Differential impacts of bullying and sexual violence on girls’ self-esteem in Nigeria: the buffering role of social support

Abstract

Background

School-related gender-based violence (SRGBV) poses serious risks to adolescent girls’ safety, learning, and mental health. This study assessed whether perceived social support – from family, friends, and significant others – buffers the negative effects of bullying and sexual violence on girls’ self-esteem in Nigeria.

Methods

Using data from 5,936 secondary schoolgirls (aged 15–20 years) drawn from the Adolescent Girls’ Initiative for Learning and Empowerment (AGILE) project, we employed multilevel linear models with random intercepts for schools to test the direct (main) effects and within-level moderation of social support and SRGBV on girls’ self-esteem, adjusting for student- and school-level covariates.

Results

Both bullying and sexual violence were significantly associated with lower self-esteem, with standardized coefficients indicating a stronger negative association for sexual violence (β ≈ −0.07, p < .001) than for bullying (β ≈ −0.04, p < .01). Perceived social support showed significant positive main effects on self-esteem across all sources (p < .001) and moderated the relationship with bullying, such that higher support from family (β ≈ 0.031, p < .01), friends (β ≈ 0.024, p < .05), and overall support (β ≈ 0.023, p < .05) attenuated its negative effect. No moderating effect was observed for sexual violence for any support source (all interaction p > .05), indicating that perceived social support did not offset the self-esteem harms associated with sexual victimization.

Conclusion

Our study demonstrates that everyday social ties, particularly support from family and peers, can protect girls’ self-esteem from the harms of bullying but appear insufficient to mitigate the severe impact of sexual violence. This suggests that effective interventions require an integrated approach: strengthening relationship-based networks to prevent and buffer bullying, alongside establishing specialized, trauma-informed responses to address sexual violence.

Supplementary Information

The online version contains supplementary material available at 10.1186/s40359-026-04083-7.

Introduction

School-related gender-based violence (SRGBV) is a particularly harmful form of violence that undermines the promise of education as a pathway to health, dignity, and economic opportunity, especially in low- and middle-income countries (LMICs) where violence against children remains widespread and consequential. Even when the true scope of the problem is obscured by unavailable or limited data, available statistics across several contexts reveal it to be pervasive. In Nigeria, for instance, population-based estimates from the 2014 Nigeria Violence Against Children Survey (VACS) indicate that six in ten children experience at least one form of violence, and one in four girls report sexual violence in childhood, with far too few receiving services or justice afterwards [1].

By definition, SRGBV encompasses “acts or threats of sexual, physical, or psychological violence occurring in and around schools because of stereotypes and roles or norms attributed to sex or gendered identity” [2], manifested in teacher- or staff-perpetrated abuse (e.g., corporal punishment, sexual coercion) and peer-perpetrated behaviors (e.g., bullying, sexual harassment, non-consensual touching, discriminatory verbal abuse) that occur in classrooms, toilets, and dormitories, and on routes to and from school. These behaviors are entangled with social norms that normalize corporal punishment, sanction verbal degradation as discipline, and privilege male dominance [3].

The educational and psychosocial consequences of SRGBV are both severe and far-reaching. Studies have shown that exposure to sexual, physical, or emotional violence is linked to absenteeism, dropout, reduced concentration, and erosion of belonging [4–7]. Girls, in particular, often alter their educational plans due to trauma or fear of revictimization [3]. Evidence from Northern Nigeria shows that children who have experienced violence are more likely to be out of school, less likely to feel safe commuting, and—among girls—more likely to experience early marriage [8], thereby hindering their ability to pursue and complete their education. The impact of SRGBV is also evident in primary education. A study of Nigerian public primary schools found that students exposed to such violence report depressive symptoms, fear of peer interactions, and diminished self-confidence and self-esteem [9], which are key psychological resources for persistence, academic identity, and later-life outcomes [5, 10].

