}+{\epsilon }_{ic}$where $y_{ic}$ is the outcome of interest for individual i in community c, GN_c is the measure of community IGN, SA_H is the measure of household IGA, SA_I is the measure of individual IGA, and X_ic are a set of controls. Community IGNs were included as a quadratic term due to suspected non-linearity in the relationship between norms and experiences of violence. The standard errors ε_ic are clustered at the community level, and gender and community sampling weights are used to make the results representative of the target population in the study area.

$X_{ic}$ includes the set of variables used in sampling and a set of controls that may be associated with the outcomes of interest. The sampling controls include: an indicator for whether there were multiple eligible adolescents in the household (only one per household was sampled); an indicator for the month the survey was administered (to control for any time trends); an indicator on whether the adult respondent was the female decision maker for the adolescent (to control for the small subset of adolescents for whom a male or non-primary decision maker was the survey respondent); and an indicator for whether the adolescent was part of the purposely sampled subset of the data (to control for differences in purposeful sampling across locations). The additional set of controls included: age of CR; gender of CR; whether the CR was enrolled in school; disability status; poverty level; family structure (living in a female-headed household and household size); and location. Missing data are less than 10% for the fully specified regression, and analysis suggest it is ‘missing at random’ so these observations are dropped from the regression.

These additional controls are motivated by the existing literature and help isolate the impact of gender attitudes and norms on violence. For each outcome of interest, we first explore associations within the full sample before investigating heterogeneity by gender and location (urban/rural). For each model, individual, household and community norms were explored separately (with the control variables), then all three measures were included together in the full model (see Table 5 for an example; due to space constraints, only the final models are presented thereafter).

Qualitative analysis

Researchers conducted individual and group interviews using interactive tools (Table 2) with adolescent girls and boys in rural and urban areas. The individual interviews used objects that adolescents selected as meaningful in their life as an entry point to explore the young person’s perceptions and experiences across six capability domains (GAGE consortium, 2019), including experiences of age- and gender-based violence within the home and community. These were complemented by a body mapping exercise with small groups of 10–12-year-olds to explore the norms and attitudes that shape adolescent transitions. A mapping discussion with parents unpacked the norms and practices around parenting of adolescents, including use of corporal punishment and violence.

Table 2. Qualitative tools with 10–12-year-old adolescent girls and boys, and their caregivers.

	Location		School		Disability		Total
	Urban	Rural	In	Out	Yes	No
Body mapping with adolescents	9	12	19	2	2	19	21
In-depth individual interviews	36	156	131	61	18	174	192
Community norms FGDs with parents	8	11	n/a	n/a	n/a	n/a	19

Preliminary analysis took place during daily and site-wide debriefings with the research team, the findings from which informed the development of the thematic codebook. Following data collection, all interviews were transcribed and translated by native speakers of the local language, then coded using the qualitative software analysis package MAXQDA. The codebook was shaped around the GAGE conceptual framework of ‘3 Cs’ (capabilities, change strategies, and contexts) (GAGE consortium, 2019) but allowed flexibility to incorporate local specificities. For an illustrative example of the coding framework on violence by parents, see Table 3.

Table 3. Illustrative example of qualitative thematic coding.

Code/s	Illustrative quote	Theme as delineated in paper
Violence from parents Normalisation of physical violence	When we made any kind of mistake, they will beat us. (12-year-old boy, Community F, South Gondar) I beat them when they do wrong thing. For example, I beat them when they do not do what they are ordered or when they do it in other way than what I want. (Mother, Community E, South Gondar)	Violence is normalised, esp in rural areas. Parents agree that violence is normalised and acceptable if adolescents don’t comply with demands for unpaid labour.
Violence from parents Psychological violence Denial of food	Adolescent boys and girls have been harassed/insulted by their family. If they are able they punish physically and if not they deny food and other materials. Then the youths would be nervous. (Older adolescent boy, Debre Tabor, South Gondar)	Parental violence is multi-level – physical and psychological
Violence from parents Severity	His mother and the father hold his neck down, then they tie him with a rope for the whole night and he was crying the whole time. They untied him in the morning and left without eating his lunch. (Younger boy, Community D, South Gondar) There is one man, he is our neighbour. He tied his child with rope with the big wood at home and punished him with rope. His mother shouted. The man beat her too then the neighbour intervened and saved both the mother and son. Girls are not beaten like that. (11-year-old boy, Community K, East Hararghe)	Violence is not only widespread, but can be extreme, especially for boys
Violence from parents Care work expectations for girls	They have beaten me because I broke the container which we use for carrying lunch for workers in the field. (10-year-old girl in Community J, East Hararghe) I do anything she orders me to do. (10-year-old girl, Community I, East Haraghe)	Girls are expected to comply with norms around unpaid domestic and care work
Violence from parents Going to school	My father got very upset at me saying, ‘What is wrong with you, didn’t I tell you to wait for me and I’ll get you registered?’ and he hit me on my shoulders with a stick. (12-year-old girl, Community E, South Gondar)	Adolescents, especially boys, are beaten for going to school – or failing to attend