In light of these consequences, the need for a safer and more supportive school environment that safeguards students’ well-being has become a matter of critical urgency. The practice literature converges on “whole school” approaches that mobilize leadership, codes of conduct, teacher training, student empowerment, safe physical environments, and parent/community engagement to create enabling climates that prevent and respond to violence [11]. Such approaches aim to transform norms and power relations, institutionalize reporting and accountability, and cultivate respectful relationships and positive discipline [11], thereby addressing the structural drivers of violence in theory. However, in practice, their implementation remains a major challenge, particularly in LMICs, where such programs are often incomplete or entirely absent [12]. This is evident in sampled Nigerian public primary schools, where barely a third reported having any SRGBV program, and fewer than half of the teachers had received SRGBV-specific training [9], suggesting considerable variability in the availability and quality of school-based support.

This context provides us with an empirical foundation to investigate whether informal support systems can serve as leverage points for students at risk of SRGBV. Theoretically, the stress-buffering model posits that social support can mitigate the adverse psychological effects of stressors by shaping cognitive appraisal, providing emotional reassurance, and enhancing coping resources (e.g., problem-solving and cognitive reappraisal) [13]. Recent theoretical advances have further proposed that the timing of support matters: whether it occurs concurrently with, prior to, or following exposure determines its function in preventing acute stress responses and fortifying resilience capacities [14]. In adolescent populations, empirical literature has consistently linked stronger perceived support to lower levels of psychological distress, anxiety, and depression, and to higher subjective well-being [15–17], with some studies suggesting that girls may both report more internalizing symptoms and be more sensitive to the benefits of supportive social networks than boys [18].

Building on this perspective, we advance two linked hypotheses. First, consistent with global and Nigeria-specific evidence, we hypothesize that exposure to SRGBV, specifically bullying and sexual violence, will exert a direct negative influence on girls’ psychosocial well-being, particularly their self-esteem. Second, in line with the stress-buffering model, we propose that perceived social support from family, friends, and significant others moderates this relationship, such that higher levels of support attenuate the negative association between SRGBV exposure and self-esteem. The conceptual model illustrated in Fig. 1 outlines two potential pathways: for girls with high support, SRGBV exposure is buffered, resulting in preserved self-esteem; for those with low support, exposure leads to diminished self-esteem.

Fig. 1.

Conceptual Model

To explore these pathways, we employed a multilevel analytic strategy that captures girls’ individual exposure to bullying and sexual violence and the protective resources in their interpersonal networks while explicitly modeling the nesting of students within schools. This approach is essential because both SRGBV exposure and access to support are structured across ecological layers: schools differ in peer dynamics, teacher responsiveness, reporting mechanisms, and overall safety. Similarly, households vary in parental support and communication, and communities differ in gender norms and expectations for girls’ education. Failing to account for this nesting structure in the dataset risks missing where meaningful variance lies and can lead to biased estimates, incorrect inferences, and misleading conclusions about relationships between SRGBV and girls’ self-esteem.

Our contributions to the literature are both empirical and practical. Empirically, we respond to a mixed record of evidence on whether social support buffers the psychological impacts of school-based victimization. For example, evidence from California indicates that higher adult support in schools is associated with lower overall serious psychological distress; however, it does not significantly interact with bullying exposure, meaning that support operates primarily as a main effect rather than as a moderator [19]. In contrast, findings from Pakistan indicated that social support moderated the effect of sexual harassment on shame and mental health, though not on self-esteem [20]. For Nigeria, however, little is known about this moderating role, despite studies documenting direct effects, such as evidence linking school bullying to reduced mental health [21]. Practically, we aim to generate contextually grounded evidence to guide culturally sensitive, relationship-centered interventions within a Whole School Approach, with the goal of reducing the psychological harm of SRGBV and fostering resilient, healthy self-esteem among adolescent girls. If social support buffers the harm of SRGBV on self-esteem, the findings will justify testing out programs that strengthen supportive relationships for girls while simultaneously advancing whole-school policies and systems to prevent, identify, and respond to SRGBV.