Results

Overall, children in the sample were young (a median age of 10.93), poor, enrolled in school (over 80%), and resided in large households (average household size of more than 6 people). Boys were slightly more likely to report being in school compared to girls (p < .05). There were differences between urban and rural locations, where urban children were wealthier (p < .001), more likely to be in school (p < .001), and more likely to live with a household head who was literate (p < .001) (see Table 4 for details). These differences were controlled for during analysis.

Table 4. Descriptive statistics of attitudes, norms and experiences of violence.

	Overall	Female	Male	Urban	Rural
	Mean or Percentage (when % indicated) (s.d.)
Attitudes and norms
Adolescent attitudes (0–16, higher scores are more gender inequitable)	6.578	6.482	6.664	5.526***	6.701***
	(2.769)	(2.788)	(2.750)	(2.588)	(2.763)
Household (caregiver) attitudes (0–14, higher scores are more gender inequitable)	7.110	7.064	7.152	5.414***	7.312***
	(2.559)	(2.603)	(2.518)	(2.907)	(2.437)
Community norms (0–12, higher scores are more gender inequitable)	8.001			7.438***	8.071***
	(1.050)			(1.241)	(1.004)
Experience of violence in the past 12 months
Household violence – either physical or psychological	66.3%	65.6%	66.9%	59.4%*	67.1%*
	(47.3%)	(47.5%)	(47.1%)	(49.1%)	(47.0%)
Physical household violence	45.0%	43.7%	46.2%	39.8%	45.6%
	(49.8%)	(49.6%)	(49.9%)	(49.0%)	(49.8%)
Psychological household violence	61.2%	61.0%	61.3%	52.9%**	62.1%**
	(48.7%)	(48.8%)	(48.7%)	(49.9%)	(48.5%)
Control variables
Age	10.93	10.93	10.93	11.04**	10.92**
	(0.827)	(0.826)	(0.829)	(0.835)	(0.826)
Asset deciles	4.333	4.354	4.313	8.532***	3.839***
	(2.506)	(2.520)	(2.494)	(1.302)	(2.122)
Adolescent enrolled in school	84.4%	82.7%*	86.0%*	96.5%***	83.0%***
	(36.2%)	(37.8%)	(34.7%)	(18.5%)	(37.6%)
Adolescent has a disability	2.85%	2.64%	3.05%	3.36%	2.79%
	(16.7%)	(1.60%)	(17.2%)	(18.0%)	(16.5%)
Adolescent lives in a female-headed household	14.7%	15.9%*	13.6%*	28.0%***	13.1%***
	(35.4%)	(36.6%)	(34.2%)	(44.9%)	(33.7%)
Household size	6.260	6.257	6.263	5.175***	6.388***
	(1.868)	(1.904)	(1.834)	(1.891)	(1.823)
Household head was literate	34.7%	34.6%	34.9%	69.7%***	30.6%***
	(47.6%)	(47.6%)	(47.7%)	(46.0%)	(46.1%)
Sample size – individuals	5448	3059	2389	847	4601
Sample size – communities	217	217	217	42	175

Notes: This table summarises the independent and dependent variables from the GAGE quantitative survey. Means/percentages are weighted to make them representative of the study communities. The sample was restricted to those that have defined attitude and norm measures. There are small differences in sample sizes across outcomes. For ease of presentation, the sample size at the bottom of each column reflects the maximum sample size for that subsample; the specific sample size for each outcome can be seen in the regression tables. Parameter estimates statistically different than zero at 99.9% (***), 99% (**), and 95% (*) confidence.