Materials and methods

Data source

This study draws on publicly available data from the Adolescent Girls’ Initiative for Learning and Empowerment (AGILE) project conducted in rural and semi-urban areas of Nigeria [22].1 The AGILE project assesses two linked intervention components: safe-space-based life skills training and combined life skills plus digital literacy training, targeting secondary school girls aged 15–20 years across three Nigerian states: Kaduna, Kano, and Katsina. The baseline survey, which formed the primary data source for our analysis, was administered through face-to-face interviews over a 44-day period between April and June 2023. It covered 8,223 first-year senior secondary school girls, 8,007 caregivers, and 270 principals in the sample. For this study, we worked mainly with the student dataset, which included direct measures of self-esteem, experiences of school-related gender-based violence (such as bullying and sexual harassment), and perceived support from family, friends, and other significant people in their lives. Our analytic sample included 5,936 students with complete data on school violence, social support, self-esteem, and other relevant variables.

Variable description and measurement

SRGBV was operationalized using two domains of school-based victimization: bullying and sexual violence, based on students’ self-reports from the previous school term.

Bullying

Bullying was assessed using the 9-item standardized and validated bullying subscale of the School-Related Gender-Based Violence Toolkit [23], spanning verbal, social, physical, and psychological aggression. These items capture experiences such as being teased or mocked, being called offensive names, being deliberately excluded from activities or peer groups, having belongings stolen or damaged, being physically assaulted (e.g., pushed, hit, or kicked), receiving threats, being coerced into harming others, and having false stories spread to damage one’s reputation. Each item was responded to and scored on a 4-point scale (0 = never, 1 = once, 2 = a few times, and 3 = many times). A composite bullying score was constructed by summing all nine items, with higher values indicating greater exposure to peer victimization within the school setting. The original development and validation of the scale reported an acceptable internal consistency (Cronbach’s alpha) of 0.73 for students aged 15–18 years [23]. In our study, the bullying subscales yielded high inter-item correlations and an overall Cronbach’s α of 0.81. The complete list of bullying items, together with their reliability coefficients, is provided in Table A1 of the supplementary materials.

Sexual violence

Sexual violence was assessed using 13 items also drawn from the sexual violence subscale of the SRGBV Toolkit [23], which capture unwanted sexual attention, coercion, and assault within school settings. The items included being spied on while undressed, exposure to another person’s private parts or sexual materials, being touched or forced to touch someone inappropriately, being subjected to sexual comments or jokes, and being offered gifts or favors in exchange for sexual acts. Each item was responded to and scored on a 4-point scale (0 = never, 1 = once, 2 = a few times, and 3 = many times). The entire 13-item instrument was summed to produce a composite measure of sexual violence exposure, with higher scores denoting more frequent or diverse experiences of sexual victimization within the school environment. The original development and validation of the scale reported a high internal consistency (Cronbach’s alpha) of 0.86 for students aged 15–18 years [23]. In this study, the sexual violence subscales yielded high inter-item correlations and an overall Cronbach’s α of 0.85. The full item wording and reliability scores are provided in Table A2 of the supplementary materials.

Perceived social support

Perceived social support was measured with the standardized Multidimensional Scale of Perceived Social Support (MSPSS) [24], a 12-item scale capturing emotional, instrumental, and companionship support from three sources: family, friends, and a “special person.” Students rated statements such as “There is a special person who is around when I am in need,” “My family really tries to help me,” and “I can count on my friends when things go wrong” on a 7-point Likert scale (1 = very strongly disagree to 7 = very strongly agree). We computed source-specific composite scores by summing the relevant items for family, friend, and special-person support, as well as an overall perceived social support score by summing all 12 items. Higher scores indicate stronger perceived support. The MSPSS has been used and validated in many Nigerian contexts, including among students [25–27]. In this study, the scale yielded a good range of inter-item correlations and Cronbach’s α of 0.83, 0.79, 0.78, and 0.75 for overall social support, family support, friend support, and special person support, respectively. The item composition and reliability statistics for the overall and source-specific scales are reported in Tables A3–A6 of the supplementary materials.

Self-esteem

Self-esteem was assessed using ten statements adapted from the Rosenberg Self-Esteem Scale (RSES) [28]. Items reflect positive and negative self-perceptions (e.g., “On the whole, I am satisfied with myself” and “I certainly feel useless at times”) and were rated on a 4-point Likert scale (1 = strongly agree to 4 = strongly disagree). Negatively worded items were reverse-coded to ensure that higher values consistently reflected higher self-esteem. We summed the ten items to generate a composite self-esteem score representing students’ global self-worth and perceived value. The RSES has also been applied to various Nigerian contexts, including among students [29–32]. In our study, the RSES yielded acceptable inter-item correlations and a Cronbach’s α of 0.72. The detailed item wording and reliability are presented in Table A7 of the supplementary materials.

Covariates

In line with extant literature, we controlled for a number of student- and school-level characteristics that could confound the associations of interest. At the student level, these included age (measured in years), visual difficulty [33] (a binary variable indicating whether the student reported trouble seeing), and boarding status [34] (whether the student lived in school). At the school level, we accounted for school location (classified as urban, semi-urban, or rural) and a school infrastructure index [35] that captured the presence of the following foundational resources: blackboards in classrooms, adequate student seating, clean drinking water, and functional toilets, reflecting the school’s physical learning environment. All school-level measures were taken from the school/principal dataset and merged with the student dataset prior to the analysis. Table 1 provides detailed definitions and coding of all covariates used in the models.

Table 1.

Variable descriptions and summary statistics (N = 5,936)

Variable	Description/Scale	Statisticsa
Self-Esteem	Total self-esteem score (21–40)	30.03 (3.69)
Total Support	Overall social support score (12–84)	76.39 (7.59)
Family Support	Support received from family (4–28)	26.44 (2.23)
Friend Support	Support received from peers (4–28)	25.19 (3.20)
Special Person Support	Support from a special person (4–28)	24.76 (4.02)
Bullying Experience	Bullying incidents score (0–27)	3.65 (4.39)
Sexual Violence	Sexual harassment incidents score (0–39)	0.89 (2.70)
Age	Student’s age in years	15.89 (0.99)
Boarding Status	Student’s residential status	Day student: 5,210 (87.8%) Boarding student: 726 (12.2%)
Visual Difficulty	Presence of visual impairment	No difficulty: 5,711 (96.2%) With difficulty: 225 (3.8%)
Location	Geographical area of residence	Rural: 2,178 (36.7%) Semi-urban: 1,063 (17.9%) Urban: 2,695 (45.4%)
School Infrastructure	Quality of school facilities (1–4 scale)	3.08 (0.67)

^aContinuous variables are presented as mean (standard deviation) and categorical variables as count (percentage)

Analytical method

We used multilevel modelling (also known as hierarchical or nested modelling) to examine how exposure to SRGBV relates to girls’ self-esteem and whether perceived social support from family, friends, or a special/significant person moderates this relationship. This modelling approach was appropriate because of the nested structure of the dataset: students (Level 1) clustered within schools (Level 2). It accounts for the non-independence of observations and partitions the variance into student- and school-level components.

Model specifications

We began with a baseline model containing no predictors to decompose the variance in self-esteem into within- and between-school components and to calculate the intraclass correlation coefficient (ICC). The model is specified as follows:

Here, ​ is the self-esteem score for student in school , ​ is the grand mean, is the school-level random intercept, and is the individual-level error term. The ICC from this model shows the proportion of the total variance in self-esteem attributable to differences between schools.

Next, we added the main effects of SRGBV (bullying and sexual violence) and perceived social support (overall, family, friends, or a special person), along with student-level covariates (age, visual difficulty, and boarding status) and school-level covariates (school location and infrastructure). The specifications are as follows:

where is either the bullying or sexual violence score, is any of the composite support measures, is vectors of student-level covariates, and ​ is the vector of school-level covariates.