Agreement with inequitable gender attitudes and norms

At the individual level, adolescents agreed with IGAs (a higher score on the scale indicates more inequitable attitudes), with a mean score of 6.578 on a 16-point scale (see Table 4). There was no statistical variation in these scores between boys and girls (girls: 6.482; boys: 6.664). However, qualitative data suggested that boys’ gender attitudes were more conservative than those of girls. There was a wider divide between respondents from urban (5.526) and rural areas (6.701) (p < .001), with adolescents in rural areas holding more inequitable views. This division was also seen in the qualitative analysis.

At the household level, female caregivers (7.110) generally held more inequitable attitudes than the adolescent cohort (6.578). There were no significant differences between caregivers’ attitudes comparing those who were reporting about their male child (7.152) or female child (7.064). However, there were differences between caregivers from urban (5.414) and rural (7.312) areas (p < .001). In addition, at the community level, urban communities reported less inequitable (7.438) norms than rural communities (8.071) (p < .001).

Experiences of violence

More than half of adolescents reported experiencing some form of household violence – physical or psychological – in the past year (66.3%), with little variation between girls (65.6%) and boys (66.9%). Similarly, the qualitative research found high levels of corporal punishment meted out by parents: ‘When we made any kind of mistake, they will beat us’, explained a 12-year-old boy from Community F (South Gondar).

Adolescent girls and boys reported a range of punishments. An 11-year-old boy in Community E (South Gondar) explained that parents ‘flog us using a stick, pinch us and punch our heads with their fist’. An 11-year-old girl from Debre Tabor (South Gondar) said that her parents ‘beat me with a cable cord. They will also slap me on my bottom’. Some adolescents reported especially cruel punishments, often for stealing. A young boy from Community C (South Gondar) explained that a friend of his was ‘tied with rope for the whole night’ and then left the next day with no food. A boy in Community F (South Gondar) added that some parents also ‘smoke’ their children, forcing them to inhale the smoke from burning hot peppers.

Children in rural areas (67.1%) reported more violence than those in urban locations (59.4%) (p < .05). The qualitative research highlighted that boys and girls were often punished physically for losing control of livestock. ‘Everyone here has been punished [flogged]. They punish you with a stick and leave you aside when you allow the cattle to graze on other people’s land’, explained an 11-year-old boy from Community H (East Hararghe). Such punishments could be severe, and cause considerable psychosocial harm. One young boy from South Gondar² reported that his brother had taken his own life following the loss of livestock and harsh punishment by his father: ‘My brother was herding goats, and one got eaten by wild animals. Then, my father beat him in the evening. Then he [the boy’s brother] hanged himself in the night.’

School attendance also appeared to cause tension between some adolescents and their parents. A 12-year-old girl in Community E (South Gondar) who left her brother in charge of the cattle when she went to school was beaten for prioritising school: ‘My father got very upset at me, saying, “What is wrong with you, didn’t I tell you to wait for me and I’ll get you registered [for school]?” and he hit me on my shoulders with a stick.’ Other adolescents reported being beaten because they did not go to school or did not do their homework. A 10-year-old boy in Community K (East Hararghe) explained, ‘If we stop studying and play the whole day outside, they will thrash us asking why we did not study’.

Overall, adolescents reported higher rates of psychological (61.2%) compared to physical (49.8%) violence. The qualitative research further nuanced these findings. While a minority of adolescents explained that violence makes them ‘frightened’ (11 year old boy, Community H, East Hararghe) or ‘nervous’ (12-year-old boy, Debre Tabor, South Gondar), for most adolescents, especially those in rural communities, corporal punishment is normalised. One 12-year-old girl from Debre Tabor (South Gondar), for example, stated that children should be beaten, because it stimulated them to learn. ‘I sometimes don’t do my homework. At such times, I want to get beaten, because I will do my homework next time.’

Relationship between violence and inequitable gender attitudes and norms

We examined adolescents’ and female caregivers’ attitudes as well as community norms to see which were associated with childhood experiences of violence.³ In the full model (see Table 5, column 4), community norms (rather than individual or household attitudes) had a significant association with adolescents’ experience of household violence in the past year (Prob > F: <.001).

Table 5. Self-reported experience of household violence in the past 12 months.

	Individual attitudes	Household attitudes	Community norms	Full model
	(1)	(2)	(3)	(4)
Scale of inequitable gender attitudes – individual	−0.007			0.000
	(0.004)			(0.004)
Scale of inequitable gender attitudes – household		−0.010*		−0.004
		(0.005)		(0.004)
Scale of inequitable gender norms – community			0.196	0.287*
		(0.115)	(0.120)
Scale of inequitable gender norms – community (squared)			0.016*	0.022**
			(0.008)	(.008)
Sample size	4979	5218	5363	4886
Adjusted R^2	0.050	0.053	0.064	0.066
Prob > F (Joint significance for community norms)			0.001	0.000

Notes: Regressions are OLS models, with coefficients displayed. All models are weighted to make them representative of the target population in the study communities and standard errors are clustered at the community level. The following variables are included as controls: whether there were multiple eligible adolescents in the household, the month the survey was administered, whether the adolescent was part of the purposely sampled subset of the data, age of adolescent, gender of the adolescent, highest level of education, disability status, poverty level, family structure and location. Parameter estimates statistically different than zero at 99.9% (***), 99% (**), and 95% (*) confidence.

The relationship between community norms and experiences of violence is curvilinear. Initially, children living in a community with more IGNs are more likely to experience violence (i.e. agreeing with one additional IGN is associated with a 28.7 percentage point increase in the likelihood of experiencing any form of household violence in the past year). However, this relationship peaks (in this case, in communities that agree with 6.5 norms on average) and reverses (with the likelihood of experiencing violence lowering about 2.2 percentage points for agreement with each additional norm). When examining experiences of physical and psychological violence within the household separately, the results mirrored the overall findings (see Table 6).

Table 6. Self-reported experience of physical and psychological violence in the past 12 months.

	Overall	Girls	Boys	Urban	Rural
	(1)	(2)	(3)	(4)	(5)
Panel A: = 1 if adolescent experienced physical household violence in the past 12 months
Scale of inequitable gender attitudes – individual	0.006	0.003	0.008	−0.027***	0.009*
	(0.004)	(0.005)	(0.005)	(0.007)	(0.004)
Scale of inequitable gender attitudes – household	−0.004	−0.008	−0.001	−0.008	−0.003
	(0.005)	(0.005)	(0.007)	(0.007)	(0.005)
Scale of inequitable gender norms – community	0.221*	0.187	0.244	0.230	0.246*
	(0.108)	(0.119)	(0.134)	(0.179)	(0.120)
Scale of inequitable gender norms – community (squared)	−0.015*	−0.012	−0.016	−0.015	−0.016*
	(0.007)	(0.008)	(0.009)	(0.012)	(0.008)
Sample size	4884	2727	2157	759	4125
Adjusted R^2	0.057	0.047	0.068	0.036	0.064
Prob > F (Joint significance for community norms)	0.117	0.290	0.181	0.441	0.121
Panel B: = 1 if adolescent experienced psychological household violence in the past 12 months
Scale of inequitable gender attitudes – individual	0.001	0.004	0.004	−0.027**	0.004
	(0.004)	(0.005)	(0.004)	(0.009)	(0.004)
Scale of inequitable gender attitudes – household	−0.003	−0.002	−0.001	−0.013	−0.001
	(0.004)	(0.006)	(0.005)	(0.007)	(0.005)
Scale of inequitable gender norms – community	0.303**	0.337**	0.355**	0.067	0.355**
	(0.112)	(0.126)	(0.125)	(0.151)	(0.125)
Scale of inequitable gender norms – community (Squared)	−0.024**	−0.026**	−0.027***	−0.005	−0.027***
	(0.008)	(0.008)	(0.008)	(0.010)	(0.008)
Sample size	4870	2149	4114	756	4114
Adjusted R^2	0.063	0.065	0.069	0.039	0.069
Prob > F (Joint significance for community norms)	0.000	0.000	0.000	0.863	0.000

Notes: See Table 5.

Differences between boys and girls

When stratifying by sex, there were similar relationships between attitudes and norms and experiences of violence. For boys and girls, community norms were significantly associated with experiences of household and psychological violence, while individual attitudes were not (see Table 6 and Table 7). However, there was no association between gender attitudes or norms and experiences of physical violence in this sub-analysis (see Table 6).

Table 7. Self-reported experience of household violence in the past 12 months by gender and location.