Finally, we tested whether social support buffers the relationship between SRGBV and self-esteem by including the interaction terms between SRGBV and support measures. Each moderation model focused on one form of SRGBV and a source of support. The model is as follows:

Here, ​ captures whether the link between violence exposure and self-esteem is dependent on perceived support levels. All focal constructs (SRGBV exposure, support types, and self-esteem) were standardized prior to analysis, which facilitates interpretation of coefficients as effect sizes in standard deviation units. The models were estimated using the maximum likelihood (ML) estimator in Stata 17.0.

Results

Descriptive evidence

Table 1 summarizes the sample characteristics based on the study variables. The analytic sample comprised 5,936 adolescent girls enrolled in 269 Nigerian secondary schools. The students were aged 15–20 years, with a mean age of 15.89 years (SD = 0.99). A small proportion resided in school dormitories (n = 726, 12.2%), while the majority were day students (n = 5,210, 87.8%). They came from diverse locations: 36.7% from rural areas (n = 2,178), 17.9% from semi-urban areas (n = 1,063), and 45.4% from urban areas (n = 2,695). Most reported no visual difficulties (n = 5,711, 96.2%), with only 3.8% (n = 225) indicating some level of visual impairment. Schools’ infrastructure quality was rated moderately high, with a mean score of 3.08 (SD = 0.67) on a 1–4 scale.

Regarding the focal constructs, the mean self-esteem score was 30.03 (SD = 3.69) on a scale of 21–40, indicating moderately high self-esteem. Perceived social support scores were also generally high: family support (M = 26.44, SD = 2.23), friend support (M = 25.19, SD = 3.20), and support from a special person (M = 24.76, SD = 4.02). The composite overall support score was 76.39 (SD = 7.59) on a possible range of 12–84. Exposure to SRGBV was comparatively low but present in this sample, with bullying experiences averaging 3.65 (SD = 4.39) on a scale of 0–27 and sexual violence incidents averaging 0.89 (SD = 2.70) on a scale of 0–39.

Analytical evidence

Tables 2 and 3 present the results for bullying and sexual violence as predictors of girls’ self-esteem. Across all models, the intraclass correlation coefficients (ICC) ranged from 0.109 to 0.115, indicating that differences between schools accounted for approximately 11% of the variance in self-esteem, while the remaining ~ 89% was attributable to differences between students. This level of clustering supports the use of a multilevel approach to the analysis.

Table 2.

Direct effects of bullying on Self-Esteem

Predictor	Dependent Variable: Standardized Self-Esteem Score
	Total Support	Family Support	Friend Support	Special Support
Bullying (std)	−0.037∗∗ (0.013)	−0.038∗∗∗ (0.013)	−0.037∗∗ (0.013)	−0.040∗∗∗ (0.013)
Support (std)	0.092∗∗∗ (0.013)	0.092∗∗∗ (0.013)	0.052∗∗∗ (0.013)	0.079∗∗∗ (0.013)
Age	0.002 (0.013)	0.003 (0.013)	0.003 (0.013)	0.001 (0.013)
Visual Difficulty	−0.160∗∗ (0.065)	−0.159∗∗ (0.065)	−0.155∗ (0.065)	−0.167∗∗ (0.065)
Boarding Status	−0.128∗ (0.064)	−0.125† (0.064)	−0.121† (0.065)	−0.127∗ (0.065)
Location (Semi-urban)	0.071 (0.068)	0.067 (0.068)	0.063 (0.068)	0.065 (0.069)
Location (Urban)	0.187∗∗∗ (0.053)	0.185∗∗∗ (0.052)	0.187∗∗∗ (0.053)	0.189∗∗∗ (0.053)
School Infrastructure	0.078∗ (0.035)	0.079∗ (0.035)	0.079∗ (0.035)	0.079∗ (0.035)
ICC	0.113	0.111	0.115	0.115

Each column represents a separate model for each type of social support. Key variables (bullying, support type, and self-esteem) were standardized (mean = 0, SD = 1). The standard errors are in parentheses

† p <.10, * p <.05, ** p <.01, *** p <.001

Table 3.

Direct effects of sexual violence on self-esteem

Predictor	Dependent Variable: Standardized Self-Esteem Score
	Total Support	Family Support	Friend Support	Special Support
Sexual Violence (std)	−0.069∗∗∗ (0.013)	−0.069∗∗∗ (0.013)	−0.071∗∗∗ (0.013)	−0.069∗∗∗ (0.013)
Support (std)	0.091∗∗∗ (0.013)	0.090∗∗∗ (0.013)	0.052∗∗∗ (0.013)	0.076∗∗∗ (0.013)
Age	0.004 (0.013)	0.005 (0.013)	0.005 (0.013)	0.003 (0.013)
Visual Difficulty	−0.166∗∗ (0.065)	−0.165∗∗ (0.065)	−0.161∗ (0.065)	−0.173∗∗ (0.065)
Boarding Status	−0.133∗ (0.064)	−0.131∗ (0.063)	−0.127∗ (0.064)	−0.133∗ (0.064)
Location (Semi-urban)	0.062 (0.068)	0.059 (0.067)	0.055 (0.068)	0.057 (0.068)
Location (Urban)	0.177∗∗∗ (0.052)	0.174∗∗∗ (0.052)	0.176∗∗∗ (0.053)	0.173∗∗∗ (0.053)
School Infrastructure	0.076∗ (0.035)	0.077∗ (0.035)	0.076∗ (0.035)	0.077∗ (0.035)
ICC	0.111	0.109	0.113	0.114

Each column represents a separate model for one type of social support. Key variables (sexual violence, support type, and self-esteem) were standardized (mean = 0, SD = 1). The standard errors are in parentheses

* p <.05, ** p <.01, *** p <.001

Direct effects of SRGBV on girls’ self-esteem

Bullying was consistently and negatively associated with self-esteem across all specifications (β ≈ −0.037, p <.01). This means that a one standard deviation (SD) increase in bullying experience was linked to a 0.037 SD decrease in self-esteem. The effect size was consistent, regardless of the type of social support included in the model. In the same models, all four forms of perceived support were positively associated with self-esteem: total support (β = 0.092, p <.001), family support (β = 0.092, p <.001), friend support (β = 0.052, p <.001), and support from a special person (β = 0.079, p <.001).

When sexual violence replaced bullying as the SRGBV predictor (as in Table 3), the results similarly showed a significant negative association with self-esteem (β = −0.069, p <.001), nearly double the magnitude of the bullying effect in standardized terms. This indicates that a one SD increase in sexual violence experience corresponds to a 0.069 SD decrease in self-esteem. Again, each support dimension exhibited a positive main effect: total support (β = 0.091, p <.001), family support (β = 0.090, p <.001), friend support (β = 0.052, p <.001), and special-person support (β = 0.076, p <.001). Collectively, these findings confirm that higher perceived support, regardless of the source, is linked to better self-esteem, while greater SRGBV exposure is related to lower self-esteem.

Several covariates showed consistent associations with self-esteem in both bullying and sexual violence models. Students who reported visual difficulty had significantly lower self-esteem (β ≈ −0.166, p <.01), and those in boarding schools also reported lower self-esteem (β ≈ −0.133, p <.05). School location played a role, with students in urban schools reporting higher self-esteem than their peers in rural schools (β ≈ 0.177, p <.001), whereas differences between rural and semi-urban schools were not statistically significant. In addition, school infrastructure quality was positively associated with self-esteem (β ≈ 0.076, p <.05).

Conditional effects of SRGBV on self-esteem at levels of social support

Table 4 presents the results of the analysis examining whether different dimensions of social support moderated the relationship between SRGBV (bullying and sexual violence) and self-esteem. The interaction results showed that bullying’s negative association with self-esteem was buffered by overall social support (β = 0.023, p <.05), family support (β = 0.031, p <.01), and friend support (β = 0.024, p <.001). In other words, the more support girls perceived in these areas, the less bullying eroded their self-esteem. However, support from special persons did not significantly moderate the bullying–self-esteem relationship (β = 0.013, p >.05).

Table 4.