	Girls	Boys	Urban	Rural
	(1)	(2)	(3)	(4)
Scale of inequitable gender attitudes – individual	−0.002	0.002	−0.030***	0.003
	(0.005)	(0.005)	(0.009)	(0.004)
Scale of inequitable gender attitudes – household	−0.005	−0.003	−0.010	−0.002
	(0.005)	(0.007)	(0.007)	(0.005)
Scale of inequitable gender norms – community	0.249	0.315*	0.132	0.328*
	(0.136)	(0.136)	(0.179)	(0.135)
Scale of inequitable gender norms – community (squared)	−0.020*	−0.024**	−0.009	−0.025**
	(0.009)	(0.009)	(0.012)	(0.009)
Sample size	2728	2158	759	4127
Adjusted R^2	0.059	0.072	0.051	0.072
Prob > F (Joint significance for community norms)	0.002	0.002	0.683	0.000

Notes: See Table 5.

While in the quantitative survey gender inequitable attitudes and norms has similar impacts on boys and girls, the qualitative research highlights that there were gendered differences in reasons for violence. Girls were mainly punished for violating gender norms. They were often socialised into obedience, as one 10-year-old girl in Community I (East Hararghe) explained: ‘I do anything she orders me to do’. When younger adolescent girls were verbally insulted or physically punished, it tended to be because they were not living up to local norms around femininity and mastery of domestic chores. ‘They have beaten me because I broke the container we use for carrying lunch for workers in the field’, reported a 10-year-old girl in Community J (East Hararghe). Girls were also disciplined physically for being seen with boys and staying out at night. As a female participant in a community group discussion in East Hararghe explained, ‘You will hear her parents saying, “Didn’t I tell you not to go there? Didn’t I warn you not to go there?” And parents will also beat and force their daughter to tell them who made her pregnant.’

But there are also signs that this is changing in response to negative impacts on adolescent girls’ mental wellbeing, with young girls increasingly exercising agency and resorting to distress migration or even suicide in response to harsh parental treatment. A male participant in a different community focus group discussion in East Hararghe noted how

In the past, girls were punished with a stick till marriage. Currently there are improvements. Parents started to peacefully advise their girls because if you punish girls with stick all the time, she may get in another serious problem like migration or suicide attempt.

Differences between rural and urban communities

Findings from the rural sample were similar to the overall model (initially, agreement with more IGNs sharply increasing the likelihood of experiencing violence, peaking and then more slowly decreasing). However, this relationship did not hold for adolescents in urban areas, where community norms were not significantly associated with experiences of violence. Instead, in these locations, individual attitudes were most influential; for example, each additional IGA agreed with was associated with a 3 percentage point decrease in the likelihood of experiencing household violence (p < .001) (See Table 7).

The qualitative research findings also underscored important rural–urban differences in community attitudes towards violent discipline. Focus group discussions (male and female) in rural communities generated candid views about the types of punishments used regularly to discipline children and adolescents. As a participant in a male community focus group in East Hararghe noted:

Starting from age 1, we use the stick to punish them because it is to prevent them from repeating something. Using a stick for punishment may continue till children reach the age of 15 years when they became youthful and a bit forceful. Parents may be afraid to punish them by using stick at this age; rather, they may use advice instead.

By contrast, norms in urban communities emerged as less supportive of physical punishment, leading parents to adjust their parenting styles. As a male participant in a focus group discussion in Debre Tabor noted:

Though my children are not here with me because they are doing different jobs in parts of the country, I raised them by tying them up and beating them hard. But now, time has changed, and even if I find mistakes in the children who live with me, I don’t beat them. I don’t dare that much to beat them. I regret the way I raised my elder children.

A female participant in a discussion in Batu Town explained that this shift in attitudes was also being promoted in parent–teacher meetings: ‘When they were a child sometimes I beat them but now I don’t beat them, mostly I bark at them. The current generation doesn’t want to be beaten by parents, so not beating children is becoming a trend.’

Discussion

This paper has examined the effects of gender attitudes and norms at the individual, household and community levels on childhood violence within the home in the past year. Overall, the results demonstrate that young adolescent girls and boys in Ethiopia are coming of age in very gender-inequitable societies, in which female caregivers share beliefs about inequitable gender roles. Adolescents also hold gender-inequitable attitudes, though less so than their mothers/female caregivers. They also tended to express more equitable gender attitudes compared with community norms. Attitudes of individual caregivers were not significantly associated with experiences of violence; however, when household norms were aggregated to create community-level measures, these did affect the likelihood that a young adolescent would have experienced household violence.