Moderation effects of social support on the relationship between SRGBV and Self-Esteem

Dependent Variable: Standardized Self-Esteem Score
Support Type	Predictor	Interaction (β)	Simple Slopes
Total Support	Bullying	0.023∗ (0.011)	Low: −0.056∗∗∗ Avg: −0.034∗∗ High: −0.011
	Sexual Violence	−0.013 (0.010)	–
Family Support	Bullying	0.031∗∗ (0.012)	Low: −0.066∗∗∗ Avg: −0.035∗∗ High: −0.004
	Sexual Violence	0.000 (0.010)	–
Friend Support	Bullying	0.024∗ (0.011)	Low: −0.059∗∗∗ Avg: −0.035∗∗ High: −0.011
	Sexual Violence	−0.009 (0.010)	–
Special Support	Bullying Sexual Violence	0.013 (0.012) −0.016 (0.010)	– –

All models controlled for student- and school-level covariates. Key variables (SRGBV, support types, and their interactions) were standardized before the analysis. The standard errors for the interactions are in parentheses. Simple slopes were evaluated at ± 1 SD of support. * p <.05, ** p <.01, *** p <.001

Simple slope analysis clearly illustrated this buffering pattern (as shown in Fig. 2). At low levels of overall support, bullying was strongly linked to lower self-esteem (β = −0.056, p <.001). This association weakened at moderate levels of support (β = −0.034, p <.01) and became statistically insignificant at high levels (β = −0.011, p >.05). The same trend was evident for family support: low (β = −0.066, p <.01) and moderate (β = −0.035, p <.001) support levels were linked to diminished self-esteem, but high family support (β = −0.004, p >.05) nullified this negative association. Friends’ support followed a similar pattern, with low (β = −0.059, p <.001) and moderate (β = −0.035, p <.01) levels linked to reduced self-esteem, but high levels (β = −0.011, p >.05) eliminated the bullying effect. Taken together, these interactions indicate that robust support from family, friends, and across sources, can neutralize the adverse association between bullying and self-esteem.

Fig. 2.

Simple slopes analysis of bullying effect on self-esteem at levels of support

In contrast, none of the social support dimensions significantly moderated the relationship between sexual violence and self-esteem (overall support: β = −0.013, p >.05; family: β = 0.000, p >.05; friends: β = −0.009, p >.05; special persons: β = −0.016, p >.05). Given that sexual violence had a larger negative coefficient than bullying in the direct-effects models, these findings suggest that the severity and potential trauma of sexual victimization may overwhelm the buffering capacity of the available support as measured here. In substantive terms, while relationship-based resources appear sufficient to blunt or even eliminate the self-esteem costs of bullying, they are not, on their own, adequate to offset the deeper harms associated with sexual violence.

Discussion and practical implications

This study examined the associations between SRGBV, specifically bullying and sexual violence, and the self-esteem of secondary school girls in Nigeria, while also exploring whether different sources of social support (family, friends, and special persons) could directly and interactively influence this relationship. Consistent with our expectations and in line with prior empirical evidence [36, 37], we found that both bullying and sexual violence are significant negative predictors of girls’ self-esteem. This means that girls who reported higher exposure to either form of violence tended to have lower self-esteem. Importantly, the magnitude of the negative association was greater for sexual violence than for bullying, underscoring the particularly damaging nature of sexual victimization in the school context. This is particularly concerning for adolescent girls, whose self-concept and personal identity are still in formative stages, and given the fact that self-esteem is a critical resource linked to mental health, resilience, academic performance, and overall life outcomes [5, 10], these findings make it imperative to identify mechanisms, both preventative and remedial, that can protect and restore self-worth among affected students.