This suggests that shifting the views of individuals and households towards more equitable norms is essential for violence prevention programming, especially for girls and adolescents in rural areas, who tend to be punished for non-compliance with gender norms. This finding supports emerging work from other studies suggesting that collective norms are important predictors for adolescent behaviour and health outcomes (Mahalik et al., 2015; Sedlander & Rimal, 2019) and that they may be more influential than individual attitudes (Stark et al., 2018).

While community norms were more influential than individual or household attitudes, the relationship between these norms and experiences of violence was complex. Initially, as community norms become more inequitable, there was an increased likelihood that adolescents would experience household violence. However, in communities where gendered norms have become completely ‘normalised’, violence rates plateau and slowly decrease – though rates of violence against children remain extremely high in all study communities. This may suggest some form of threshold effect, where increasingly, inequitable gender norms are no longer associated with continued increases in violence; or it could be the result of young people trying to avoid violence within the home (e.g. by running away, distress migration or suicide attempts). Further research is needed to understand these dynamics.

While experiences of violence may often be gendered, we found little difference in the influences of individual and household attitudes and community norms on violent experiences within the home. For both sexes, community norms were more influential than individual or caregiver attitudes. As these adolescents go through puberty, it is likely that gendered expectations deviate more distinctly. Our qualitative research demonstrates that gendered differences affect the reasons for violence against children at this early stage of adolescence.

There were more marked differences between the influences of individual attitudes and community norms when comparing rural and urban areas. As in the overall sample, in rural areas, community norms had most influence on rates of violence. However, this was not the case in urban areas, where individual attitudes were most influential, and where adolescents who held more equitable views reported more violence. This may suggest that young adolescents who deviate from community norms are experiencing higher rates of violence as a consequence, or it may simply be that urban areas have less cohesive communities.

Data limitations

There are some important limitations of our analysis. First, all data were cross-sectional and as such cannot demonstrate causation. Second, all data were self-reported, so it is possible that social desirability bias affected disclosure. Third, we used broad measures of violence, utilising experiences of household violence in the past year as a simple binary outcome. We did not investigate intensity of violence, which might illuminate other gender differences. Furthermore, these results reflect the influences of gender attitudes and norms on young adolescents on a specific form of violence (within the home). Different results could emerge if the same analysis was conducted with older adolescents, or for different forms of violence (e.g. in school). This will be the topic of future research.

Conclusions

Overall, this paper adds to our understanding of how community norms and individual and household-level attitudes are associated with experiences of violence among adolescents. While norms appear to work differently in diverse contexts within the study sites, overall it appears that community norms are more influential than individual attitudes for experiences of household violence for this age group. This highlights the importance of programming and communications outreach efforts that prioritise shifting wider gender and adolescent-age related norms, as part of a package of interventions to reduce violence against children. This package could include improved reporting systems, capacity strengthening for social workers, healthcare workers and justice officials to identify and respond to violence against children in the home, and parenting awareness sessions on positive disciplinary approaches. Learning is possible from examining the approaches of successful community-based programmes that have shifted gender and social norms related to violence against women in low-resource settings (see for example: Abramsky et al., 2014; Glass et al., 2019; Pulerwitz et al., 2015), though more research is needed to understand the effectiveness of such programmes on violence against children – particularly in regards to young adolescents.

Acknowledgements

The authors wish acknowledge that the GAGE program is funded by UK AID from the UK Department of International Development (DFID). The views expressed and information contained within are not endorsed by DFID, which accepts no responsibility for such views or information or any reliance placed on them. For more information please see www.gage.odi.org. We also wish to thank Megan Devonald and Eric Neumeister for research assistance support.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Additional information

Funding

GAGE is a longitudinal research programme funded by the Research and Evidence Division of the Department for International Development UK as part of UK Aid.

Notes

1 Results from the LPM instead of a logistic regression are displayed for ease of interpretation. Results are robust to the use of logistic regression.

2 Details changed to preserve child’s anonymity.

3 Correlations between scales were moderate, but not high; and coefficients remain stable across specifications.