In terms of protective factors, our results showed that perceived social support from all three examined sources (i.e., family, friends, and special persons) was positively associated with self-esteem. Regardless of violence exposure, girls who reported stronger support networks tended to have higher self-esteem. When we examined moderation effects, we found that social support, particularly from family and friends, played a clear buffering role in the bullying-self-esteem relationship. Specifically, girls with low or moderate levels of these supports experienced significant declines in self-esteem when bullied, but those with high levels of support did not show a statistically significant link between bullying and self-esteem. This pattern supports the buffering hypothesis of social support [13], which posits that support is most protective under conditions of high stress. For adolescent girls navigating school environments where violence is prevalent, such support may provide emotional validation, practical advice, and a sense of belonging that help counteract feelings of worthlessness or isolation. In practice, this suggests that fostering strong, reliable connections, whether through family engagement, peer support programs, or broader school-community initiatives, can meaningfully reduce the psychological toll of bullying.

Interestingly, the same buffering effect was not observed for sexual violence. None of the three sources of support moderated the sexual violence–self-esteem relationship, meaning that even high levels of perceived support did not offset the harm caused by sexual victimization. This divergence from the bullying results warrants close attention. It suggests that the severity, stigma, and emotional trauma associated with sexual violence are so overwhelming in this context that available support resources, as they currently function, are insufficient to address the resulting psychological damage. Among other factors, cultural norms that silence victims, fear of retaliation, and a lack of specialized trauma-informed support within both families and schools could all contribute to this gap. Therefore, education policies should adopt a dual-track approach, implementing prevention and intervention measures tailored specifically to bullying on one hand and sexual violence on the other, reflecting their unique dynamics, severity, and long-term consequences.

Summary, limitations, and directions for future research

Despite the insights from this study, some key limitations should be noted when interpreting and generalizing its findings. One of them is the cross-sectional nature of the data, which restricts the ability to make causal claims. Consequently, the results reported here should not be interpreted as establishing a causal link. To address this limitation, we recommend that future research employ longitudinal designs that allow for stronger conclusions regarding temporal relationships and causality. Similarly, given the self-report nature of the data, there is a potential for response bias, particularly in relation to sensitive topics. Future studies should consider multi-informant approaches (e.g., teacher observations and counsellor logs), anonymous digital reporting tools, and vignette-based measures to reduce bias and improve response accuracy.

Methodologically, our use of multilevel modelling in this study, while appropriate for the nested nature of the dataset, was insufficient to identify the specific school-level variables that influence students’ self-esteem. To overcome this, we recommend that future research incorporate school climate and safeguarding indicators (e.g., teacher training coverage, code-of-conduct enforcement, availability/quality of reporting systems) and test cross-level interactions to examine whether supportive school ecologies amplify the buffering effects of interpersonal support.

In summary, this study advances our understanding of how school-related gender-based violence (SRGBV) affects the self-esteem of secondary school girls in Nigeria and clarifies the role of social support in mitigating these effects. The findings indicate that promoting social support not only buffers the negative impact of SRGBV but also fosters overall self-esteem and resilience. Consequently, designing, testing, and implementing interventions that actively cultivate consistent access to social and emotional support is essential. Future programs must integrate this universal support with targeted, trauma-informed interventions for specific harms, such as sexual violence. Such a dual approach can help students cope with adversity, enhance their well-being, and support their academic success.

Abbreviations

SRGBV

School-Related Gender-Based Violence

AGILE

Adolescent Girls’ Initiative for Learning and Empowerment

LMIC

Low - and Middle-Income Countries

VACS

Violence Against Children Survey

MSPSS

Multidimensional Scale of Perceived Social Support

RSES

Rosenberg Self-Esteem Scale

ICC

Intraclass Correlation Coefficient

Authors’ contributions

Both authors contributed equally to the study.

Funding

No funding was received for this study.

Data availability

The dataset used in this study is hosted on the World Bank’s Microdata Library at [https://microdata.worldbank.org//catalog/6239](https:/microdata.worldbank.org/catalog/6239) and carries the reference identifier NGA_2023_AGILE-IE_v01_M.

Declarations

Ethics approval and consent to participate

The original AGILE baseline evaluation was approved by the Nigerian Health Research Ethics Committee (NHREC). Informed consent and assent for those younger than 18 (with parental consent) were obtained from all students. This current study requires no further ethical approval because it used de-identified publicly available data.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.